Chat:World/2021-05-15
CouscousAgha: good luck!
ZarthaxX: nice!
ZarthaxX: doing opp prediction clock?
ClockSort: yes i treat the opponent as a first-class player
ClockSort: and the end of 90ms i can play both sides of the board :P
ZarthaxX: what?
ZarthaxX: first class player?
ZarthaxX: im confused lol
NinjaDoggy: yea first class you know
NinjaDoggy: like first class flight or something :P
ClockSort: you know, like, front of the airplane
NinjaDoggy: LOL
ZarthaxX: :rofl:
ZarthaxX: too late for this talk
ClockSort: what i mean is, i dont take any shortcuts with player 2. so they get a full search just like player 1
ZarthaxX: but is this simultaneous?
ClockSort: yes
ClockSort: @ninjadoggy it's thanks to you that i pushed through and built the full mcts :thumpsup:
ClockSort: hmm, that should not be thumps.
CouscousAgha: hey ClockSort give me back my points
CouscousAgha: XD
ClockSort: if i hit legend now then i'll be fresh for Google Code Jam tomorrow :D
CouscousAgha: how did i lose with 1 trees
CouscousAgha: LOL
ZarthaxX: how do you make it simultaneous
ZarthaxX: damn
dbdr: maybe just the rollout part is simultaneous?
ClockSort: i used a weird hack
ZarthaxX: wanna know in pm :)
ClockSort: i consider all my moves and all opponent's moves, pick the best of each using standard UCT, then apply both at once according to game rules, which reaches a new state.
dbdr: ok, this is DUCT then
ClockSort: i didn't know the name :)
Zenoscave: Eh GCJ is tomoroow?
dbdr: it means if each player has 10 moves, you have 100 children, right?
ClockSort: google shows me this paper, though: http://mlanctot.info/files/papers/sm-tron-bnaic2013.pdf
UndercoverToad: codejam today (in my timezone at least)
UndercoverToad: in 10h
ClockSort: @dbdr yes in theory, but in practice i fill the children on demand, so they don't all get created because some moves are bad enough to not be picked enough to be fully explored
ClockSort: and i prune a lot of children before starting
dbdr: right
ClockSort: is this standard DUCT?
ClockSort: lol wow the paper i linked basically explains what i do :P
ClockSort: Decoupled UCT
dbdr: yes, I think it is
dbdr: it's the obvious extension of MCTS to simultaneous, I think
ClockSort: i should get points for inventing it, but i see it's section 4.2 here: http://mlanctot.info/files/papers/sm-tron-bnaic2013.pdf
dbdr: we have a new MSmits
dbdr: reinventing existing algos :)
ClockSort: obvious in hindsight, this is the first MCTS so I've been learning/researchinga ton
dbdr: I forgot what Smitsimax was called
ClockSort: clearly not researching enough ha
ClockSort: btw @dbdr congrats on #2 Rust, hoping you take the top spot by end of contest! :thumbsup:
UndercoverToad: at the moment c# seems easier to win than rust
dbdr: thanks ClockSort
dbdr: I'm happy if another rust is high :)
ClockSort: i love rust too, i learned it just because it was cool
iggy12345: I'll be happy if I can get out of bronze
dbdr: I already got the language achievement, in Java and in Rust
ClockSort: one of my favorite cryptocurrencies (Grin) is written in Rust
iggy12345: wassup guys
dbdr: not ready to use Rust in contest yet ClockSort?
Numby: did someone say you could get the turn input from replays?
dbdr: Numby if you print it yourself
Numby: Ah
ClockSort: in fall 2020 contest i used C# because i didn't even know C++ yet. Now I've learned C++ enough to get to Round 2 Code Jam and ~1750 Codeforces rating. Not ready to switch to Rust yet :D
iggy12345: psh, why would you switch to rust when you can use C?
iggy12345: it's obviously superior
UndercoverToad: C is not C++
iggy12345: yah
UndercoverToad: and Rust has similar speed
iggy12345: because C is better
CouscousAgha: debatable
dbdr: don't feed the troll ;)
iggy12345: woah woah woah
iggy12345: I am not a troll
MMmm____MONKE: shit, what time is it for you guys?
UndercoverToad: 6:20am
CouscousAgha: 4:20am
ClockSort: i'm in USA, it's not too late
iggy12345: 11:20 pm
dbdr: time is -02D 03H 39MN 53SC
dbdr: no need for timezones
iggy12345: how the heck are you 2 days behind us
ClockSort: it's :evergreen_tree: :deciduous_tree: time
UndercoverToad: contest end
iggy12345: oh
iggy12345: duh
abenner: anything I can do about rust compiles timing out, I swear my code isn't that degenerate
CouscousAgha: btw what will happen when timer hits 0
MMmm____MONKE: dam 6;20am and 4:20 am
dbdr: there is even a time change twice a year, like in reality :)
CouscousAgha: let's say if someone submitted their code 10 sec before the timer hits 0
CouscousAgha: what would happen
iggy12345: well
jrke: submission will complete
UndercoverToad: submits and promotion will finish
CouscousAgha: so they gonna be stuck at the bottom of the rank?
jrke: you can submit even 1 sec before ending of contest it will be completed
jrke: no not stuck after ending of contest all pending submissions are completed
dbdr: I resubbed just before promotion to legend. started to get matches in gold, then in legend
jrke: then there mega rerun for legend players
dbdr: I wonder if that other submit also continued in parallel :)
UndercoverToad: only top100 in legend get a rerun
Zaphus: anyone here using a brutaltester on windows? just grabbed one, struggling with it (Zenoscave)
NotSureWhyThisWorks: Yeah it works for me
Zaphus: it half works - it runs, but complains about negative scores - and never kills my .exes so I have dozens of them running
NotSureWhyThisWorks: hmm I was using it mainly for python so it might be different
Zaphus: might give it a run with python starters to see if that is different
Counterbalance: bots probably time out
dbdr: Zaphus make sure not to print too much to stderr
Lysk: Zaphus I was having the same problem, I don't know how to fix it but a if the bot is slow enough everything works...
Lysk: some folks recommended me to use cin.good() to check for valid inputs and kill the bot otherwise
Counterbalance: my fork with the -timeout 2000 didn't help?
Lysk: not for me
Lysk: even with 2018 brutal tester :p
Lysk: so, it's weird but I have a piece of code that tries to predict opp move
Lysk: if I run it then brutal tester works
Lysk: if I remove it then the problem with threads not dying
Counterbalance: hmm one thing I just though of.. shouldn't make a difference, but I use g++ on cygwin for the bots
Counterbalance: though i used to feed it VS binaries too.. as long as they're console apps that should work too
UndercoverToad: what's that timeout 2000 fork?
Counterbalance: just the cli branch of my fork.. it has more options
UndercoverToad: the SDK has some constraints, did you get around them?
Counterbalance: yes, but for timeouts of 2s it's not necessary
Counterbalance: i have a patch for the timeout and for excessive stderr
Lysk: what is bugging me is that it's the first time this is happening (same gcc and java version as in fall challenge)
UndercoverToad: by default the SDK allows a total of 30s, you confuse me
UndercoverToad: that's 15 bot interactions
UndercoverToad: there is a timelimit for the first turn (default=1s) which can be extended and doesn't even count for the quota if I read the code correctly
UndercoverToad: and the per-turn timeout
ClockSort: i'm cutting too early, and by day 20 i have nothing and opponent has 8 large trees
Counterbalance: here's the patch from the fall challenge, it still applies on master: https://github.com/Counterbalance/FallChallenge2020/blob/cli/engine-timeout.patch
UndercoverToad: I see, you modified the SDK part too
Counterbalance: ahhh crap the -timeout wasn't in my branch, but yes, the sdk modification is needed
Counterbalance: that explains why it didn't barf all over the place
Counterbalance: anyway i coded my own referee that's runs 1000 games in 10s instead of 20 minutes (the difference is less with a search bot, but losing the 20 minute overhead is nice)
UndercoverToad: i did that last contest
UndercoverToad: this game feels to much rock paper scissors for me to trust batches
UndercoverToad: that's how Maug s reached 3rd: run an offline batch and then accidentally submit the wrong file
Counterbalance: wrong as in not the batch for nr 1?
UndercoverToad: the bot that was losing offline
SirLoxley: Need some help with the offline Game ... I want to run it from command line but it keeps telling me "Directory src/main/resources/view/assets not found."
SirLoxley: mvn exec:java -Dexec.mainClass=Spring2021
this is how I run it (from the root dir)
Counterbalance: cp -r src/main/resources/view/assets src/main/resources/view/assets/
Counterbalance: mvn package
Zenoscave: I fail to see the diff in the dirs
SirLoxley: cp: cannot copy a directory, 'src/main/resources/view/assets', into itself, 'src/main/resources/view/assets/assets'
Zenoscave: is it casing?
iggy12345: does the directory exist?
iggy12345: maybe you need to make assets/assets
Counterbalance: lame
Counterbalance: well,checkl if .../assets/assets/ has files.. if so, then it's okay
iggy12345: oh, but you are trying to copy a file directory into the same directory
Kellthazar: Im running from Intellij Idea just fine...
iggy12345: that seems redundant
Counterbalance: see https://github.com/Counterbalance/SpringChallenge2021/commit/940b5926c4c12595c87c197eabd2bc6d190da73d
Counterbalance: that's a slighly nicer workaround
Sanguigno: ca
SirLoxley: let me take that
SirLoxley: Ok, have to give up for the moment. Thanks though
Counterbalance: np. That copy command should fix it though - duplicate all files in ../view/assets/ to ..../view/assets/assets/
SirLoxley: still get nothing in localhost / test.html
SirLoxley: that's the actual problem
SirLoxley: anyway have to run, hiking day
Counterbalance: have fun
NotSureWhyThisWorks: could someone give some tips on how to do move generation for a search algorithm when there are a variable number of turns between days. I've been stuck on this for a while and I can't find a way to do it efficiently. I have all the specifics (calculating seed options, grow options, etc) but no glue
Whiskee: I personally don't simulate days but turns, and WAIT is a turn just like the others since it can be at any point in the chain
NotSureWhyThisWorks: I tried that first but I found comparing the evaluations didn't make as much sense
XeroOl: yeah, you need really good heuristics for that to work
XeroOl: that can compare competently across days
Whiskee: WAIT turns are those that update shadows, energy, etc. while all turns have a potential value of future gains
Whiskee: if you track what you can generate from your trees, then the day itself doesn't matter
XeroOl: yeah, it just means I'm going to have to write a heuristic that quantifies potential gains
Whiskee: it does
NotSureWhyThisWorks: and finding all the next possible turns from each turn is a lot of redundant calculation. But I don't know of a clean way of fixing this yet
XeroOl: ^ that's a good point
XeroOl: right now what I have so far is pretty wasteful
XeroOl: I'm at the point where I'd just be happy to get my sim done before the competition is over
NotSureWhyThisWorks: mine just keeps timing out lol
Whiskee: again, it isn't the only approach, but there are techniques to reduce redundancy if you go that route :)
Whiskee: I create a hash that is unique for any combination of actions with the same targets happening within the same day
UndercoverToad: is it just my impression or are submits slower than usual?
Whiskee: yup
jrke: yeah slower
jacek: happy Caturday
aCat: happy :-)
kovi: submits are definitely slow
UndercoverToad: congrats for legend aCat
aCat: thanks
aCat: I feel weird given waht I have inside
aCat: just optimized heavy pruned search
aCat: (but still got ideas to improve)
aCat: on actually one-element heuristic
UndercoverToad: the boss is weak, legend isn't what we are used to call legend
aCat: I know
aCat: and actualy that's make me feel weird even more
aCat: I mean - 2 days of making lagend of 30 people legend of 100
aCat: is hmmm some achivement
aCat: and this is...
aCat: waht - 400 people at the end?
UndercoverToad: not if submits keep that speed
aCat: It actually makes me loose motivation than other way oarund
aCat: *around
aCat: maybe because having a boss is a clear goal to optimize
aCat: I'm not so used to be in top places and just to optimize agains a bunch of other lplayers and their meta :thinking_face:
UndercoverToad: don't worry, you are not in the top places yet ;)
aCat: euler these ARE problems of the first world
aCat: were talking 100/6600
UndercoverToad: that's top 5% but not top 1%
aCat: I'm the memeber of the 5%
aCat: ^^'
aCat: don't worry I'll be working on this
aCat: but today got deadline for the COG reviews
aCat: 6 papers ;p
UndercoverToad: does that mean you have to work on them?
aCat: read, review
aCat: some I laready read
aCat: one is really interesting
aCat: 2 medium, others meh
Westicles: Interesting, looks like terry u16 prints his odds of winning every move
Marchete: noob question, beam search is per day (with N turns packed in the same Beam node)?
Zenoscave: I don't think many do that
jrke: zenoscave are you still heuristic
Zenoscave: no
Zenoscave: haven't been for days
aCat: from day 4 he see > 0.8 against me
aCat: what the predictions he has :D
aCat: must be great
aCat: Marchete you have beam for every action depth
Marchete: One question on top players
Marchete: https://www.codingame.com/replay/555729619
aCat: but its better (at least huge difference for me) to synchronize
Marchete: frame 24
Marchete: why grow 25 instead of 4?
Marchete: there are 3 size 2 trees, and one of them seems more suitable, more centered on better richness, without nearby shadows
Zenoscave: sun in next turns increases due to spookyness i guess. not sure though
Marchete: there aren't directional shadows
Marchete: at all
Zenoscave: hm. in the next 6 turns?
Marchete: 25 even has a friendly shadow
Zenoscave: true. Not ure then
Marchete: Maybe it's because at N depth that decision was irrelevant
Zenoscave: who knows
Marchete: yeah, but as I'm a little retarded I wanted to know
dbdr: rust #1, gg terry_u16! :)
Kellthazar: Gratz!
kovi: testing through submit is hard when submits take 2hours
aCat: yeah, timings are awful :(
Zenoscave: It's even worse when your code times out in IDE often
Zenoscave: but not in submit
aCat: true - I'm still not suer about my timelimit settings - keeping them low for safety but thats clearly suboptimal
Zenoscave: Yeah it's hard to guage currently
UndercoverToad: i have these IDE timeouts too
UndercoverToad: with time limit set to 80ms
Zenoscave: I usually take 1.5x in IDE
Zenoscave: so try 60ms for IDE
Zenoscave: but of course you can't guage it's effectiveness in submit at that point
dbdr: 60 or 80 ms might have less impact than you think
Zenoscave: impact in what way?
dbdr: playing strength
abenner: dbdr is there any magic rust macro for performance like the c++ pragmas
UndercoverToad: rust is compiled in release mode on CG, no need
dbdr: :+1:
jacek: :unamused:
abenner: oh cool is target-cpu=native also part of the rust flags?
dbdr: I don't think so
kovi: is it just me or there are time jumps nowadays?
dbdr: abenner you can use fn annotations like: #[cfg_attr(target_arch = "x86_64", target_feature(enable = "popcnt"))]
dbdr: time jumps?
kovi: hmm, maybe sort is slow
tarriel: #[target_feature(..)]` can only be applied to `unsafe` functions\nsee issue #69098
ANJACE_Z: how to raise ur rank?
abenner: yes you can wrap safe functions around them though so it's not so bad.
geppoz: any clue why that referee for me, it works on my centos VM, but doesn't on windows linux subsystem (ubuntu) ?
abenner: I am running it on wsl
abenner: maybe check your java install?
geppoz: well it seems always all bots go in timeout at first move
geppoz: so it seems not related to java installation, they starts correctly (and I also see some of their stderr before they die)
ddreams: dbdr was it you that asked about the missing sprites locally?
ddreams: anyway, in case anyone else has the same problem I figured it out
ddreams: the assets need to be under assets/assets/, not directly under assets
ddreams: geppoz if there is any output from the referee (like warnings) it will stop
geppoz: ty, will check
geppoz: but I mean, not the brutaltester, just this: java -jar spring-2021-1.0-SNAPSHOT.jar -p1...-p2...
ddreams: hmm, ok
geppoz: it works if I put random move, but as soon as I put any "thinking" bot, timeouts
ddreams: timeout can also mean crash though
ddreams: which java version do you use?
ddreams: i had problems with a newer one, so installed 1.8
Michael_Howard: When you tap HISTORY on the left of the IDE and mouse-over the "About XXX hours ago", are the times in my time or CEST?
dbdr: ddreams I have ./src/main/resources/view/assets
kovi: hmm, i wonder if i resubmit now...will it be finished before final recalc?
geppoz: ddreams , that assets folder solved sprite problem for me
jrke: are submissions slow due to server or large pending submissions?
geppoz: well wtf it seems that referee is starting with full grown trees :D
geppoz: that's why my bots timeout
jrke: geppoz change league level
geppoz: where?
UndercoverToad: #!/bin/sh
java -jar cg-brutaltester-1.0.0-SNAPSHOT.jar -r "java -jar -Dleague.level=3 spring-2021-1.0-SNAPSHOT.jar" -p1 "dotnet ../Bot/SpringChallenge2021/bin/bots/arena/SpringChallenge2021.dll" -p2 "dotnet ../Bot/SpringChallenge2021/bin/bots/ide/SpringChallenge2021.dll" -n 1000 -t 4
jrke: are you talking about offline refree or brutual tester?
UndercoverToad: -Dleague.level=3 in particular
geppoz: offline refereee
jrke: gameRunner.setLeagueLevel(3);
geppoz: ty
UndercoverToad: you can do that as a command line argument too if you use Zenoscave's fork
UndercoverToad: maybe even without, didn't test
geppoz: I am using that
geppoz: seems it is needed then
dbdr: geppoz where did you copy the assets folder?
jrke: can you gimme link for zeno's form
geppoz: just i created another assets
Whatar: I had to remove the other two league from inside the source code even using the fork
geppoz: assets/assets
ddreams: dbdr yep, but in ...\Local\Temp\codingame\assets
UndercoverToad: https://github.com/LSmith-Zenoscave/SpringChallenge2021
ddreams: so you have to repeat assets :p
geppoz: all inside main\resources\view\assets\ , in main\resources\view\assets\assets\
ddreams: ^
geppoz: ty ddreams
UndercoverToad: but why does it find the background but not the other assets?
ddreams: finally local referee is useful
ddreams: I think it's the sprite library
dbdr: it is weird indeed
dbdr: you need to rebuild the jar after that?
kellbooby: hey
geppoz: y
jacek: https://staticfishki.iplsc.com/fishki_prod_2013_02_19/77887ff9b3b374b88075dfdae18545e0.jpg
jrke: i have modified referee a bit to play as many matches i can
Maxboyer: Hey guys, i'm trying to code in Groovy and most of the time, i get timeout, but my code is quite simple ... is it something to be expected?
ToshiTuringMachine: hi all. Did we have R language before? Now I don't see it anymore....
jacek: never?
Zandy156: does (-1 % 6) = -1 in C# ?
ToshiTuringMachine: what a pity. wanted to learn some R
jacek: yes
SabertheLost: is the battle arena superslow?
jacek: yes
jacek: friday deployment :heart:
jrke: 310 results found for «Score - Computing»
jrke: :fearful:
jrke: looks like time for a break
elderlybeginner: I think they are computing battles with paper and pen now.
Xblue707: hi
ToshiTuringMachine: wow good job devs preparing starter kits for all languages
ToshiTuringMachine: but the server looks overwhelmed
timbersson: Hey
timbersson: Could anyone help me get rid of the the following message when I run the current challenge in brutaltester...
timbersson: "10:21:07,166 ERROR [com.magusgeek.brutaltester.GameThread] ERROR StatusLogger No log4j2 configuration file found. Using default configuration: logging only errors to the console."
timbersson: It looks like I need to configure logging somehow... but how?
timbersson: I don't really know Java :disappointed:
amzh: I have similar messages, but simulations still work fine
geppoz: someone answered yesterday
geppoz: you need to put an xml somewhere
timbersson: can I find that answer somehow?
actualCoderTrevor: timbersson I have no idea if this is right but I saved a tab earlier and I think someone said you need to edit pom.xml with these changes: http://chat.codingame.com/pastebin/2b78c7e0-87fa-4f0f-a90a-1c77f74307d3 but I haven't done it myself so I may be way wrong
timbersson: thanks trevor, I'll give that a try
actualCoderTrevor: You're welcome let me know if it works
geppoz: log4j2.xml in \src\main\resources
actualCoderTrevor: Oh yeah it could be that file
geppoz: yes it is
timbersson: Yup that got it. Awesome. Thanks both
geppoz: np, has same problem yesterday, someone helped me ;)
geppoz: *I had
Westicles: for the groovy guy, it tends to time out unless you use: import groovy.transform.CompileStatic;
UndercoverToad: ToshiTuringMachine 11:20AM wow good job devs preparing starter kits for all languages but the server looks overwhelmed
UndercoverToad: most starters were shared by the community
UndercoverToad: staff is informed about slow servers
Zanoshky: do you guys know the github of a project where you specify which values needs to be changed to find best version of your algorithm
ddreams: timbersson put that into a log4j2.xml file and start the referee (inside the -r part) with -Dlog4j.configurationFile=log4j2.xml
ddreams: oh
ddreams: I was scrolled way back
UndercoverToad: you can try https://github.com/eulerscheZahl/ParameterFiddler not sure how well it does for this game
Zanoshky: thank you Toad
ddreams: that looks interesting
kovi: those cgbench spammers...
UndercoverToad: they can't anymore because of new play constraints
kovi: i know...but what other excuse can we come up for such slowness
ddreams: the parameterfiddler doesn't take high dpi into account :D
Nerchio: it seems like the problems come up at similar time every event lol
UndercoverToad: we have a :eyes: reaction from Thibaud already
dbdr: \o/
UndercoverToad: the param fiddler has a command line version too if you prefer that ddreams
ddreams: Thanks, I'll try that too
ddreams: can read it, it's just tiny
dbdr: how slow are the submits?
UndercoverToad: yes
UndercoverToad: 20% in 1h
dbdr: :scream:
DomiKo: 15% in 1H :D
Westicles: another memory leak?
Nerchio: Domiko simulation takes computation from all the servers
Nerchio: :thinking:
CouscousAgha: those mct bots are mining the servers
geppoz: while waiting for servers... question about montecarlo
geppoz: how should be choosen the random move, I mean
UndercoverToad: randomly?
UndercoverToad: you can try to make some more likely than others
geppoz: if i have 50 possible SEEDS, 1 COMPLETE and 1 WAIT
UndercoverToad: was about to say that
geppoz: WAIT should be choosen with 1/52 prob?
UndercoverToad: that's up to you but a higher WAIT probability will most likely give better results
geppoz: it is very different to chose 1/3 from SEED, WAIT, COMPLETE
geppoz: then choose a random SEED
UndercoverToad: you can try both and see what gives better results
geppoz: well
geppoz: problem is that is very different implementation
geppoz: in first case I had to count every single SEED possibility
geppoz: before to choose
MSmits: you can also random in phases.
MSmits: random a grow3 or wait, random a grow2 or wait, etc. until random a seed or wait
MSmits: no need to consider all actions every time
MSmits: it's messy and branchy
geppoz: yes, that's the point
geppoz: you are sayng the theory should work fine in that way?
geppoz: choosing 50%-50% consecutively ?
MSmits: it could work fine, it needs to be tested. Make a parameter for it :)
MSmits: but a lot of domain knowledge is important
MSmits: put in the bot what you know of the game, to lighten the load
MSmits: for example, you know with growth, it's always better to grow3 before grow2
MSmits: so why ever allow your bot to do the reverse?
geppoz: yes that is mathematically
geppoz: but other things not
geppoz: in theory SEED can be better as first in some case
geppoz: rare case
MSmits: yes
MSmits: but how about this
MSmits: consider it as an option for the first action in your search
MSmits: and then nowhere else in your search
MSmits: so it can be chosen as a move
MSmits: still tons of pruning done there
geppoz: uhm in the rollout shoudn't you go pure random to have an "impartial" evaluation?
MSmits: pure random is not the same as impartial, unless thats some kind of precisely defined technical term
MSmits: the idea of the rollout is to make it realistic
MSmits: realistic playouts are important, because they make you converge faster
MSmits: pure random is almost never realistic
MSmits: of course you take some risk by pruning in the random rollout, but as you get deeper, you can correct it
MSmits: btw, i feel like we're talking mcts now
jacek: hm?
MSmits: ?
dbdr: how can he be #1? https://www.codingame.com/replay/555809312
Whatar: maybe he throws the game when he can predicts a loss
dbdr: yeah, was thinking the same
jacek: or crashing after some date
jacek: planned obsolescence
dbdr: also not collecting points when he's winning anyways https://www.codingame.com/replay/555809988
NotSureWhyThisWorks: how do they even measure the probability of winning? Is it with mcts?
dbdr: possibly
MSmits: you can check how many nutrients are on the board and how many trees left maybe. It's a reasonable way to estimate.
MSmits: but i think it's silly to throw the game
CouscousAgha: are these submissions time normal at this time of the day ?
CouscousAgha: or is it because of the weekend
NotSureWhyThisWorks: maybe I will print my evaluation score compared to opponents to pretend like I am doing the same
actualCoderTrevor: It's an unorthodox strategy to be sure
actualCoderTrevor: Maybe it's to keep you from optimizing against the bot? Or just to keep you guessing lol
NotSureWhyThisWorks: print("WAIT", "simulations= " + r.randint(100000, 400000))
Westicles: In a bunch of his losses he thinks he is at 0.99 then drops suddenly to 0 and sort of gives up
NotSureWhyThisWorks: interesting
kovi: mcts solver tend to do that i was thinking about when doing breakthrough
MSmits: solver seems painful to write for a simultaneous mcts
ddreams: UndercoverToad needed to add "rank: 1 0" to the end of the brutaltester end of game output
ddreams: to make the fiddler work
Fubuchi: arena is too slow now, maybe it is because of weekend
MSmits: mmh i named my bot exe "0.exe" and now my windows defender thingy thinks it's a trojan horse
Astrobytes: lol
MSmits: crap it got deleted
MSmits: evil windows
Astrobytes: :rofl:
MSmits: I was teaching the other day and my students have to learn html/css so they're making a webpage
MSmits: this guy wanted to mail his files so he could work at home
MSmits: so i said zip it and mail it
MSmits: apparently we have a process running that deletes all zipfiles automatically
MSmits: :(
MSmits: I can understand doing this for unauthorized exe files, but come on
Astrobytes: hahaha, classic workplace issue
ddreams: so many companies now that are sending emails with stuff like " MailScanner has detected a possible fraud attempt from "eur02.safelinks.protection.outlook.com" claiming to be"
MSmits: yup
ddreams: and it's a link to their own site
Astrobytes: Last office I worked in, we had different levels of user with different privileges. That occasionally does not go well.
MSmits: yeah my zip files dont get deleted either
MSmits: it's just for students
UndercoverToad: right ddreams, the Param Fiddler assumes an older Brutaltester protocol from before the SDK
MSmits: but my exe files do
MSmits: i remember trying to play a game at work, copied it from USB and then it just deleted the exe file
ddreams: I had already modified that brutaltester line to include the scores from the game, so was easy
Astrobytes: Just rename it to something else
MSmits: if i could start it in 100 ms or something i could still play the game :P
ddreams: couldn't you run it from the usb drive?
ddreams: or would it delete it from there too :o
MSmits: i seem to remember that worked too, but a bit slow
MSmits: that would be really evil
MSmits: deleting from people's USB
UndercoverToad: evil doesn't help, I need eval
ddreams: to keep you safe
MSmits: same
ddreams: try an evil eval
Astrobytes: Assuming ownership of any and all peripherals attached to the system
Astrobytes: I thought you gave up euler?
MSmits: he's still submitting, so i guess not
UndercoverToad: just a few tiny modifications, nothing big
UndercoverToad: codejam in 3h
Astrobytes: ah ok
MSmits: have fun UndercoverToad
ddreams: gl
**UndercoverToad is nervous
Astrobytes: lol, there are submits still going from 8am
ddreams: I rejoined the fun too
MSmits: local testing seems the way to go :)
UndercoverToad: 9am my timezone
UndercoverToad: Thibaud already did :eyes:
Astrobytes: yeah I saw
ddreams: first 10 battles in no time, then nothing
MSmits: Yeah I am guessing there's a priority queue; IDE battles -> first 10 -> rest
natofp: Im trying to run brutaltester locally on windows, I did almost everything but when i launch it i get error: https://pastebin.com/86N5vtLk
natofp: May anyone take a look please and tell me whats wrong?
ddreams: too new java
ddreams: install 1.8
ddreams: or java8
MSmits: natofp i am assuming you're playing ghost in the cell?
MSmits: cg-referee-ghost-in-the-cell.jar
MSmits: because if were trying to do this for the contest, then you have the wrong referee
Michael_Howard: UndercoverToad have you eaten the UndercoverMice whose running wheels power the servers?
ddreams: you probably want this for this contest: https://github.com/dreignier/SpringChallenge2021
natofp: I put referee from here there: https://github.com/LSmith-Zenoscave/SpringChallenge2021
ddreams: and you renamed it cg-referee-ghost-in-the-cell.jar ?
UndercoverToad: no Michael_Howard. and there must be a dead mouse under my roof, the room smells horribly
UndercoverToad: i wish that was a bad joke
MSmits: UndercoverToad tell your cat to eat his dinner
ddreams: ok, well, the other problem is that your java is too new
UndercoverToad: her*
MSmits: UndercoverToad tell your cat to eat her dinner
natofp: ok so i will try to reinstall java and check after this
natofp: so it should be java 1.8 right?
ddreams: yep
Astrobytes: My menacing spider has migrated to the other side of the room overnight.
Michael_Howard: UndercoverToad probably from their diesel powered soviet-era server machines.
natofp: thanks a lot i will try this out
MSmits: then you won the staring contest Astrobytes
Astrobytes: Evidently!
ddreams: ten minutes for game 11, and I lost it
UndercoverToad: i'm climbing some ranks. mostly because stronger players are in submit
MSmits: Yeah weekends always give you the feeling you're ranked higher than you are, because everyone is submitting
MSmits: feeds disappointment in the end result
UndercoverToad: i'm used to it already
MSmits: to disappointment?
UndercoverToad: yes
UndercoverToad: probably no tshirt for me this time. but there's still codejam to save the day
MSmits: well you put the bar very high. I am very impressed with your succes
MSmits: I could not get a beamsearch working that well
Nerchio: C++ and rust everywhere again ? :P
UndercoverToad: you still have a day
MSmits: yeah and i havent given up, we'll see
ddreams: two days even
UndercoverToad: my submit is so slow I have a progress circle but not even an arrow to indicate the score change anymore
therealbeef: my submit just stopped at 14%, not even starting the next battle
UndercoverToad: [CG]Thibaud — Today at 1:02 PM I've pinged the devs. No news so far
Murat_Eroglu: How do you implement MCTS here? it is not a real turn based game, is it? Can anyone help me in any direction, so I can learn
Nerchio: from response on discord i'd assume the fix will take some time
FrancoisB: submitted about 2h ago, before dinner, it's now at 70%
UndercoverToad: that's all we can say. there is an issue with submits. it's reported, now we can only wait
therealbeef: ah, thanks toad
Michael_Howard: It still seems to be moving, just very slowly.
Michael_Howard: As more submits come in but none get finished, it will keep getting slower.
UndercoverToad: we've seen such cases before. it will get slower and slower, then someone reboots the server or does some other magic
Astrobytes: *download more RAM
UndercoverToad: :smirk:
Numby: When two players complete a tree in the same turn how is it determined who completes first? (this matter cuz nutrients)
Nerchio: its in the rules
Nerchio: both get same nutrient and its -2
CouscousAgha: hopefully the devs would be kind enough to fix it on a weekend
cegprakash: Numby both can't place the seed
cegprakash: you'll be refunded the seed cost but u can't use that tree on the day after
cegprakash: it'll become dormant
cegprakash: Nerchio he asks because he couldn't find it
cegprakash: coz it's sometimes difficult to search
Nerchio: yeah but im not sure which question you answered
ddreams: just some random info :D
Nerchio: he asked about completion not seeding
UndercoverToad: you can seed in a range that's equal to the size of the tree
Nerchio: you get income for trees not in shade
cegprakash: I see.. Numby both get same nutrients
UndercoverToad: because they aren't spooked
Nerchio: they aren't spooked because they are not in shade
UndercoverToad: - Keep it clear and consise - Avoid flavour text https://www.codingame.com/playgrounds/40701/contribute---help/contribution-guidelines
UndercoverToad: someone not listening to the own guidelines?
Nerchio: its weird they want to force their goals on me.. "End the game with more points than your opponent."
Nerchio: what if i just want to plant trees?
UndercoverToad: all I want is a nice pattern
UndercoverToad: but the opponent keeps placing at bad spots
FrancoisB: should a coop game really!
kovi: i like that idea
Sanguigno: One plants and the other chop?
kovi: anyway...i think now they changed the game rules...all we can do is wait...
actualCoderTrevor: Bots that play together on a team of two versus another team of two would open up so many doors for complexity... Could be cool? :thinking:
Nerchio: would certainly be nice to blame the other bot and not mine for losing
Nerchio: :relieved:
UndercoverToad: we had code a la mode for this
MSmits: actualCoderTrevor try code a la mode, it's not 2 vs 2, but cooperative
RoboStac: code a la mode worked like that. There was a lot of angry bot blaming
UndercoverToad: smits liked my bot being aware of the overall situation
actualCoderTrevor: Oh cool I'll check that one out. I'm still pretty new to CG
MSmits: did you accept my plates UndercoverToad?
UndercoverToad: every time according to you
MSmits: ok, i wasnt sure, just that many people didnt :rage:
Nerchio: i didn't play that one
Nerchio: sounds fun :D
UndercoverToad: 3 player game, 1 has a break
UndercoverToad: so you have 2 players in a kitchen, preparing dishes to collect points
UndercoverToad: each bot gets 2 out of 3 total games, scores for those is added
actualCoderTrevor: Hey this looks like overcooked with bots. I like it.
UndercoverToad: it is heavily inspired by that game
kovi: fireworks is also coop
MSmits: full heuristic bot won contest
UndercoverToad: we also have a bomberman (HyperSonic)
MSmits: that's a fun game to code for, but a real timesink
UndercoverToad: i got a chance to see the winner's code at CalM. storing states as strings internally
MSmits: you get hooked
MSmits: funny UndercoverToad :)
MSmits: I guess it doesnt matter
UndercoverToad: no. just wasn't a good read
MSmits: did he/she do this because of inexperience?
MSmits: heuristic bots dont require a lot of coding expertise, but are mostly about understanding the game
UndercoverToad: no idea. I actually consider him a skilled coder
MSmits: ah ok, that answers that then
UndercoverToad: also strong on RAIC
Astrobytes: who won calm?
UndercoverToad: morozec
MSmits: I haven't lost a single game since my submit started half an hour ago :P
Astrobytes: Ah yeah
MSmits: that's 15 games :P
UndercoverToad: so you won the first 10 smits?
UndercoverToad: oh wow
MSmits: I am around rank 10-30 with this bot, so it's not that weird
MSmits: (gold)
UndercoverToad: i know, in legend it would have impressed me
MSmits: yeah
MSmits: the main reason i want to get legend now is the huge boost in ranking
MSmits: i will probably get top half there
ddreams: I realized I never submitted my best bot
UndercoverToad: that sentence could be from struct
Noyotens: anyone in legend
ddreams: 4-1 vs MSmits
UndercoverToad: i'll take a nap before codejam
ddreams: Noyotens stop asking "anyone typescript/legend/whatever" just ask your question
MSmits: 4- 1?
ddreams: played in the ide four times
MSmits: ohh ok
UndercoverToad: five
ddreams: right, counting
MSmits: he already forgot his loss
MSmits: selective memory
ddreams: hehe
ddreams: uhh one from the last battles :p
Noyotens: ok anyone have replay to how beat Miyaza
ddreams: look at the top gold for that
MSmits: you wont be able to see why they win though
MSmits: there isnt a specific trick to beating boss
MSmits: it's just... playing well
ddreams: it's also not enough to just beat the boss
Aweorih: anyone knows how u can run multiple "MultiplayerGameRunner" if u run ur bot locally in the checked out source of the game? (like running multiple games in parallel)
ddreams: you also need to beat other gold opponents
ddreams: Aweorih brutaltester
Noyotens: i know
Noyotens: but my alghoritm is like :shit:
MSmits: poopsearch?
Noyotens: yeah lol
MSmits: innovative
ddreams: looking through the shitty moves and finding gold
Noyotens: https://www.codingame.com/share-replay/555829942
Noyotens: myaza vs myaza
ddreams: all the cpu is spent on boss self fights
MSmits: dont slow down my submit :P
jacek: time to promote in csb :imp:
ddreams: 15-1 so far
MSmits: lol, whyyyy
ToshiTuringMachine: 3 hours for 48 games in wood1
ToshiTuringMachine: I will never see gold
MSmits: your bot with continue promoting after deadline hits i think?
MSmits: will
MSmits: you just cant test your stuff
MSmits: on leaderboard
ToshiTuringMachine: rules are different in other leagues?
ToshiTuringMachine: :stuck_out_tongue_winking_eye:
MSmits: from bronze and up they are the same
MSmits: so get to bronze, then write a good bot
MSmits: but you can look up the rules in preparation
ddreams: you can also run it locally
TobiasA: do you have to win against a bot 100% of the time to get to the next league
ddreams: no
TobiasA: ok thanks
MSmits: just slightly over 50%, if you're near to 50% it may take multiple submits to drop rating of boss
MSmits: during which you will promote other players
Scarfield: you have to win enough so that your score is higher than the boss score at the end of submit
ddreams: lol multiple submits with this broken state
MSmits: or when you stay there, under boss, other players may push you over
ToshiTuringMachine: if you forget anything you lose 6 hours
MSmits: (only if you're near and win vs players)
derjack: is it only me or clash submits are slow?
ToshiTuringMachine: codingame has never been so slow, and I've been here since 2016
MSmits: derjack lol
MSmits: always creative with trolling :P
ddreams: this submit will take me to legend
MSmits: I hope so
ddreams: in six-seven hours
MSmits: then you can stop wrecking me
ddreams: :D
Hoyer: does anyone have good resources for tree structures in Rust? I decided to learn Rust for this challenge, but I'm really stuck on this and google hasn't been extremely helpful (outside of telling me it's a limitation of Rust)
MSmits: dbdr maybe
MSmits: though he's known this stuff a long time so not sure if he still knows where to find
ddreams: Trees are naturally occurring world objects in Rust. Players can harvest trees for Wood using any gathering tool such as a Rock, Hatchet, or Pickaxe.
Marchete: wish you luck to pass legend Kodle and MSmits :D
CoderDJD: I'm literally stuck with bronze, any context which might help?
MSmits: thanks Marchete
MSmits: i am mostly param fitting today, maybe coding a new search to experiment with, while I wait
MSmits: CoderDJD efficient seeding, try not to shade your own trees. Watch lots of games. during a turn if you grow something to size 3 and something to size 2, grow the larger first
MSmits: seed last (generally)
jacek: ^W^ https://img-9gag-fun.9cache.com/photo/aEpAXPo_700bwp.webp
MSmits: generally not needed to have more than 1 seed at a time
Kodle: haha thanks Marchete! finally managed to port my python search over to c++ for the speed up haha, seems to be doing alright fingers crossed :)
CoderDJD: Thanks MSmits, good luck with legend!
MSmits: thanks
pn2501: Hoyer you can look for arena tree https://dev.to/deciduously/no-more-tears-no-more-knots-arena-allocated-trees-in-rust-44k6
MSmits: Hoyer i am not sure how it works with Rust, but in c++ i just make a giant array and index to it
MSmits: is that not an option?
SabertheLost: My submit just beat my friend. So maybe top 500 silver.
MSmits: gj
Porsuk: I have seen people make the forest creatures say stuff
Porsuk: how do they make it?
MSmits: cout << "COMPLETE " << bestChild->actionIndex << " " << message << endl;
ddreams: add text after your command
AlMualem: print(Action,"your text")
Porsuk: like print("GROW 25 my comment")
Porsuk: ah
Porsuk: okay
Porsuk: will try, ty
AlMualem: de rien
CaptainHamburger: MSmits code leaked?:scream:
jacek: ?
dbdr: Hoyer: a performant way for trees is to represent nodes as indexes in a large Vec / Array
ddreams: 21 battles in one hour, nice
TobiasA: oh i thought it was only me
FrancoisB: Hoyer: Google Rust Arena
elderlybeginner: 52 in 3 hours :wink:
CaptainHamburger: Im winning i see: 27 battles in 2 hours
ddreams: and one of them against my previous bot
struct: hello
TobiasA: hi
ddreams: hi structure
ddreams: do you get pinged from that, btw?
struct: no
Hoyer: pn2501 MSmits dbdr I was also having trouble doing it that way, since you can't initialize an array with empty values and I couldn't get my Node struct to derive from Copy
struct: you can initialize an array with empty values
Hoyer: in rust?
dbdr: doing MCTS?
struct: ah in c++ I can
struct: GameState* Agent::states[3] = {new GameState[1000000], new GameState[1000000], new GameState[1000000]};
struct: static GameState* states[3];
elderlybeginner: how to compile contest referee (referee.java) into jar?
dbdr: Hoyer: do you want it initialized or uninitialized (unsafe but instantaneous)?
ddreams: elderlybeginner use https://github.com/dreignier/SpringChallenge2021
ddreams: but you have to use maven, e.g. mvn package
ddreams: if that's what you asked
natofp: Im running brutaltester on Windows locally and exe programs which i test stop working after the game, does anyone know what might be the issue?
cegprakash: 1 day 19hrs left.. Me Panics
ddreams: if c++ probably not checking if stream in is still good
elderlybeginner: i installed mvn, make target, now how to get referee
Hoyer: dbdr I think I could make my program work either way
cegprakash: elderlybeginner get the referee jar from someone who already built
Hoyer: Right now my focus is completing in time :sweat:
natofp: thanks a lot again ddreams
elderlybeginner: anyone eager to share referee? :wink:
cegprakash: but from I experience I think brutaltester can help u improve 50 ranks not more
dbdr: Hoyer: http://chat.codingame.com/pastebin/312ad417-7f31-4b29-b677-79dede1d13d2
dbdr: this should be very fast, but uninitialised, so up to you to initialize before use
dbdr: each node
Hoyer: hm yea makes sense... I guess you keep the node array in the main function then? I was trying something like this as a static variable, but Rust really didn't like that either
Hoyer: Lots of occasions where I need to change my way of thinking with this language
Hoyer: :grimacing:
dbdr: if you use a static you will need more unsafe ;)
CoderDJD: good luck to anyone whose tryna pass some level. Have fun.
dbdr: I store it inside a struct, which I initialize in main,before the main loop
Hoyer: that makes sense
Hoyer: alright I'll get to work, thanks for the advice everyone :)
ddreams: elderlybeginner after mvn package the referee is in the target directory
struct: damn top is so close to each other
natofp: OK i still got the problem that when testing locally programs stop working, im using exit(0) when i wait during 23-th day, shouldnt it be ok? or should i close those programs with some command?
ddreams: should be ok, so attach a debugger and see where it's stuck
elderlybeginner: ddreams am I missing something https://i.imgur.com/HrIMSpu.png
ddreams: looks like you have compiled brutaltester
elderlybeginner: and I don't see referee
ddreams: https://github.com/dreignier/SpringChallenge2021
ddreams: ^ again, there's the referee
elderlybeginner: there is uncompiled referee.java in sources, brutaltester uses jar files this is all my guessing as I have no idea about java
ddreams: you are completely ignoring what I'm saying, so...
TobiasA: this taking a long time
TobiasA: i submitted a while ago and it is still in 15%
elderlybeginner: ddreams :wink: it's not about ignoring, I'm just like a child which does not understand
ddreams: brutaltester does not contain this game, you have to download it separately and compile it
CouscousAgha: any update on submits speed?
ddreams: 24 games in two hours
ddreams: so codingame is still broken
Noyotens: y'all testing is same time
Noyotens: this broke the server
CouscousAgha: well its the last weekend for the contest so i guess the servers werent ready for this
Astrobytes: This is ... fun.
Fluxor: same here.... 25 games in 2h
Astrobytes: there are submits still going from 5.5 hours ago
CouscousAgha: 40 in 3h
Fluxor: I guess it is useless to try and deduce anything from the result, since all IA will have changed in meanwhile...
Fluxor: Well, I do use brutal tester locally, and I submit only when I have at least + 20% win after running 500 games against myself... but unfortunately it tells nothing about the impact on the ranking. I happens that improving against myself just makes things worst against other...
Fluxor: at the end, only the real arena can tell...
ddreams: same
MSmits: this is true
ddreams: though every time I've reached 70%+ i've improved ranking, but that's not in legend
ddreams: rank 30 in gold now after 25 fights, one loss
Scarfield: 596 submits atm :'(
MSmits: rank 41 gold 28 fight 3 losses
ddreams: Fluxor: do you get any significant difference in win percentage after 150 games?
aCat: Magus nice !
aCat: I see now changes, something is happening
aCat: (except very slow resubmits ofc)
elderlybeginner: brutaltester uses jar file as referee, how to compile referee.jar from https://github.com/LSmith-Zenoscave/SpringChallenge2021
struct: looking nice MSmits
Noyotens: #save_codingame_server
aCat: better use Game runner directly from this project
struct: also my promoted bot on gold that was like 0.01 away from boss
struct: is 40th legend...
MSmits: it's not enough yet, but i hope a param fit will get me there
Fluxor: Before I resubmit, I was happy to have climbed from 38th up to 24th in legend, without submitting anything... but I guess now it is just because all the top players that resubmitted were stuck playing games at the bottom, too bad :-(
jacek: you have pretty avatar
Fluxor: ahah you too :-D
CouscousAgha: lost brothers
ddreams: elderlybeginner as I said before: mvn package
ddreams: will make spring-2021-1.0-SNAPSHOT.jar
ddreams: that's the referee/game
Michael_Howard: Just for a change, a reverse roast request - what is Miyaza doing wrong in this game I (or we) should be exploiting more? https://www.codingame.com/replay/555848522
ddreams: it's your bot, right? :)
Michael_Howard: I didn't write the boss, no :upside_down:
Michael_Howard: And he normally murders me, this game is unusual.
ddreams: oh, yeah, it was Bernard_Lowe
UndercoverToad: let us know if you need another murderer
ddreams: he probably started on the wrong side of the board
ddreams: silly mistake
Michael_Howard: UndercoverToad I'd *love* to get to the place where you can murder me.
UndercoverToad: oh, you can't even select you as an opponent
UndercoverToad: maybe you are in submit?
Michael_Howard: Yeah.
UndercoverToad: in slow sneak submit
lpellieer: how do you guys evaluate your moves ?
ddreams: double eval = evaluate(g);
ZarthaxX: kodle is using c++ OH MY GOD
ZarthaxX: the python guy D:
Numby: Is it normal if I played 2k games from start board with random moves that all 2k end up as ties :S
VizGhar: with same seed completely normal :)
ddreams: it's not normal to play 2k games with random moves though :p
ddreams: jk.. you're doing mcts?
Numby: I'm not changing seed and keep calling Random.nextInt so should be fine
Numby: Yeah doing mcts or trying to lol
UndercoverToad: kodl e gave up on python? i'm genuinely surprised
Numby: I already tested by game advancing logic and it looks good. I basically replayed an entire game from a replay end to end by just selecting the action and apply it and end state matched perfectly
struct: Just wait until I submit my pascal bot
Astrobytes: Well, this is a productive day.
UndercoverToad: you don't have a Pascal bot
UndercoverToad: as a codejam player, is it bad to be happy about the slow CG platform today so I don't miss a lot?
Astrobytes: :rofl:
Astrobytes: not at all :D
elderlybeginner: is this java version issue (referee): https://imgur.com/b9TNU3S.png
ddreams: yes
ddreams: how's the code jam going?
reCurse: 639 computing o.O
DomiKo: 36% mmmm
Zanoshky: my bot beats boss 10/10 rounds, but I am still not promoted, why?
DomiKo: 64 battles in 4H :D
Astrobytes: This is almost funny.
Zanoshky: where is the codingame employess? why not spin off more was isntances?
Zanoshky: aws*
reCurse: Because it's not your money?
Astrobytes: Thibaud (community manager) pinged the devs. Heard nothing since.
CouscousAgha: because you dont pay anything
Cjx_1: glad its being looked at. This is killing me--soo close to gold
Cjx_1: just want to crank out a few more iterations lol
reCurse: Ping doesn't mean looked at
Noyotens: 2 games in houy
Noyotens: *hour
ddreams: I pay a lot because I code for the contest instead of for clients
Astrobytes: ^ what reCurse said
reCurse: Also long weekend in France so maybe people are out of reach
reCurse: Great timing
ddreams: would be happy to pay my part of the servers
Astrobytes: Perfect time to have a contest.
ZarthaxX: true
Astrobytes: Cause nothing will go wrong.
ZarthaxX: so basically results are known already
UndercoverToad: that's what you get for hosting a contest during codejam
ZarthaxX: cant beat without submit testing
ZarthaxX: what toad :rofl:
UndercoverToad: what i'm writing makes no sense
Astrobytes: How unusual :P
ZarthaxX: haha
CouscousAgha: UndercoverToad how long is codejam
UndercoverToad: 2.5h, starts in 45min
ZarthaxX: submits are literally freezed
ZarthaxX: damn
CouscousAgha: oh it's just gonna start? lol
reCurse: Not frozen just very long
UndercoverToad: but it's round 2 already, you have to be qualified from a previous round to play
CouscousAgha: i thought it's already on
ZarthaxX: reCurse sorry i meant that
UndercoverToad: i'm preparing right now (starting IDE, tasting the interactive problem setup, ...)
ZarthaxX: this is not gonna get better anytime soon
ZarthaxX: :/
ZarthaxX: popularity hurts :P
ddreams: undercovertoad how does it taste?
dbdr: don't eat the problem! bad toad!
Astrobytes: :rofl:
Cjx_1: do we not do lab saftey in cs?
Astrobytes: ZarthaxX: BUT THE NUMBERS THO'
ddreams: you can disappoint so many people at once
ZarthaxX: damn numbers astro
Astrobytes: Indeed.
ZarthaxX: highest ever excepto for the accoutnant?
Astrobytes: Donno. Haven't looked tbh.
ZarthaxX: toad has to know
natofp: did anyone have issue with programs stopping during local testing? im dying from tilt
geppoz: natofp are you sure you are launching correct league level? I had same problem launching wood league...
geppoz: runner.setLeagueLevel(3);
AlMualem: i have issues with battles :/ too long
geppoz: in CommandLineInterface.java
SabertheLost: good damn, I gave cost to grow seed the same as grow tree size 2. It almost never thougth it was ok to grow from seed
Cjx_1: the wood spirits are prepareing their counter offensive
tobk: This is getting ridiculous. Submitted 4h ago, now at 40%...
reCurse: It's going to get more ridiculous don't worry
CouscousAgha: wait until codejam starts
AlMualem: one Eternity later.....
struct: I think its time for a csb promotion
ddreams: 657 submissions calculating
ddreams: codingame has it's own code jam
Astrobytes: hah!
CouscousAgha: tfw it's been almost 4 hours and u still didnt reach the point to know if your bot is better or worse than before
JohnCM: think now need to wait a while for the rankings
tobk: also, I think rankings are pretty messed up because so many are still calculating and thus do not have their proper rank
ClockSort: i'm 8th gold and afraid to resubmit due to the backlog. is it common to drift into the next league?
ZarthaxX: ClockSort yes
UndercoverToad: "tfw it's been almost 4 hours and u still didnt reach the point to know if your bot is better or worse than before"
i'm in that state for days despite submits finishing
ZarthaxX: but submits are too slow for you to get pushed fast
JohnCM: clocksort i think you can just wait for someone to bump you up to legend
ZarthaxX: haha toad :P
elderlybeginner: when is the best time to jump up a league? when it's just open?
UndercoverToad: depends on whether CG can repair their servers ClockSort
ClockSort: hi @JohnCM <3
JohnCM: i just hope to beat this silver boss this time, haha got lazy to code this challenge
UndercoverToad: at the same time if you resubmit it won't be any faster
UndercoverToad: just wait it out
ClockSort: thanks @UndercoverToad
JohnCM: intending to do it purely heuristic-based. my silver code was like 30 lines
JohnCM: now i need to add more stuff for tree positioning to get gold
elderlybeginner: my gold code is like 30 lines still :joy:
JohnCM: wow that's advanced heuristics
elderlybeginner: +starter
UndercoverToad: you can always improve your heuristics with a simulation
daniserrano7: I got to gold for my first time ever!!!
ddreams: gratz
elderlybeginner: simulation needs to score the position, which needs to put all shadows on the board, which leads us to +50 lines at least :joy:
ClockSort: i'm at 1200 lines... keep writing :P
daniserrano7: But my code is 300 lines :sweat_smile:
UndercoverToad: around 650 lines including dead code
struct: o.o
UndercoverToad: didn't get my opponent sim to do any better than without :(
JohnCM: all the best clocksort
JohnCM: haha i actually intended to do monte carlo tree search but was too lazy to code out the game sim
UndercoverToad: i don't think that this game is a good candidate for plain MCTS
kovi: in the morning i was jking about last submit before recalc...but now...
UndercoverToad: but some top Japanese players are tweeting about DUCT
ClockSort: i'm using DUCT MCTS which is why i'm not in Legend :D
ClockSort: j/k
JohnCM: what's DUCT MCTS? got link to that
ClockSort: section 4.2 here: http://mlanctot.info/files/papers/sm-tron-bnaic2013.pdf
daniserrano7: Do you think its possible to reach legend with pure heuristics?
UndercoverToad: yes, there are some heuristic bots in legend
JohnCM: definitely, if you can code the heuristic well enough, you can reach legend
ClockSort: getting setup for google code jam now. see you in a couple hours!
ddreams: gl ClockSort
daniserrano7: cool!
ToshiTuringMachine: hi ZarthaxX how are you?
reCurse: I get the impression there's something O(n^2) going on
Wontonimo: yesterday i finished a MCTS , it performs worse than my MC-1ply bot (rank 140th), probably around 600th overall.
geppoz: MC-1ply means?
jrke: damn 52% battles in 5h
CouscousAgha: im sure someone is mining ETH with their bot
reCurse: I wouldn't be surprised if you stopped everyone submitting, then queue by batch of 100 people submitting, it would be 3x faster
Wontonimo: MC = montycarlo. 1ply means no tree search, just check all first moves geppoz
JohnCM: haha think i only get one good submit before comp ends
JohnCM: means a greedy AI per move is decent alr
struct: oh well back to brutaltester
Astrobytes: I've lost what little motivation I had
struct: good thing I got promoted yesterday
kovi: is this hurt your smurf testing recurse?
kovi: oh btw karliso is bronze...
CouscousAgha: are smurfs allowed in spring challenge ?
reCurse: If I have a smurf I'm not doing too good now am I
dbdr: reCurse, according to JBM it's the match scheduler that gets overloaded
reCurse: I think I heard CG say something like that before with their DB
reCurse: Thought they fixed something :/
dbdr: sudo reboot
dbdr: * it works now *
reCurse: It's a long weekend in France right? What are the odds everyone is unreachable?
struct: yes
Astrobytes: The total absence of any communication would support that theory
JBM: i mean, they picked this week *because* it's a long weekend
jrke: what will they do for battles now 600+ ongoing submits
JBM: it's not like it's a surprise to them
reCurse: The total absence of communication would have made CG go out of business 3 years ago, not the point
dbdr: live dangerously!
ddreams: Astrobytes no communication? I've heard several times today that Thibaud made an emoji comment this morning
Astrobytes: Yes. That's all.
JBM: i'm sure he told the devs about it, yes
UndercoverToad: he also wrote a line
ddreams: oh my
Astrobytes: Oh yes. Thanks toad.
dbdr: :D
DomiKo: Good luck in Code Jam!
UndercoverToad: IDE :ballot_box_with_check: sheet of paper :ballot_box_with_check:
DomiKo: pen?
UndercoverToad: do you play too DomiKo?
UndercoverToad: pen :ballot_box_with_check:
DomiKo: ofc
UndercoverToad: same name as here?
DomiKo: yap
UndercoverToad: oh no, cat entered the room
UndercoverToad: why now?
UndercoverToad: just go to sleep
Astrobytes: They know things euler. Always.
UndercoverToad: parked her at the window
Astrobytes: good luck with that :D
dbdr: inb4 euler's cat gets sick
ddreams: over the keyboard
UndercoverToad: 40s, see you later
ddreams: gl
dbdr: glhf
Astrobytes: best of luck to you both
Gabbek: good luck!
DomiKo: another slow servers :(
kovi: maybe too many wrote heuristic + dogecoin mining instead of real search
reCurse: Another funny theory I have is the initial 10 matches get higher priority, people wonder why submit is stuck, resubmit, repeat...
struct: They probably do
struct: There is no other explanation on why they get computed so fast
dbdr: nice theory
MSmits: reCurse +1 That's a good one
MSmits: also because they want a better start, they do it a few times
Wontonimo: the first 10 seem to be executed all in parallel, where as the rest are serial
MSmits: 32 wins, 4 losses, rank 7, at 29%... maybe it's enough for legend?
struct: probably msmits
struct: come back in 4 hours
dbdr: you need 100% for legend
MSmits: lol
dbdr: that's the hardest
kovi: at this rate...you dont need to comeback
Gabbek: I've submitted 7h ago and it's on 50%
kovi: slowdown is exponentioal
Gabbek: so I imagine you should wait about 10h :(
Noyotens: guys why i cant debugging
struct: oh god
struct: Guess I will wait until 1 am or something to submit
kovi: and +1 for recurse...i looked at red circles on leaderboard...the more submission the more slower it gets
DaNinja: chopped trees causing log jam
Noyotens: guys why i cant debugging
- please: tell me
MSmits: that emote doesnt work
ddreams: Noyotens how can anyone answer that?
struct: print to stderr
kovi: in the morning it was 2hours/100%, then 2hours/30%...now additional 5 hourse for another 30%
dbdr: kovi: 696 results found for «Score - Computing»
Noyotens: i can answer
Noyotens: but
NotSureWhyThisWorks: the number of red circles is crazy, I just checked lo
Noyotens: i cant do console.error
Marchete: maybe if we all resubmit at once it will be fixed!
ddreams: creative way to clear up the jam
Wontonimo: glad i stayed up late last night and got in a last submit. it went fast
Noyotens: still not working ;_(
Wontonimo: Marchete :joy:
ddreams: I did a speculative submit instead of letting my gold #1 bot stay
Noyotens: still not working ;_( :cry:
reCurse: Stop with the spam and useless questions
MSmits: yay another game, rank 4 now :P
CouscousAgha: no update?
MSmits: I just got 1. A game fininished!
MSmits: I'm just brutaltesting locally... hopefully that will eventually translate to something good on the leaderboard overnight
MSmits: we can at least, still code our ideas
ddreams: true, but I miss being able to test against the people in legend
MSmits: yeah
MSmits: there's probably some way to do it if you know the agent ID
MSmits: I guess CG bench can help you there
reCurse: Let's make things worse!
MSmits: no, i meant, run 1 game
MSmits: not 100 :P
reCurse: Why 1
MSmits: because IDE doesnt let you select agent id
MSmits: of legends
MSmits: when you're in gold
jrke: whats CG benchmark btw
reCurse: Ohh
jrke: i don't know about that
jrke: any link?
reCurse: That's fixable by a few lines of greasemonkey
MSmits: it's a program where you basically run IDE without browser, so you can chain games and get statistics, but its not really working now. Nor should you run it when the server is taxed like this
reCurse: Wasn't there a whole extension based on that?
MSmits: i was merely mentioning it because it allows you to put in agent ids
reCurse: CG Enhancer I think
MSmits: reCurse yeah, but it doesnt work anymore i think
reCurse: Ah
MSmits: not sure
reCurse: Sucks
Noyotens: i am gonna use netural networks
Noyotens: the AIs is coming
Fluxor: MSmits: same here, I got some nice improvements according to brutal testers results, but I hope it translate for real...
MSmits: if you cant spell it, maybe try something else :P
MSmits: I'll be quite happy just to get legend
CouscousAgha: jackpot! i got 2 games at the same time
MSmits: even though many people got in there with the so called *easy boss*. It was not easy for me at all
MSmits: CouscousAgha one game from a different players' submit
MSmits: doent count for your 100%
CouscousAgha: never felt this good since the first time i hit top10 in gold
struct: I think im going to try duct, i still should have enough time
MSmits: do some duct struct
Payalord: I can't run my contest code in IDE at wood 2 league. Is it disabled?
MSmits: put some puct in the duct as well
struct: I never did it though
struct: But I dont see my current bot going any further
Marchete: never & 2days & duct
Astrobytes: DUCT or delete
Marchete: I don't know
MSmits: struct, If you do Duct, i have an easy way to test without completely messing things up
MSmits: start at the end of the game. Duct the last few turns
MSmits: then work your way back
MSmits: keep rest of bot for earlier turns
Noyotens: i cant use AI without dubegging :cry:
MSmits: end of game is simple/easy
Payalord: So for those who is late for the contest Play My Code doesn't work anymore in wood 2 league right?
struct: ill try it, thanks
Noyotens: no i am in gold
CouscousAgha: Payalord everyone are having server issues
MSmits: Payalord the servers are overloaded
actualCoderTrevor: Plot twist: Noyotens is an AI
Payalord: Ok i see
Noyotens: no i am human
Payalord: so it works but not for everyone
actualCoderTrevor: That's just what an AI WOULD say
Payalord: lol
Noyotens: ur right :bot_face:
Noyotens: why there no emoji
MSmits: they dont get created on demand
Payalord: hack the chat and implement bot emojis lol
MSmits: :robot:
Noyotens: MSmits u also think i am AI
MSmits: maybe a bugged AI :)
Noyotens: :no_mouth:
MSmits: oh, i thought ClockSort went legend with his mcts, but he's still here
MSmits: in gold
Payalord: Detroit: become human
Noyotens: i bugged the server that's why u cant submit *bugged AI sound
Payalord: I think I'll pass this contest. Too late for me
ToshiTuringMachine: guys what's the command to make a jar from spring challenge referee?
Altaire: mvn clean install
DaNinja: mvn package
Wontonimo: no Payalord, you could get to Silver in the time remaining with just a ifelse bot
jrke: i am in gold with if else bot
Payalord: I can't even test my code man
cegprakash: Payalord
Payalord: Play My Code in IDE here doesn't work at all
alchemsti: I think you can get to bronze by just spitting back the last viable move (they give you a list) and maybe waiting on the first turn.
Wontonimo: MSmits my MCTS can't break 400th Gold.
cegprakash: sort all possible moves by target index and print the move with lowest index that takes u to gold
cegprakash: we are all in gold
cegprakash: with 1 sort u can come to gold
Payalord: I was late to the contest
struct: Wontonimo is it single player mcts?
Payalord: And no option even to test the code now in wood 2
Payalord: When i click play my code in IDE nothing output
Payalord: No info at all
Payalord: No turns nothing
cegprakash: reset code
Payalord: tried
Payalord: same
Payalord: restarted browser reloaded page
MyCCODE: Who knows how to solve rubik cube
Wontonimo: struct - yes, single player + very minimal heuristic for rollout = 600th gold
cegprakash: Payalord reset code should print "WAIT"
IamQuan: :>
IamQuan: heh...
Wontonimo: struct - i may make it 2 players tonight. meh
jrke: i think that minimax can give legend
struct: im going to try dutct yeah
struct: duct*
Payalord: Ok i have tested from firefox now and it worked finally
Wontonimo: nice
Payalord: So might be something with my chrome then
MSmits: jrke if heuristics can give legend, then any search can also
MSmits: so minimax too, yes
jrke: my bot is having patience
jrke: https://www.codingame.com/replay/555873961
jrke: lol
MSmits: Wontonimo I think mcts can work here, but in general, i find it pretty hard to configure a search (pruning/ eval) in such a way that it beats a good heuristic bot
Wontonimo: i'm not sure heuristic can get there anymore with so many new search bots crowding top of Gold
MSmits: that shouldnt matter
MSmits: if they are better than the boss, they are out
Wontonimo: Agreed MSmits, there is definitely a lot of heursitic in my MCTS
MSmits: correction, if they are better than the boss, they are out... tomorrow
Wontonimo: well, that ain't me then. my bot doesn't beat boss regularly
BrunoFelthes: omg, submits are stuck
Wontonimo: I was chasing your progress all last night MSmits :)
cegprakash: I don't know if my bot is climbing rank because it suddenly starts winning or submits are stuck for others
ddreams: yay, I found a bug in my bot, wonder if fixing it improves it :p
Astrobytes: No. By Marchete's Law it will be orders of magnitude worse.
MyCCODE: I'm having a temperature quiz where you have to write a program that will return the closest temperature to 0 but I don't know how to check if we have to do the "POSITIVE" or the "NEGATIVE" number in the output
Wontonimo: (someone isn't working on the contest)
geppoz: :D
aes-salm: I TEST IN ARENA about 30 minutes ago and BATTLES IN PROGRESS 14%, did anyone have the same problem?
ddreams: 700 people have the same problem
Wontonimo: yeah, it's overloaded now
TobiasA: i have the same problem
TobiasA: mine is 30% in 3 hours
BrunoFelthes: all contests, same problem
ddreams: 28% in 4 hours here
MSmits: 31%
ddreams: If it's the matchmaking that is the issue, CG should make a contest about it
MSmits: got my first real ladder match vs boss
ddreams: share it?
MSmits: https://www.codingame.com/share-replay/555878421
ddreams: sim perfect?
MSmits: oh yeah, every turn i check the sim vs next state of the game
MSmits: to see if it got predicted correctly
MSmits: i try all opponent actions and make a list of states
MSmits: it should be among the list
ddreams: ah
MSmits: if it's not, then my sim has a bug. I forgot to remove it after testing, but kinda like it
FrequentlyMissedDeadlines: North Korea proposed to lend their computer! it should double the compute
ZarthaxX: got a battle too MSmits!
ZarthaxX: what a great day to be alive
ZarthaxX: lol
MSmits: hehe
MSmits: hopefully my bot does well. Maybe i can pull him down a bit
MSmits: (the boss)
MSmits: oh you're legend ZarthaxX
MSmits: nice
Michael_Howard: Wanna bet it's all from one uselessly exponential line of matchmaking code that someone at Codingame wrote drunk in 2015.
ZarthaxX: oh yes got into it when legend opened, first time i didnt have to fight for it :)
ZarthaxX: you are getting there too MSmits prob
MSmits: starting to think I might yeah
MSmits: have more improvement locally *I think*
MSmits: hard to be sure
_martynas: Tried 1v1'ing 1st gold player for fun: https://www.codingame.com/replay/555883236
jacek: back. do submits work now?
JasperV: no
JasperV: 757 computing atm
ZarthaxX: thats not a sign of improvement necessary MSmits
Pick8chu: Guys
Pick8chu: What is your seeding weight algorithm? I seem to make bad choices finding a better spot to seed
ItsNotABug: 762, exclusively going up
Pick8chu: btw how do you know how many computing?
ItsNotABug: leaderboard -> score -> computing
ItsNotABug: 774 now
ZarthaxX: well, we can already call it gg and gz the top 3 right?
aniski: technically, it happened during all the recent contests, didn't it?
Pick8chu: first time so idk
ZarthaxX: hey ToshiTuringMachine!!!
Pick8chu: it seems pretty bad tho, I submitted mine like hours ago and still less then 30
ntroPi: I have a consistent timeout ... but it goes away if I add some error output? I hate Heisenbugs.
aniski: I submitted mine 5 hours ago, 32%
ItsNotABug: Yeah. At least the last 4
Numby: GUYS
Numby: I found why all my sims end the same
Numby: I forgot to undo my trees as dormant at the end of each day -_-
Pick8chu: How am i supposed to know which algorithm is better if I can't see the result :joy:
Pick8chu: wdym?
Pick8chu: you can undo your trees?
ZarthaxX: he meant in his sim
ZarthaxX: he didnt reset their state
Pick8chu: what is 'sim'
ZarthaxX: simulator
ZarthaxX: you implement a simulator of the game in your code, t
ZarthaxX: to use it for trying possible actions and
ZarthaxX: keeping the best
Pick8chu: wait
Pick8chu: how do you mean that you implement a simulator of a game in your code?
Pick8chu: your code is just to get out the best possible actions atm isn't it?
ntroPi: yes but to know what rally is the best we simulate the game
ntroPi: really
Pick8chu: how can you simulate the game?
Pick8chu: so you meant, to get the best possible action, every turns, you try every actions and see what's the best and choose that?
Pick8chu: So basically every turns you would estimate possible scores when you reach the end? is that possible?
ntroPi: no basically for every possible turn, simulate how the game would play out (assuming only simple moves until the end)
MSmits: You need a way to look at the current game state and give it a score that tells you how good this is for you
MSmits: then the simplest way to use a sim is to try each action
MSmits: then calculate the new gamestate
MSmits: evaluate it (give it a score)
MSmits: take the highest scoring one
Pick8chu: that's crazy that you can do that
ZarthaxX: are you beaming smito?
ZarthaxX: i got another game!
MSmits: this is not a very effective way to do things in this contest. for this contest you need more depth (more turns ahead)
MSmits: yes beaming
MSmits: I am brutaltesting parameters. Found some good improvement i think
ZarthaxX: nice
ZarthaxX: need to set up that some day
MSmits: normally i dont have a lot of succes with it
MSmits: i tend to pick my parameters well enough that any improvement comes down to overtraining that doesnt translate to leaderboard
MSmits: this time i did not pick them that well
Pick8chu: Could you explain little bit of how you could do 'You need a way to look at the current game state and give it a score that tells you how good this is for you'
Pick8chu: cause right now what i'm doing is
MSmits: you can count the number of trees you have of each size
MSmits: count the number of suns
MSmits: game score
Pick8chu: oh
MSmits: the way your trees are shading eachother
MSmits: the future income
Pick8chu: so like calculate the weight of it
MSmits: yes
cegprakash: fn getScore { return player[me].score + player[me].sunpoints/3 } Pick8chu
ddreams: kind of how you'd evaluate the board if were playing yourself
MSmits: if you know chess, then the best example for you is the simple material score that is used
MSmits: pawn 1 pt, knight 3pt etc.
Pick8chu: but to see if it actually is making a progress, you gotta submit a lot of time right?
MSmits: not sure what you mean by submitting time
ddreams: no, you can play in the IDE or use brutaltester
Pick8chu: brutaltester
Pick8chu: is that a program?
MSmits: Pick8chu if you are in a lower league, you should not worry about brutaltester
MSmits: just try different things in your bot that make sense to you
Pick8chu: I'm on gold league rn
MSmits: ok, then you might have to worry about it :)
Pick8chu: 400 something, i wanna be in the legend
ddreams: what have you done so far?
MSmits: well it's a bit late to actually write a sim and search if you have not done that before
ddreams: sounds like we're talking with you as if you were in bronze
Pick8chu: I was doing weighting but to see what i changed actually made it better, I had to submit everytime and see if my rank got up
Pick8chu: Haha sorry bear with me, I just started this
MSmits: currently it's impossible to do that
MSmits: better test vs yourself locally
Pick8chu: And I really appreciate your help
MSmits: brutaltester is used for that
Pick8chu: okay, i'll google brutaltester
Trottel-Nit: Just pick games from Last Battles that you lost, push them to IDE and check if your new version would win them.
MSmits: yeah, someone else might be better at explaining what to do
ddreams: you can download the game code and run it yourself, and brutaltester automatically runs it over and over for you
MSmits: i always need people to explain it to me also :P
Pick8chu: you guys are really nice, thanks a lot
MSmits: Trottel-Nit thats a bit hard when the opponent uses randomness
MSmits: or when you do
Pick8chu: Yeah, i felt that too
Pick8chu: with one guy i thought i made a good progress but
MSmits: or even if they use the full calculation time and their search terminates at a different point
Pick8chu: if I submitted it, it wasn't liking
TobiasA: after the contest you guys have to teach me what you are talking about
ddreams: my bot is random only in how much CPU cg is giving it
MSmits: stick around and do multiplayer arenas guys
MSmits: a lot of them are fun
ddreams: I wish I had time to do more contests
ddreams: last time I tryharded was in 2016 or so
Pick8chu: ddream why not?
Trottel-Nit: Yeah, thats right but at least it gives a hint and is much faster than submitting every time. And with brutaltester you can only play against your own bots, can't you?
Pick8chu: are you busy ?
ddreams: work and family
CouscousAgha: my motivation is slowly disappearing because of this :/
MSmits: Trottel-Nit yes thats true
Pick8chu: ah. good for you then
ddreams: became a father last year
MSmits: unless someone shares a compiled bot with you. I dont recommend that
Pick8chu: Wow
Pick8chu: congrats man!
MSmits: gj
Pick8chu: how is it being a father! you must be so happy
ddreams: hehe, it's very fun
ddreams: many people say the first months are so boring, but that's not my experience
ddreams: like "all they do is eat and sleep"
MSmits: I thought it was pretty boring :P
Trottel-Nit: boring? Who the hell said THAT? :-)
ddreams: but they are communicating quite early
ddreams: MSmits right now :D
Trottel-Nit: I loved the first weeks and months. Never felt so much love :-)
MSmits: my daughter is 14 :P
Pick8chu: wow
MSmits: so i've been there
TobiasA: must be nice
Pick8chu: I never imagine having kids haha
Pick8chu: Maybe I should start thinking, must be good
ddreams: how old are you tho?
MSmits: it's ok, I always wanted 0 or 1 child, since my wife wanted 1 or 2, there was overlap
Pick8chu: I'm 26 international age
ddreams: and your national age?
ProG_r_aM_m_er: just gotta beat silver, i can do this lol
MSmits: wait what is international age
Pick8chu: I think I would want to have 1 girl and 1 boy
actualCoderTrevor: metric age?
Pick8chu: lol
Pick8chu: so i'm from korea
Pick8chu: and in korea, when you were born you are 1 yo
MSmits: oh ok so not an alien from another solar system with different size of year
Trottel-Nit: crazy
Pick8chu: and you get older every new years
Pick8chu: lmao
Pick8chu: no i wish
MSmits: well it's more accurate
ddreams: lua must have been made by a Korean
MSmits: you're 9 months old when you're born
MSmits: so 1 is more accurante
MSmits: accurate
actualCoderTrevor: ddreams lol I get it
Pick8chu: I think either china or korea
Trottel-Nit: Never thought about that there could be different ways of counting your age ... interesting.
Pick8chu: that's fair, yeah true MSmits
Trottel-Nit: So, do you celebrate birthdays? Or is it all together on new year?
Pick8chu: we do celebrate our own birthday
Trottel-Nit: just not getting older ...?
Pick8chu: but we all celebrate it on new years too,
Pick8chu: now you said it, it does make sense
TobiasA: wow that's cool
TobiasA: imagine everyone having the same birthday
MSmits: reasons to celebrate don't have to make sense :)
Trottel-Nit: Solved the birthday paradoxon
Pick8chu: lol haha
Pick8chu: okay i'm gonna try to download that brutal something
jacek: introduction to algebra?
Numby: What are people using as their timeout (ms) limit? Dunno if my code has too high of a variance on time taken for an iteration but it frequently times out even at 60ms limit :( I start timer after reading first input
somnek: xqcL
ddreams: timeout can also mean crash
Marchete: only 18hours to pass smits
ZarthaxX: :rofl:
Pick8chu: it says '<path_to_maven>/mvn package', what should i type?
DomiKo: mvn package
ddreams: if you're in the brutaltester directory and have maven installed, just mvn package
Pick8chu: okay trying
Pick8chu: Oh working! thanks lol
ddreams: then you get the jar file in the target directory
ddreams: now you have to get the game, aka. referee
struct: Whats gold boss current rating?
ddreams: 50.27
ddreams: Pick8chu https://github.com/LSmith-Zenoscave/SpringChallenge2021
ZarthaxX: you can check on any replay struct
ddreams: 807 computing now
Pick8chu: actually the build had failed
Pick8chu: No compiler is provided in this environment. Perhaps you are running on a JRE rather than a JDK?
Pick8chu: hmm
ddreams: you need java8
Pick8chu: yeah i thought i jsut downloaded it
ddreams: newer javas give you problems
Pick8chu: Oh
Pick8chu: jdk-8u291-windows-x64.exe
Pick8chu: i got this one, is this seems okay?
ddreams: looks ok to me, I used chocolatey
ddreams: choco install ojdkbuild8
Pick8chu: I'm learning so many things today lol
Pick8chu: I found i have jdk1.8.0_291
Cjx_1: so disheartened by these submits </3
ddreams: yeah... sucks
HDainester: hey is it still possible to rank up fro wood to higher leagues? A friend of mine is now in there for quite a it, winning every game, but he doesn't rank up (stuck at rank 24).
struct: submits are extremely slow atm
ddreams: HDainester not really at the moment
ddreams: practically broken
Numby: How slow is extremely slow
Cjx_1: normally to beat the bot per league (and thus level) it requires top 1-3
Astrobytes: Overloaded server.
Numby: like hours? :O
Cjx_1: ask them to check the leaderboards if their above that boss
ddreams: played 45 battles in 5 hours
HDainester: is there a chance this will be fixed in time?
Astrobytes: Currently no indication as to when (or if) it will be fixed.
Cjx_1: yeah, submits literally not going rn. Hours for just a handful of battles
ddreams: 30% in 5 hours... so 15 hours for perhaps levelling up
Cjx_1: i just had a big group of battles go
actualCoderTrevor: The first ten seem to be in parallel but after that it will crawl
Cjx_1: meh--seems like false hope
Cjx_1: prolly ment as a buffer
ddreams: the number of people waiting for scores is just increasing
ddreams: 816 now
Pick8chu: definitely not asians, it's 1am here
Pick8chu: okay just build successfully
ddreams: probably a lot of asians in the queue then, since it's been like this for over 8 hours at least
Astrobytes: Well. I'll come back later and check on this. I'm off to watch some paint dry instead.
Pick8chu: that's fair deduction haha
Pick8chu: How can i make refree .zip as .jar?
ddreams: mvn package makes a jar
CogitoErgoProxy: does this platform have troubles with stream api of java? it says that i've not read all the input (i use the stream after the standard code)
Marchete: in the eval, how can you control the COMPLETE part? I understand it's on endgame (day >15 , malus when more than Nx tree3), but what other way to control that?
ddreams: what makes you keep trees around?
PatrickMcGinnisII: if the diff in score is > nutrient value maybe
Cjx_1: i just go off differnece in total trees lol
Marchete: trees generate sun
Marchete: sun is the economy on the game
Marchete: trees3 generate score
Marchete: score is the final objective
Cjx_1: right on. Thats why even though its a super simple metric, if one has more trees than the opponent, shadows not considered, theyll be generateing mor sp
Pick8chu: how do you play your js code on brutaltester?
Pick8chu: i did 'js mycode.js'
UndercoverToad: rip codejam dreams :sob:
DomiKo: yeah
CouscousAgha: is it done?
DomiKo: we should do C2 :/
jacek: ?
UndercoverToad: domi and me both 56 points
UndercoverToad: i thought that would be enough
DomiKo: D1 matching euler?
DomiKo: me too
UndercoverToad: D1 random search
DomiKo: lul
UndercoverToad: just in a loop, for 300ms
PatrickMcGinnisII: MSmits congrats
UndercoverToad: i think i could have done D1 is a similar manner even (quick look at the analysis)
MSmits: ohh i am above boss
MSmits: but not certain, submit needs to finish
MSmits: lots of wins vs boss though
Smelty: u in legend now?
Smelty: :0
MSmits: 5-1 vs boss
MSmits: maybe tomorrow smelty
Smelty: wow
Pick8chu: good job!
MSmits: at the speed this is going
Smelty: i see that ur first in my league
Smelty: and IM 1258 in my league
Smelty: ....being the noob that i am
MSmits: I guess UndercoverToad was right
MSmits: thanks for your confidence again :)
Smelty: oo.o
ZarthaxX: smashing it smito
ZarthaxX: gz!
MSmits: I have a new version ready to go with 65% wr vs my old one too
ZarthaxX: sheez
MSmits: if that means anything
MSmits: it migth be overfitting, who knows
Marchete: submit it again ezpz
TheBatMan_TM: hey can someone help me with this?
TheBatMan_TM: I am back with a web app doubt
ZarthaxX: are all those just optimized params?
MSmits: yeah
MSmits: i hadnt fitted anything before
ZarthaxX: damn
TheBatMan_TM: but last time somebody helped me, so i am trying again
ZarthaxX: didt change any term?
ZarthaxX: wow
MSmits: not really no, not in a while
Marchete: how do you adjust weights then Smits?
TheBatMan_TM: source: https://replit.com/@kaushikkalesh/Online-Notes
ZarthaxX: wow
Smelty: i invested everything into knight moves and now my ai sucks
MSmits: currently i do it like this. Say I suspect a good value is between 0 and 400
TheBatMan_TM: can someone tell me why my delete_note() isn't working?
Smelty: hmm
Marchete: like x*costs+y*sungain+z*positionalscoring
MSmits: i create bot 0, 100, 200, 300, 400
TheBatMan_TM: i think the problem is in index.js
MSmits: have them all fight
Smelty: k
MSmits: 10 battles x 400, so 4000 total
MSmits: then add up the wrs for each version
Smelty: huh
Smelty: Msmits ur over a point above the boss i'm confident in you promoting
Marchete: I don't have a suspect of what my x,y,z could be
MSmits: yeah seems likely now
Marchete: I was doing unit tests
PatrickMcGinnisII: i'll celebrate my small victories. :(
Marchete: like "test seeding"
Marchete: and it's seeding correctly
MSmits: what do you mean test seeding?
Smelty: yay won 2 out of first 10 battles....:(
MSmits: oh
Marchete: scoring seed position
Marchete: from all free positions, check score
MSmits: ah ok, well i have two params that do somethign with seed score
Smelty: yes
Pick8chu: I kept getting 'Negative score during game 5 p0:-1' this error
Pick8chu: does anyone knows?
MSmits: well shadows mostly
Smelty: ._.
Smelty: Pick8chu what lang
Marchete: I have N weights for that :D
Cjx_1: pika, thats an invalid oputput deal i think
Pick8chu: js
Pick8chu: js ..
MSmits: generally, you pick 1 param you keep fixed (I keep my score param at 1000)
Smelty: well...javascript..hmm
Marchete: sungain now + sungainfuture+etc etc
MSmits: and then you try to fit the others relative to that
Pick8chu: but i used exactly the same code that i used in IDE
MSmits: you can do some easy calc to see what a realistic range would be
PatrickMcGinnisII: 21st in my language...rofl
Smelty: try going past condition only if >-1
MSmits: you wouldnt want suns to be more valuable than score for example
Cjx_1: you doing anything in parralell?
actualCoderTrevor: Pick8chu is trying to get brutaltester to work - and so am I with similar results (python)
MSmits: Cjx_1 how do you mean?
Pick8chu: Oh hey haha
actualCoderTrevor: I've been lurking to see if someone helps you / if I can figure it out first
actualCoderTrevor: In my case (and maybe yours) it seems like the referee isn't outputting anything
Pick8chu: well mine too it kept saying 0%
Pick8chu: but i guess that's just cause i'm returning -1
Marchete: I usually just MC the weights until I hit jackpot :D
natofp: you guys can just run the referee with inputs without brutaltester and see if that works
DannyBoy1024: Does testing in arena also become super slow for you guys as well?
Marchete: but I need to improve the search algo too...
natofp: for me even this crashes input programs :))
Marchete: too many things too little time
Marchete: :S
Cjx_1: yeah danny, its broke rn
MSmits: yeah i still need to try a different search also
MSmits: just to see if it works
MSmits: should get to coding
DannyBoy1024: I launched a test 4h ago and now it barely progressed 60%
Michael_Howard: Greatest feeling ever: finding the awful bug that's messing your whole bot the day *before* the contest finishes.
MSmits: then it;s getting beter
MSmits: better
MSmits: 4h 60% i mean
actualCoderTrevor: natofp I did try that using the command brutaltester gives me for "playback". Well it tell me the command should end with -d but the referee says it needs an option for that or something and I didn't provide one
Cjx_1: folks looking at full shadow lengths or just immedietely adjacent places, or disregarding them? Im yet only looking at adjacent squares only lol
MSmits: I look at full shadows
MSmits: for big trees thats important
Numby: where do I spend 95% of my time during sim? create a list of possible actions :(
MSmits: i dont create lists during sim
Marchete: I do a sun gain calculation Smits
MSmits: me too Marchete
Marchete: for all 6 days
Marchete: I add some decay
MSmits: for full game here
Numby: how do you store the possible actions
Cjx_1: smart, I homebrewed too much and now I dont wanna go back and rewrite the amount Id have to to easily get full shadows lol I prly will have to today tho lol
Marchete: I have that too
MSmits: I haven't added decay
Marchete: but with a low weight
Marchete: I'm not sure about endgame sungain
Cjx_1: numby, theyre stored for us if you start with one of the templates--in numerical order of the cell as well\
MSmits: me neither
Pick8chu: Oh i got the same thing -d
PatrickMcGinnisII: MSmits u definitely did something...yesterday u were avg. beating me by 10 pts, now you are around 28 ... and i did change some stuff for the better on my end
MSmits: not sure about a lot of things
Marchete: I calc sungain current + expected growth+expected growth2+engame sun
MSmits: PatrickMcGinnisII yes fitting parameters better
actualCoderTrevor: If I just copy-paste a bunch of text here it will automatically make a paste link right?
Marchete: I don't check "neighbours" or shadows
Marchete: but sungain futue
Marchete: future
Numby: I mean when simulating a game and at a certain state you first have to find what are the possible actions. and there's a number of them so need to store them in some list right?
MSmits: but if you do future sun gain you do shadows too
Marchete: it's faster for me too
PatrickMcGinnisII: Marchete where r u on leaderboard?
Marchete: I mean not just shadow
MSmits: btw future sungain might not properly take into account seed positioning
actualCoderTrevor: http://chat.codingame.com/pastebin/6ca439c3-13fa-440a-96dc-4c6838762909
Marchete: ofc
MSmits: it takes many turns for a seed to give enough income for it to matter where it is placed
kovi: speedup on servers!
Marchete: but I don't get any better idea
Marchete: I'm like 3000th PatrickMcGinnisII :D
MSmits: tbh, all i have is a bunch of features that made sense on some level, i gave them a param (obviously a wrong one at first) and am fitting them today
actualCoderTrevor: And the link it gives me looks like this: https://imgur.com/JRZLorn any ideas?
PatrickMcGinnisII: I'm still lurking aroun 1.1k overall
DaNinja: java -jar cg-brutaltester-1.0.0.jar -r "java -jar -Dleague.level=3 -Dlog4j.configurationFile=log4j2.xml spring-2021-1.0-SNAPSHOT.jar" -p1 "../x64/Release/Spring2021-ai1.exe" -p2 "../x64/Release/Spring2021.exe" -t 2 -n 30 -l "./logs/"
PatrickMcGinnisII: Marchete, u at 1471...smh
DaNinja: works for me
actualCoderTrevor: Thanks I'll try that!
Marchete: it's the referee.jar published somewhere?
Marchete: spring-2021-1.0-SNAPSHOT.jar
Marchete: I know maven and sh*t
Nerchio: guys what do you think about search ignoring complete and doing just economy,
Marchete: no
Nerchio: i feel like completes need to account for enemy so they are kinda useless in a solo-search
PatrickMcGinnisII: ok, my echo dot just went off, who is phishing my network?
Nerchio: i override almost all completes of my search anyway
Marchete: I don't know :D
Marchete: don't generate completes until endgame
Marchete: problem solved
Marchete: or trees3 > 4 or 5
Nerchio: yeah kinda makes me wanna try ignoring completes in search tbh
PatrickMcGinnisII: i don't cut unless tree3 is >3 no matter the running score
Marchete: directly don't generate the moves
Marchete: in early game
Nerchio: if only submits were working
MSmits: they're a little better now
MSmits: but might still be very slow
MSmits: dont all submit at once now :)
Nerchio: yeah i'll wait
PatrickMcGinnisII: i think it's been an hour for 30%
Scarfield: since the cost of completing is more expensive the more trees of that size you have, there should be an optimum of when to start thinking about completing i guess. could try limiting it, instead of completely disregarding it(?)
Nerchio: my point is that eval is hard
Cjx_1: thats a very hard tipping point to figuer out I think
Nerchio: when to sell trees and when not
Nerchio: so search can pick suboptimally because it tries to sell
Nerchio: if you rule out completes you can have much easier eval focusing on economy
Cjx_1: honestly, I think its likely better to just use the opponents score and your trees as metrics for complete, but my endgame still sucks to be fair lol
MSmits: Scarfield it's a hard one to optimize, the opponent also enters into this. If he starts eating neutrients....
PatrickMcGinnisII: i figured out that if the cost of growing a tree3 is >10 then it's not economic
MSmits: it helps to hard-prune some things like, dont complete trees when you have few of them halfway through the game. Not because your search cant find it, but because it lightens the load on the search
MSmits: with beamsearch that means less states to beam
Pick8chu: could anyone help me to set up brutaltester?
Scarfield: yes of course, but a bad estimate of it, might still be better than removing complete from search entirely(?)
Pick8chu: it just kept saying i got negative score
MSmits: time outs
MSmits: i think
MSmits: your bot is crashing somehow
Pick8chu: 02:00:35,347 ERROR [com.magusgeek.brutaltester.GameThread] Negative score during game 3 p0:-1 02:00:35,347 ERROR [com.magusgeek.brutaltester.GameThread] Negative score during game 3 p1:-1
MSmits: this is my usage:
MSmits: java -jar .\cg-brutaltester-1.0.0.jar -r "java -jar Referee.jar" -p1 "./0" -p2 "./20" -t 8 -n 400
Pick8chu: which language do you use?
MSmits: bots are called 0.exe and 20.exe
Pick8chu: I see
MSmits: that doesnt matter, these bots are executables, compiled
MSmits: not sure how this would work with a non-compiled language
DaNinja: nice to have 8 threads! :)
Pick8chu: I uses js and it maybe cause of how i conpiled it
MSmits: yes that helps DaNinja, but i had to stop my yavalath solver :(
Scarfield: 824 results found for «Score - Computing» :'(
UndercoverToad: this morning when it start to get slow, there were 100 in submit
MSmits: yeah they fixed something at least, submits are moving
MSmits: just very slow
dbdr: you're still running it MSmits? :o
MSmits: yep
dbdr: wow
dbdr: :D
MSmits: me and 3 others above boss
MSmits: oh you mean yavalath solver?
dbdr: yeah
CouscousAgha: oh i think it's fixed now
MSmits: yeah, i restarted it later, it's now parallelized. I restarted because I polluted it with counterbooking
MSmits: want to run solver without interference
MSmits: two starting moves will solve automatically, but i still wonder if any of the other ones can be solved
MSmits: academic interest mostly :)
dbdr: awesome!
UndercoverToad: others call it obsession
UndercoverToad: but i'm impressed by it
MSmits: those are not mutually exclusive
MSmits: obsession is often amazing :P
MSmits: thanks UndercoverToad and you were right, i got there in the end
MSmits: still some ways to go but +2 pts should do it
UndercoverToad: i also predicted that you will rank above me
MSmits: yes you did and it's a possibility. Fitting params does a lot of good
MSmits: but i cant say for sure
UndercoverToad: i might run some offline batches tomorrow to hope for the best
MSmits: it well not hurt at least
UndercoverToad: except for overfitting maybe
MSmits: I am liking the method of creating 5 bots and have them all fight
UndercoverToad: did anyone get around the log file warnings for the brutaltester?
MSmits: it reduces overfitting due to drift
kovi: hmm, my code performs better when servers are ok
MindController: Holy, 7 hours into submit and only 70% done
kovi: 100%=#1
UndercoverToad: it's reported but staff is on a long weekend
DaNinja: I created a log directory, that fixed the errors
UndercoverToad: i still hope that they'll do something about it soon. we just don't know more
UndercoverToad: just a folder "log"?
MSmits: so they just log back in on monday and go like: "So, what'd we miss?"
DaNinja: yes
dbdr: gg kovi
UndercoverToad: anything else to consider? e.g. command line arg
DaNinja: -l "./logs/"
DaNinja: at the end
UndercoverToad: i'll try
UndercoverToad: 19:12:22,894 ERROR [com.magusgeek.brutaltester.GameThread] Error during game 4 19:12:22,895 ERROR [com.magusgeek.brutaltester.GameThread] ERROR StatusLogger No log4j2 configuration file found. Using default configuration: logging only errors to the console. Set system property 'log4j2.debug' to show Log4j2 internal initialization logging.
DaNinja: try this
DaNinja: "java -jar -Dleague.level=3 -Dlog4j.configurationFile=log4j2.xml spring-2021-1.0-SNAPSHOT.jar"
UndercoverToad: ERROR StatusLogger File not found in file system or classpath: log4j2.xml
UndercoverToad: let's see where I can find that
DaNinja: https://github.com/dreignier/cg-brutaltester/tree/master/src/main
MSmits: 1 more battle for renarrd
MSmits: 49.83 > boss 49.46
UndercoverToad: ok, now just compiling the referee again with a setting other than info
UndercoverToad: or replacing info in that xml
UndercoverToad: no, the latter didn't do the trick
Pick8chu: does anyone use js?
Pick8chu: it gives me readline is not defined
ddreams: http://chat.codingame.com/pastebin/08caa0c6-3928-4aec-846a-44a3622ba008
MSmits: LeRenard is promoted to legend. Congrats
ddreams: UndercoverToad
LeRenard: Thank you for your help
timbersson: UndercoverToad someone gave me this earlier:
timbersson: http://chat.codingame.com/pastebin/3ed88507-2e83-460f-bef5-20c541f2b99a
timbersson: put in src\main\resources\log4j2.xml
ddreams: so battles are finally fixed
ddreams: yay
Cjx_1: if this arena ever finishes, I think I may be low 300's in silver, maybe even mid 200's. Im personally hella happy lol :D Going for gold so hard lol
struct: recurse submited :D
struct: only saw now
jacek: :scream:
Astrobytes: lol
ddreams: I have a better bot too, but don't want to submit it...
ddreams: this one was beating msmits and he's numba one
UndercoverToad: then reCurse might reach legend on Monday
struct: lol
reCurse: Yeah thanks to that stupid scare...
reCurse: I choked
UndercoverToad: losing matches in wood1 already? :P
NinjaDoggy: lol
reCurse: I seriously thought the problem wouldn't get fixed and even promoting to bronze before end would be a close call
UndercoverToad: which was your alt? i would say probably someone from Japan
reCurse: Anyway nothing to see move along
UndercoverToad: wait, they did something to improve it now?
reCurse: Yes
struct: its better atm
UndercoverToad: nice
ddreams: still scoring over 800, but I got from 30 to 60 in lunchtime
ddreams: so something has unjammed
reCurse: Backlog will take a while to go through
UndercoverToad: we know that from league opening back in the days of from CSB promotions
reCurse: Thanks for fixing it CG but no thanks for the heart attack
DomiKo: Toad don't beat me :(
UndercoverToad: then don't fight me
jrke: recurse is that you - https://www.codingame.com/profile/316314dc83b90a6bacf4438dc504e1940285672
ddreams: starting fights you can't win isn't a good idea
UndercoverToad: you got me at codejam by a few ranks already. not that it would matter
reCurse: Streaming account jrke
reCurse: Had separate logins for everything when streaming
reCurse: Super restricted HDD access etc
UndercoverToad: not to accidentally leak anything?
reCurse: Yeah
UndercoverToad: wise decision, would do the same if i would be streaming
reCurse: Also streaming on secondary monitor
reCurse: So primary gets all the new crap
UndercoverToad: then again i would be a horrible entertainer. no one would want to see me for more than 5 minutes
reCurse: You're implying I wasn't
UndercoverToad: didn't watch long enough to tell :innocent:
reCurse: Don't blame you
Cjx_1: if you tell me your wearing a toad outfit rn, id watch :P
Astrobytes: I thought you were pretty good at it tbh
MSmits: 5 mins is quite long. You could cover a song in that time UndercoverToad
UndercoverToad: and get a twitch strike
Cjx_1: maybe occassionall ribbits? Thatd be funny lol
Astrobytes: haha
Scarfield: you were streaming this contest reBless?
jacek: stream this https://www.youtube.com/watch?v=zHU2RlSCdxU
reCurse: No a long time ago
Cjx_1: a streamer HexTree_ did at least some streams of it in python
Scarfield: ah okay, thought i missed something :p
Scarfield: saw most of your locam videos on youtube before i started learning c++, probably worth having a 2nd look :)
reCurse: Nooo
struct: time to put it down :D
Cjx_1: F, now IM watching hypnotoad for 10 hrs....</3
reCurse: The only reason I don't is because I mostly forget about them
ddreams: wonder if the game would be less asymmetrical if the starting trees were diagonally opposite
PatrickMcGinnisII: U guys need a life ALL GLORY TO THE HYPNOTOAD
pszemsza: is the Spring Challenge going to be available later in Bot Programming?
reCurse: Yes
struct: I wonder if pb 4 is still playing
ddreams: if they
ddreams: if they're not in the chat room, can they get pings?
reCurse: Only if they log in before history gets erased
struct: no
ddreams: no tab completion on pb
dbdr: lol, opp stops playing because he thinks he won, but... https://www.codingame.com/replay/555964133
kovi: wow
ddreams: lol, nice
Pick8chu: does anyone know how to compile javascript into exe?
Scarfield: xD
Pick8chu: i used node pkg but apprantly readline is not defined
ddreams: no, but you shouldn't need to
kovi: i think its because he detect and support another rust bot
Pick8chu: wdym?
Alonius: Are there any changes in game rules in gold and legend leagues?
ddreams: you're trying to play locally, right?
Pick8chu: java -jar cg-brutaltester-1.0.0.jar -r "java -jar spring-2021-1.0-SNAPSHOT.jar" -p1 "js myCode.js" -p2 "js myCode.js" -t 2 -n 1000 -l "./logs/"
Pick8chu: yes i tried this
ddreams: what's js though? node?
Pick8chu: js myCode.js doesn't work so i thought
Pick8chu: yes i guess
ddreams: uhm
Scarfield: no Alonius, rules stay the same from bronze
ddreams: try writing node myCode.js instead
ddreams: haven't tried node on windows, but that's what I'd do elsewhere
Pick8chu: i just did and Negative score during game 1 p1:-1
Pick8chu: it seemed work tho
Pick8chu: time.��RED�� Player 0 was disqualified. ��RED��Player 1 has not provided an action in time.��RED�� Player 1 was disqualified.
Pick8chu: wait, is this timeout errors?
Lysk: probably yes
nikita_lutsuk: \
Pick8chu: Why tho? the same code would work fine in IDE
Pick8chu: it does work
Pick8chu: ...hmmmmmmmm
KelvinAndHubbles: tf the silver boss is in 29th place
KelvinAndHubbles: submits must be reallly slow
ddreams: Pick8chu probably playing in wood league
TobiasA: is the boss supposed to be higher lower?
TobiasA: silver boss that is
Pick8chu: @ddreams like go back and play there?
KelvinAndHubbles: well, I mean considering who ever beats it gets promoted, yeah
Scarfield: there are submits from 12.00 ish, they have been stuck for loong, but its starting to help.
KelvinAndHubbles: it's just theres about 30 submits currently beating it that hasn't finished
tobk: my submit from ~9h ago finished just now...
ddreams: Pick8chu no, the default league when playing locally is wood leage
ddreams: *league
Pick8chu: oh
Pick8chu: but I tried with the starter Kit
ddreams: so you have to write, I think, -D league.level=3
Pick8chu: like the basic
ddreams: oh
Pick8chu: there's no way it would give me a timeout.. but I should try that too
ddreams: ok, well, if you add "-s" to your options, you'll get a web server where you can inspect the output
Pick8chu: I think the problem is
Pick8chu: javascript doesn't have a funcion called readline
reCurse: It's hacked in by CG
Pick8chu: but somehow on codingame they use it to get inputs
BlitzProg: The more I try to fix my heuristic, the less I understand it. How am I even supposed to play
reCurse: Someone posted an implementation a long time ago
Pick8chu: oh they did?
Pick8chu: where?
Pick8chu: like community on CG?
reCurse: Forgot sorry
Astrobytes: May have been the discord. Though it might have been pasted in the forum somewhere too.
Lysk: -s is for swap, do you mean -l "./logs"
UndercoverToad: it was discord
Lysk: oops
Scarfield: DiscoBytes
Pick8chu: there's one on french but i have no idea lol
ddreams: Lysk that's in brutaltester, if you run the referee by itself it starts a server like an IDE playout
Pick8chu: https://www.codingame.com/forum/t/readline-et-print-en-js/2327/5
UndercoverToad: JS readline:
UndercoverToad: http://chat.codingame.com/pastebin/888f4c52-4ed5-4177-9374-74e141d4d5b8
Pick8chu: OHHHHHHHHHHHHH
Pick8chu: THANKSSSSSSSSSSSSSSSS
Astrobytes: ffs toad
Lysk: oh, got it, ddreams (thanks for the help thes other day for cin.good() by the way :) )
Pick8chu: ah.
Pick8chu: still doesn't work...
UndercoverToad: did you find it too Astrobytes?
Astrobytes: Yes!
Pick8chu: fd = fs.openSync('/dev/stdin', 'rs')
UndercoverToad: in the Ocean of Code thread. That will be deleted at some point
Pick8chu: this doesn't exist
UndercoverToad: i still don't get why some channels disappeared. can't you just make them read-only?
Astrobytes: Clutter?
UndercoverToad: fountains of knowledge
Cjx_1: ok I got it. I just need to redesign game.board I think. This will work, and be worth the time. (he said without much belief in the words lol)
Astrobytes: Who am I to question
Scarfield: Submit count dropping 729 results found for «Score - Computing»
ddreams: Lysk yw :)
UndercoverToad: 724
UndercoverToad: we are on the same page Scarfield
Enknable: omw to gold! :)
CouscousAgha: holy * the top 20 spots in gold is literally at war
CouscousAgha: everyone are fighting for legend
Scarfield: we were literally xD
CouscousAgha: beautiful
UndercoverToad: those are the legend bots stuck on slow submit
Numby: Lol 3 people above boss now
Cjx_1: I think all the way to silver will be quite heated for the rest. I cant imagine the savagry in platinum lol
Scarfield: on the same page that is
Numby: 4*
Lysk: wow 363 submit in gold oO
MSmits: damn i am dropping, very near boss now
ZarthaxX: F
Lysk: still a good margin above the boss score
ZarthaxX: its a bit random
ZarthaxX: you had a good streak smito :(
MSmits: yeah
UndercoverToad: https://www.codingame.com/replay/555982654 a new opening - and it's 1 sun better :o
UndercoverToad: should check again with me upgrading the large tree later maybe
UndercoverToad: my opening is poor
ZarthaxX: hardcoded opening? :P
DomiKo: why not use search instead of opening?
UndercoverToad: i do
UndercoverToad: but my search is bad
DomiKo: yeah I beat you
DomiKo: :stuck_out_tongue:
Jelle: How to start the escape room game?
nader5: how to learn on this website
aangairbender: Jelle go to https://escape.codingame.com
Pick8chu: GG
Jelle: Yes I started and got a number and a link
Jelle: but then nothing happens
ddreams: I think people say check your email
ddreams: spam? though I have no idea really
Jelle: Nope also not there, thanks for thinking along
CouscousAgha: LOL TOP GOLD IS LITERALLY AT WAR
ddreams: I don't like that brutaltester is printing output ignoring draws
CouscousAgha: are you guys witnessing this
Twelve0: and gold boss's score is increasing!
MSmits: oh come on!
MSmits: 0.06 below boss now :(
Scarfield: just resubmit
MSmits: lol
Scarfield: xD
ddreams: lol
CouscousAgha: i dont htink u need to resubmit if you are 0.2-0.1 below boss
ddreams: I will resubmit soon
CouscousAgha: one push from another player and you are promoted
ddreams: well, I'll wait the last 10% nwo
Scarfield: you will get there im sure, if not now then just a little push
MSmits: yeah i guess so, it's really annoying though, i have a better version offline
MSmits: but if i submit it, it will take forever
UndercoverToad: then submit and go to bed
MSmits: not yet my napnap time
ddreams: it's progressing at a good speed now
ZarthaxX: ^
Scarfield: oh, yea :/ looks like the submits are going faster and faster though
dbdr: UndercoverToad still waiting for your morning submit? :D
UndercoverToad: dealing with my codejam disappointment
ddreams: oh, what happened?
dbdr: :(
UndercoverToad: took a short walk after and of course got the idea how to solve a problem
ddreams: brains
ZarthaxX: ;(
UndercoverToad: i needed top1000 and reached 1194. that's what happened
struct: thats normal
ddreams: ack, that sucks
ClockSort: that's rough, sorry @UndercoverToad
MSmits: that's why i dont like short contests.. I need walks to get ideas too
MSmits: sucks
PatrickMcGinnisII: I walk to the fridge
dbdr: I spent the whole day trying to make a more clever search work. I just realized I was using a stinky old eval that ignores the opponent :upside_down:
UndercoverToad: i see you failed that round too ClockSort
ClockSort: I got 2845th place, yeah
MSmits: lol PatrickMcGinnisII
UndercoverToad: robo at 2856
ddreams: you care about the opponent now, dbdr?
ddreams: :p
ClockSort: I couldn't even solve Matrygons ... i was going for the biggest test set on it
UndercoverToad: i solved small first and then returned for big
UndercoverToad: but on C and D it remained small
alchemsti: sorry to hear that UndercoverToad
ClockSort: solving D small put me ahead of all the other small guys because it was 1pt more, haha
UndercoverToad: i wrote a random solver and let it run 300 ms on each testcase, taking the best result as the answer
alchemsti: Google's code Jam? Did you have a motivation other than a cool contest?
ClockSort: it was a tough round, you needed to solve two of B, C, and D to qualify
dbdr: ddreams I was trying to, if not for that bug
UndercoverToad: i'm collecting tshirts, that was my main point of motivation
ClockSort: :sweat_smile:
reCurse: Wear them all at same time and aim for world record
UndercoverToad: that could cause damage to them
alchemsti: Ha! That's a good motivation, IMO
dbf: UndercoverToad - if you have too many t-shirts, check this one: https://1.bp.blogspot.com/-RT69bnWgb3g/UsmkKfza4RI/AAAAAAAAfZY/GksJbTYqe4U/s1600/IMG_20131229_205945.jpg
DomiKo: Petr I see
ClockSort: http://chat.codingame.com/pastebin/606d224a-5c5d-4740-9664-d6d85c9f069c
ClockSort: @UndercoverToad see above pastebin
UndercoverToad: dbf that's horrible
MSmits: submit fininished. 0.09 below boss
MSmits: just needs 1 battle
MSmits: damn a loss
CouscousAgha: I got you MSmits
dbf: UndercoverToad I think in this mode they will be save better than if you wear and wash them regulary :)
ddreams: practically in legend then, MSmits
Zenoscave: MSmits that's rough
Zenoscave: are submits faster again?
zasmu: me too, gonna kick the boss :)
ddreams: yes Zenoscave
MSmits: well I'll go take a break, I may get pushed over later
ddreams: they're working well now
Scarfield: not well id say, but definately better
kovi: maybe just wait to be pushed (boss pulled) by pending submits
Zenoscave: I feell asleep during my last submit
ddreams: still have a huge backlog, but 300 are gone now
ddreams: only five hundred-ish left from 820 or so
Scarfield: i saw some submits that were 8 hours and counting :O
Zenoscave: Last submit i did was 2... So it got worse
ddreams: mine finished now after 7 hours, with like 70% the last hour
UndercoverToad: dbf i'm also impressed by the number of different colors. All I get is black
Zenoscave: https://www.codingame.com/replay/556006710
Zenoscave: That's a copied bot, no?
DomiKo: UndercoverToad in Russia they like colors
ClockSort: @UndercoverToad were you close on large C or D in the code jam?
ddreams: copied from?
UndercoverToad: i thought it wasn't even necessary
actualCoderTrevor: Does anyone have the referee and/or brutaltester working with Python? It seems like the referee and my poor little script just aren't communicating. This is the closest I've been able to get: https://imgur.com/jedDOeK
UndercoverToad: on C I realized that the last 1 represents the largest cake
UndercoverToad: from there it's just 2 recursive calls of both sides
UndercoverToad: and a binomial
Zenoscave: Is CJ over!? I just woke up
UndercoverToad: over
Scarfield: (╯°□°)╯︵ ┻━┻
Scarfield: ?
Scarfield: poor zenos table im afraid
TobiasA: i semi-hardcoded my bot to get from around 500th to 42th rank in bronze but it wasn't enough to defeat the boss and now i don't know what to do
ddreams: do you still seed next to your own trees?
Zenoscave: The table has too much stuff on it otherwise...
MSmits: aww screw this i am submitting my new bot :P
Scarfield: gl
dbf: UndercoverToad, raic-2020 could be black also :)
TobiasA: yes
ddreams: I'm on the way to do the same
Zenoscave: MSmits how much better is it?
ddreams: TobiasA then figure out a way to avoid it
ClockSort: @Zenoscave it's probably worse XD
MSmits: locally 65-35
MSmits: not sure otherwise
ddreams: after how many matches?
Zenoscave: After how many matches?
MSmits: 4000
Zenoscave: did you play benches?
Zenoscave: or just your bot?
TobiasA: i've been trying values for the score but it doesn't work
Zenoscave: this game is very rps
MSmits: i did 400 games per game with 5 different bots, each with a different param value
MSmits: so i had them all play against eachother
TobiasA: like to avoid seeding next to mine the tree size has to be greater than 1 but the bot doesn't know that
MSmits: this one came ahead each time
MSmits: and 65% vs old
Zenoscave: Gotcha.
ddreams: almost certainly better then
MSmits: you would think so
ddreams: that kind of local improvement has always given a big improvement in ranking for me
MSmits: won all first 10 anyways, should speed things up
MSmits: but gonna take a break ttyl, hope this submit makes it
ddreams: for sure, gl though :)
MSmits: thanks
actualCoderTrevor: ... If I compile the starter.cpp file, I can use that with the referee no problem. But the starter.py file is no good. I'm just launching it with "python starter.py" is that wrong? I have python 3.9.0.
Enknable: what errors u get?
UndercoverToad: open a terminal and try that command there
actualCoderTrevor: https://imgur.com/a/KCeM0iL
UndercoverToad: windows: do you have python as a path variable?
actualCoderTrevor: Yeah I'm able to launch the script from the command-line
UndercoverToad: flush stdout
UndercoverToad: print(command, flush=True)
Enknable: "?RED?"
actualCoderTrevor: Hmm okay I'll try that...
ddreams: smart toad
UndercoverToad: i literally have a tshirt that tells you to do that :rofl:
UndercoverToad: https://img.printfection.com/18/3872/BDcI0IOU3xhVh3n/American+Apparel+2102+-+CodinGame+-+Black+%28Front%29.png
SabertheLost: I win against bossoot 1 in 10 now. Still not good enougth.
ddreams: lol
DomiKo: :joy:
SabertheLost: I want to play agenst myself. (my ai vs myself)
KelvinAndHubbles: you can, delete the other AI down at players options
ddreams: I guess he means himself as human player
ddreams: not too hard to make it
SabertheLost: I want see if my code is "smarter2
actualCoderTrevor: Sadly the shirt didn't help :( Also I showed you the error for my own bot the starter.py is just slightly different but seems to be the same root cause: https://imgur.com/F6awU0h
KelvinAndHubbles: ahh, well then just write this game as an actual game
ddreams: need to disable timeouts in the referee, and make a program that asks you for the move
Zenoscave: actualCoderTrevor what referee are you using...
UndercoverToad: that looks like your bot has a problem actualCoderTrevor
UndercoverToad: do you play with bronze rules?
Zenoscave: If it's mine I don't send the action's
actualCoderTrevor: UndercoverToad this is just the starter.py
ddreams: actualCoderTrevor it has a stack trace there
Zenoscave: which is why the index[0] would fail
ddreams: aha
actualCoderTrevor: Yes I was able to make it that far with "java -jar -Dleague.level=3 -Dlog4j.configurationFile=log4j2.xml spring-2021-1.0-SNAPSHOT.jar"
ddreams: right, Zenoscave has removed that part of the output
UndercoverToad: you don't send actions Zenoscave?
UndercoverToad: how dare you?
Zenoscave: For time reasons.
ddreams: I saw that he removed it in a later commit
ddreams: I still use an earlier version with the actions
Zenoscave: I'll put a full version up
Zenoscave: my bad
actualCoderTrevor: wait... the -Dleague.level=3 isn't doing anything
ddreams: nice way to remove the timeouts
DaNinja: actualCoderTrevor how are you running the viewer?
ddreams: I recompiled the entire codingame sdk
DaNinja: this works for me
DaNinja: java -jar ./target/spring-2021-1.0-SNAPSHOT.jar -p1 ../x64/Release/Spring2021-h1.exe -p2 ../x64/Release/Spring2021.exe -l .\logs/game4.json -s -d seed=1009211670800188956
ddreams: DaNinja add "-s"
UndercoverToad: he has -s
actualCoderTrevor: DaNinja it works fine with compiled .exe's but not python scripts
ddreams: UndercoverToad is -Dleague.level=3 the right syntax?
Zenoscave: It doesn't work if you don't generate your own actions.
UndercoverToad: the python starter to be precise
UndercoverToad: yes ddreams
DaNinja: actualCoderTrevor to get the graphics, copy all files in main\resources\view\assets\ to main\resources\view\assets\assets\ then recompile
actualCoderTrevor: DaNinja thanks I'll try that in a bit. But I think it should work without them(?) I mean it seems to work with an .exe like I said.
actualCoderTrevor: UndercoverToad to answer your earlier question, I see the same behavior with the version at https://github.com/dreignier/SpringChallenge2021 as well as https://github.com/LSmith-Zenoscave/SpringChallenge2021
ddreams: they have the same removal of output
Zenoscave: I merged dreigniers changes into mine
Zenoscave: It always sends a action count of 0
Zenoscave: I'll probably undo that change but keep the rest
Magus: I changed that to make the referee faster :D
actualCoderTrevor: So this is different from what the IDE is using?
Magus: my code never used the possibleMoes
Magus: *possibleMoves
actualCoderTrevor: Yes the Python starter code definitely uses that
actualCoderTrevor: (As does my script)
Zenoscave: I figured Magus. I didn't think people would use mine so heavily before generating possible moves
ddreams: is it you that made the brutaltester?
ddreams: it's cool
jacek: AutomatonNN where are you :(
Zenoscave: jacek lol
Magus: ddreams: Yes I created brutaltester. I'm not the only contributor, eulerscheZahl and some other codingamers made some pull request on it.
ddreams: Nice of you to share
Magus: I wanted to add a feature for the confidence interval in the statistics. But never find the time to code that.
Zenoscave: Based on wilson's CI test?
Zenoscave: I can add that soon if you want
TobiasA: this is the coolest thing i have ever seen :sob:
TobiasA: https://www.youtube.com/watch?v=hQ4ryudP4j8
Zenoscave: https://github.com/LSmith-Zenoscave/SpringChallenge2021/releases/tag/v2
Zenoscave: try this one actualCoderTrevor
Zenoscave: no need for -Dleague.level flag either
Zenoscave: It only allows bronze rules
jacek: TobiasA maybe because you havent seen THIS https://www.youtube.com/watch?v=dQw4w9WgXcQ
DomiKo: jacek :thinking:
DomiKo: that link looks familiar
jacek: youtube is pretty popular site after all
ddreams: QQ domiko
dbdr: MSmits still the same submit?
ddreams: he resubmitted
dbdr: yes, I'm silly, just had a look at the submit time
KelvinAndHubbles: submissions going faster yet?
MSmits: this bot should be better, but not seeing that much evidence of it yet
Zenoscave: jacek HA I have ads so it didn't work
TobiasA: i knew it yet i still clicked
dbdr: why? did not promote?
Astrobytes: wait, your first one didn't make legend?
MSmits: nope
MSmits: got stuck -0.2
Zenoscave: I still listened anyways. I like this song
dbdr: ouch
ddreams: I was sure mine would make it too, since it seemed to beat msmit so much
MSmits: could have waited, but i need to test my new param values
MSmits: and supposedly it has 65% so..
Astrobytes: fair
UndercoverToad: vs yourself where every tiny change has effects
dbdr: did you read the stats right? ;)
MSmits: yes i know UndercoverToad
dbdr: or is it 35%? ;)
MSmits: it's a risk, but i still have the old code. Can just resubmit
MSmits: lol no, i looked right this time :)
MSmits: going afk a bit, dont want to watch this, too stressful :P
ddreams: you use swapped position while testing, I hope
dbdr: we'll look for you
dbdr: and you'll make it
dbdr: I had a long day, lots of frustration. but what an illumination. happy I found it and did not stay in hte mystery of why it barely works
TobiasA: damn rick astley got moves tho
UndercoverToad: did you submit a working version of it already dbdr?
dbdr: yep, it's in :)
ddreams: are you also considering the opponent now euler?
UndercoverToad: no :/
jacek: selfish
UndercoverToad: tried, was on par at best
CameronWatt: Lol =+ != += how did i do this
actualCoderTrevor: Zenoscave thanks that gets me a lot farther but I still get a weird "?RED?Player 1 has not provided an action in time.?RED?" on round 56 with my own script. I gotta go for a few minutes but wanted to let you know I tried it. It's definitely an improvement.
Zenoscave: It means ou timed out
Zenoscave: that's on you
ddreams: or you crashed
Numby: I think my code's timing out on garbage collect. When running it locally it can go to 1GB~4GB memory lol
VizGhar: wow o.O
Numby: Like each turn is consistently taking 65ms then I just time out
jacek: your code is garbage eg
ddreams: you have a max of 700mb ram, so perhaps it's killed
CouscousAgha: boss is at 49.63
CouscousAgha: :D
aCat: Nomby in what language you write?
aCat: Numby
aCat: you can prealocate memory and use factories
aCat: or sometimes even turn GC off
Numby: In java, I think my search tree is gobbing up memory. I've been re-rooting my tree if opponent selected a turn that's in the tree
Numby: So I can keep a part of the search tree that's relevant but I guess eventually the parent portion of the tree gets GCed and I time out
Numby: At least that's my theory...
Illedan: :wave:
Astrobytes: hey Ille
ddreams: Hei Ille
UndercoverToad: did you sleep since yesterday Illedan?
Illedan: Didn't do anything on this
Illedan: But I'll start now
Illedan: Tomorrow I don't have to wake up at 7
Enknable: Numby I think you wanna preallocate (pool) your states at the beginning. Then just keep track of which state are in the current tree
dbdr: gogogo Illedan
Illedan: I have a wierd idea. Wanna see if it pays out :)
UndercoverToad: no, C#C#C#
dbdr: cheese?
Zenoscave: rust
Illedan: I tried to fill the board with seeds. Didn't help too much
UndercoverToad: weird, not wierd. every single time you do this ;)
Illedan: I need a full rewrite
Illedan: Yeah that word is weird
dbdr: I thought about it, but must be too slow/expensive
Illedan: Fun though, enemy had no good actions in the last 3 rounds
Illedan: xD
dbdr: game is well balanced, that's why it's hard to eval
dbdr: got a replay Illedan?
Illedan: I guess you can do it on highly blocked maps?
Illedan: nope
dbdr: possibly, yeah
kovi: but how can you afford it
struct: arent seeds a bit expensive for that?
Illedan: https://www.codingame.com/replay/556057454
Scarfield: shadows galore
actualCoderTrevor: Zenoscave yeah it seems spamming stderr is a lot more costly on my local machine than in the IDE. I removed the spam and now I was able to see a game to completion. \o/ Thank you so much!!! p.s. I finally realized ?RED? is just so the IDE makes the text red, lol.
Zenoscave: stderr is very costly
Illedan: Anyway, off to code something working. No sleep until euler beaten :nerd:
UndercoverToad: https://www.extremetech.com/wp-content/uploads/2012/09/simpsons-mr-burns-blocks-out-the-sun1-640x353.jpg
Nezinau: Hey, could you give me a hint what to think about in pod racing (coders strike back) when you get to pods colliding stage?
reCurse: Negative triple velocity
UndercoverToad: you don't have to think about collisions until you reach gold
UndercoverToad: finding an efficient path is the most important in lower leagues
MSmits: dont some people ignore collisions until legend?
MSmits: or is that not possible?
Astrobytes: Even in legend, yes
Nezinau: thx, will work on my angles and boosts then :)
YannT: MSmits plz don't block my legend push :(
MSmits: well if my bot does well vs boss, i might pull him down for you
MSmits: I did for renard earlier. I got to 6-1 pulling boss down, he promoted, then started losing :P
YannT: I have a great winrate vs the boss too, I just can't get to it in a push :/
MSmits: oh ok
YannT: grr I just seem to finish around top10, I'm just if I can't get there I'll pass
MSmits: thats annoying Yannt :(
MSmits: keep submitting
YannT: I'll apply the kovi method until it works I guess :D
MSmits: maybe it pushes your competitors out of the league
MSmits: like me :)
Noyotens: https://www.codingame.com/replay/556066418
cegprakash: I don't know how to test a depth > 1 its so damn hard
Noyotens: the AI broke my code
Noyotens: at 48/138
Noyotens: https://www.codingame.com/replay/556066418
VizGhar: hey C++ devs... BS... how you store nodes? in vector, array, or something else?
Michael_Howard: Submit is definitely getting faster.
Lysk: arrays
struct: array
kovi: YannT: today i have only submitted 2
MSmits: not like you had a choice kovi :)
kovi: valid
dbdr: :D
YannT: kovi: oh man, you're slipping ;)
CameronWatt: all ur base are belong to us
AllYourTrees: sup
Astrobytes: lol AllYourTrees
AllYourTrees: :wave: :D
Zenoscave: hey astro
Razielwar: top are all in mcts?
Astrobytes: pewpewpew zeno!
Zenoscave: PEWPEWPEW
MSmits: doubt that Razielwar
MSmits: but there might be some
AllYourTrees: aren't they all beam search?
MSmits: many probably
MSmits: but not all
MegiAKAHammatopvcwchat: hey jojo fans
Zenoscave: what's jojo
AllYourTrees: oh shoot... "You reached the limit of plays for a period of time"
AllYourTrees: didn't even know that could happen
Razielwar: ahah
Astrobytes: They lowered the threshold recently, very annoying
reCurse: Just submit instead kekw
AllYourTrees: :scream:
AllYourTrees: does anyone know the time period/game limit?
VizGhar: MSmits going legend :) I'll need one more year probably to get there
Astrobytes: Quite A Few, In A shorter Period Of Time Than Previously
AllYourTrees: damn just what i feared
PRO_G_r_AM_m_ER: almost beat silver with my 50 lines of code that does something :(:)
MSmits: VizGhar this one might again get stuck right under boss =/
VizGhar: how long are you working on solution smits?
MSmits: what do you mean
VizGhar: well 20 hours? 60?
VizGhar: I dedicated huge amount of time... but it seams I only learned how to code in C++ :D
MSmits: ohh, well most of the week when i wasnt working
Lysk: MSmits "Sim perfect", so much trolling ;)
MSmits: :)
BrunoFelthes: What Miyaza means?
Zenoscave: persons name i believe
Magus: Miazaki
Twelve0: name of gold boss?
MSmits: Miyagi?
Magus: https://en.wikipedia.org/wiki/Hayao_Miyazaki
Magus: (the gold boss name)
Zenoscave: sorry msmits...
Scarfield: looks like a solid submit PEWcave
Zenoscave: I have a good w/r against most of league except boss... haven't been able to test much due to timeouts
Zenoscave: Ope lost to smitty
Zenoscave: YW MSmits
MSmits: thanks :P
Zenoscave: Won against boss!! MSmits i'll drop em for you
MSmits: i just lost against boss
Zenoscave: well don't do that
MSmits: i just won against you
Zenoscave: Don't do that either
MSmits: got it
ItsNotABug: I got my first win against boss. But then got beat when i did a new game :/ so im taking the positive
AllYourTrees: i cannot figure out when a good time to complete a tree is
Zenoscave: :D
MSmits: hold your ear against the trunk and wait
MSmits: until it says
MSmits: "You complete me"
AllYourTrees: oh duh yeah i never read that value from game input, ty!
Zenoscave: lol MSmits that's enough weed for one day
DevLorenzo: Hi, has anyone completed level 1 of the coding escape game?
dbdr: how poetic MSmits :)
VizGhar: nice :D
MSmits: I can do poetic :)
Astrobytes: But don't give up the day job just yet :D
MSmits: It'd be poetic if i can finally get out of gold
Zenoscave: what about perl poetry
Astrobytes: ^
MSmits: i saw that
MSmits: whats special about it?
Zenoscave: it's perl. and poetry
Zenoscave: duh
Astrobytes: It's poetry, in perl
MSmits: I see
MSmits: you make good points
Zenoscave: two losses to boss :/
Astrobytes: chomp die
MSmits: damnit
Westicles: The kids these days are into hash poetry
DevLorenzo: I'm stuck to find the number of universes
Westicles: Write a poem that sha256 hashes with the most consecutive eights or something like that
Zenoscave: ... why
MSmits: Zenoscave do you have +50% vs boss?
Scarfield: heightku
MSmits: i have no idea what i have
Zenoscave: no idea either
Zenoscave: too many timeouts in IDE
MSmits: ah yeah
MSmits: I got only 1 match vs boss this submit
Zenoscave: you've got a 67% against me currently
Zenoscave: I've gotten 3
MSmits: top gold is all very close in rating
Zenoscave: Makes it a real barrier
MSmits: yeah i guess so
Scarfield: i think it was ille who said when he finally made it to legend he got straight to no. 40
YannT: I failed around top10
YannT: I'll repush with incremental changes until I make it to legend in a stroke of luck :>
Zenoscave: YannT perhaps wait a bit. so many concurrent commits we just squabble and no one promotes
YannT: it's okay at current pace I'll get there in 2 hours anyway
Zenoscave: lol true
YannT: also I think I have a decent winrate against you, I saw my bot stealing wins here and there :D
YannT: I need your ELO points for the greated good (me going legend), sorry
Zenoscave: NOOOOOOOOO
Zenoscave: not my TrueSkill
YannT: :joy:
DevLorenzo: Is there a way I can find help with the coding game escape game?
Zenoscave: DevLorenzo there's an active competition. That'll be the main focus til monday here
CouscousAgha: i think the only way stopping me from getting to legend is to figure out when to stop cutting trees
MSmits: zeno, i need 0.35 to reach boss.
MSmits: get on it
MSmits: 0.24 now
Zenoscave: I need you to quit beating me. you're most of my losses
MSmits: most of my losses are the green camel
Zenoscave: That's the second most for me
ddreams: Hi Zeno
kovi: told you to stay wit 0.09 diff
MSmits: i had a new version to test
Zenoscave: Once his submit finishes I'll probably have better luck
Zenoscave: .14 below rn
MSmits: on my last submit, when i was around your % I was 2 pts ahead of boss =/
Zenoscave: Yeesh
Zenoscave: What happened/
MSmits: i started losing
Zenoscave: Already doing that
YannT: then just kovi through it man :)
MSmits: well i meant on that submit
MSmits: i am now too yes
ddreams: I have an idea... hmm.. bbl
MSmits: I will try to kovi through
Zenoscave: ddreams, no ideas!
YannT: :D
Zenoscave: only dreams
MSmits: or maybe someone can just kovi boss down to my level
Zenoscave: I'm trying
YannT: it's pretty low right now, it seems to stay between 50 and 49
Twelve0: hahaha
MSmits: well you're wrecking me currently zeno :P
Zenoscave: Sorry msmits....
MSmits: ah well
Zenoscave: Skot_z is my only issue currently
MSmits: yeah i'll be glad when he is gone
Zenoscave: It'll make submit much easier next time for you ;)
MSmits: possibly
YannT: I repushed a version with tree reuse between turns
Zenoscave: the camel has a 82.5% w/r against me
YannT: little hope it does anything good
Zenoscave: < .01 between me and boss...
PRO_G_r_AM_m_ER: now i somethimes beat silver lol :cry:
Zenoscave: I need to shut my mouth. I keep losing when I talk
MSmits: or maybe because you dont use all caps
Zenoscave: PEWPEWPEW
YannT: worse case you'll just land riight next to it and netflix until you get promoted man
TobiasA: i just got to silver league
TobiasA: woooohoooo
MSmits: gj
TobiasA: i am the winner of the entire game
TobiasA: because if you didn't know silver is the new legend
ddreams: gz :)
TobiasA: now i don't think i can make it to gold before monday but i'll try to make it to top 500
skotz: sorry zeno :)
Zenoscave: resub smitty?
MSmits: yeah
Zenoscave: any changes?
TobiasA: wait nothing is different?
TobiasA: in silver?
MSmits: lowered beamwidth
Zenoscave: TobiasA no changes after bronze.
YannT: I like how your bot says sim is perfect and then it loses
MSmits: well sim was perfect
YannT: :D
MSmits: the search wasnt though
Zenoscave: what does "perfect" mean here
Astrobytes: It doesn't say search was perfect :P
YannT: "II perfectly simulated this defeat" *explodes*
reCurse: "All went according to plan, I lost"
MSmits: it's just applies all possible actions to current gamestate and my own chosen action. Then keeps that list till next turn and checks against new gamestate
ddreams: muhahaha
MSmits: if one is exactly the same, then the sim is perfect
Zenoscave: Gotcha
reCurse: "My author has purposely programmed me wrong, as a joke"
Zenoscave: lol recurs
Astrobytes: hehehe
YannT: yoo tree reuse seems to actually bring something
YannT: push is doing good
MSmits: you cant reuse them
MSmits: when you complete them they're gone
Zenoscave: Baduntiss
Astrobytes: dearie me :D
YannT: I'm sorry, in your words I meant "sim perfect" MSmits :)
MSmits: :)
Zenoscave: I have a sim perfect prune array
MSmits: oh noes
MSmits: noone is above boss now
YannT: jokes aside I get it, my tree reuse I just pushed is essentially the same thing
YannT: look in child nodes for the move opponent played
YannT: if found, great
MSmits: yeah i use that in boardgames
YannT: if not, sucks :D
dbdr: brutaltester does not support tournaments (more than 2 bots in a 2P game)?
MSmits: it does
MSmits: oh wait
MSmits: no
Zenoscave: dbdr no
MSmits: 3p 4p games though
dbdr: ok
Magus: brutaltester can handle to 4 players
Magus: (if the referee accept 4 players)
MSmits: i looked for that option too
Magus: but there is no "tournament" thing
MSmits: i now manually manage tournament thingy in excel
Magus: But .... I like the idea :D
Magus: maybe after this contests I'll work a little on brutaltester
Magus: I want confidence interval in the output
reCurse: Mortal Tester
Magus: and tournaments seems nice
Counterbalance: grats zeno
dbdr: esp as api benching is almost dead
Zenoscave: YannT don't you dare
YannT: coming through
MSmits: uh oh
Zenoscave: Counterbalance not done yet
YannT: (hopefully)
Counterbalance: no but it's a good sign
MSmits: you would think so
Zenoscave: I feel like that was a jinx
Counterbalance: :]
Zenoscave: -_-
Magus: reCurse: fear my skill to find project names ! My local "training arena" is called Abathur. And my tool to parse/analyse my replays JSON is called Overlord :D
reCurse: I still think Mortal Tester is the most appropriate name to have one with tournament feature
MSmits: Overlord is a good name for a cat
reCurse: ^
Zenoscave: All cats are named that
Magus: my wife refused my names proposals for our cats
Zenoscave: we just don't know it
MSmits: what were they Magus
Zenoscave: Please tell me one was "horse"
Magus: MSmits: Stackoverflow and Jira
MSmits: lol
Magus: According to normal people, they are not good names
Counterbalance: arena tester name suggestion: Battle Royale (wrong author though)
reCurse: According to everyone
Astrobytes: wtf man
Magus: But, I have named one of our rat Mutex
reCurse: I was hoping for something funny
Magus: It's still something, I suppose
Astrobytes: Oh you have rats?
YannT: why do you hate your pets honestly
jacek: and he locked himself in
Astrobytes: Rats are awesome
Magus: I love my pets :D
AntiSquid: https://youtu.be/kAJSA_f1_gY?t=11
YannT: :joy:
Zenoscave: Mutex is good. I like it
MSmits: wait, are these rats pets magus or does your house have issues
reCurse: Why not both
Magus: pet rats
reCurse: Make the best out of a ratty situation
Magus: 8 females, 7 males
jacek: you should name your cat alfador :thinking:
Astrobytes: Wow, you have a whole bunch of rats!
Magus: (actually, 15 rats, 2 cats, 2 parakeets)
YannT: Magus: so that's why you needed thread safety, because of too many concurrent rats?
Astrobytes: Do they get along with the cats or they're not allowed to mix?
reCurse: Must smell nice
Magus: Cats never tried anything with the rats. They sniff them and that's it.
YannT: Zenoscave: sorry mate :/
Magus: Rats want to play with the cats, but the cats don't want :D
Astrobytes: hehehehe
Zenoscave: 9 loss streak :/
MSmits: damn
MSmits: I feel your pain
Zenoscave: Right when Counterbalance said gratz
YannT: took a in on the boss
MSmits: them Dutch people always sucking the fun out of things
Zenoscave: lol MSmits
YannT: bruno is kicking my ass though
Counterbalance: muhaha :grimacing:
Astrobytes: Sucking the smoke out of spliffs you say MSmits
MSmits: that too
Zenoscave: you are the literal devil counter
MSmits: thats the devil, only he can read
Zenoscave: read what?
MSmits: things
Astrobytes: Zeno was getting high, then the Dutch came along and smoked the rest before he had a chance.
MSmits: he oh wait thats literate
Zenoscave: neato
ItsNotABug: Was ~860 in gold. New submit is already top 300 after the first 10% :D
Zenoscave: Zeno doesn't smoke
MSmits: because we took em
Astrobytes: It's metaphorical!
Zenoscave: It was the drink that killed me
YannT: get outta here already BrunoFelthes you're cramping my style :p
Zenoscave: I was hit by a guiness truck
MSmits: I'm on my way back up, hopefully this one has a better shot
Zenoscave: I should eat today
Zenoscave: I need a break
BrunoFelthes: ahuahuahua
ItsNotABug: fack. Another couple games and its dropped me back to ~500
MSmits: or just some day
BrunoFelthes: YannT you push me up
MSmits: nice one BrunoFelthes, grats
MSmits: now get out of our way
YannT: I noticed :p
Ajay.komirishetty: Hey Gang
Zenoscave: see this is when you say gratz. He has finished the submit
YannT: go on, promote, I still have 50% and I4d rather you not be there :D
Zenoscave: gratz BrunoFelthes
ddreams: gz BrunoFelthes
Zenoscave: HAHAHAHAH INSTA KARMA YannT
YannT: ah shit :D
MSmits: 16th now and it's less than 2 points to boss
MSmits: so close all
Zenoscave: MSmits quit it. I am at 95% no beating me
YannT: it's okay, might as well take the L's at 50% and kickass on the tail end of the push
TobiasA: the moment you realize you are the only one from your country in a contest:cry:
MSmits: I am hoping for some more boss battles, only had 2 last time
ddreams: you're number one, TobiasA
MSmits: both losses
JRFerguson: i feel you Tobias
JRFerguson: Uruguay here, always happens to me
TobiasA: i am only one not number one
BrunoFelthes: ahuhauahu, TobiasA, Brazil is a very big country, and we are not so many people here
jacek: uruguay eh https://www.youtube.com/watch?v=4paLVu0gNcQ
Astrobytes: think you're best to wait for a push Zeno
JRFerguson: lol jacek, i knew that Simpson bit
Astrobytes: Otherwise this will be the luckiest submit in the history of today
ddreams: I've lost 1 out of 8 vs. zeno
TobiasA: haha that video
JRFerguson: anyone has any good tips on how to simulate opponent ?
TobiasA: this is my first contest ever
Zenoscave: I'll just wait for push
JRFerguson: i just ignore opponent
TobiasA: in anything
Zenoscave: who keeps losing to bos?
elderlybeginner: i don't even sim my turns or board
Zenoscave: boss*
MSmits: I lost 1 :(
YannT: not me, I went 2-1 against boss
YannT: and now I fell :(
YannT: I blame Bruno
ddreams: 2-1 against boss too
Zenoscave: 6-7 heare
ddreams: Though I think I've never lost against YannT, my bot is his kryptonite
Zenoscave: I know that's not a word. I know I can't spell shush astro
MSmits: are you gonna wait Zenoscave?
Zenoscave: yeah msmits
MSmits: probably good
TobiasA: those numbers in the leaderboard what are they?
Zenoscave: I'm hungry and need food
Zenoscave: Trueskill rank TobiasA
MSmits: ranks and ratings TobiasA
Eglerion: Any tips on how to eval game state?... I tried tons of different things over the last hours but my sim still can't beat my initial submit with hardcoded focus on seed, grow, complete in the middle of the board..
YannT: can all og my kryptonites please promote switfly and clear the way? :p
TobiasA: the decimal ones
TobiasA: with arrow beside it
Zenoscave: yes TobiasA trueskill ratings
ddreams: Trying, YannT :D
elderlybeginner: egl
TobiasA: oh
peerdb: Eglerion i have the same issue :(
Zenoscave: bah
elderlybeginner: Eglerion - fix the main bug :joy:
MSmits: oooh beat boss!
JRFerguson: Eglerion: score, points per tree, points per lvl 3 tree, wait till the end and count sunPoints for each player
JRFerguson: congrats MSmits
MSmits: then lost to the tric :(
elderlybeginner: which means: don't seed in your possible shadow
MSmits: no i mean 1 game lol
MSmits: almost zeno
Ra777: Hello, anyone knows what is the limitation for games per time period in codingame?
AntiSquid: yes
Ra777: do you know what that limit is?
AntiSquid: yes
ddreams: I do too
MSmits: can you tell him what the limit is AntiSquid
ddreams: or no
ddreams: you mean in the IDE?
casmith789: he can
MSmits: :P
MSmits: jinx me Counterbalance i am above boss
Counterbalance: gratz
MSmits: :)
MSmits: it really works
MSmits: my next game was loss vs boss
AllYourTrees: oof
elderlybeginner: MSmits, I think I saw you over the boss yesterday :joy:
AllYourTrees: i feel like your next submit will happen
ddreams: he was over boss earlier too
VizGhar: gogo smitsy
dbdr: MSmits was born above the boss
Counterbalance: ra777 look at the bottom of the statement (and next time ask what you want to know, not if someone knows it - these people take everything quite litterally here :))
VizGhar: you dont know how it feels dbdr :D
dbdr: what?
VizGhar: defeat boss... you were always step ahead of boss
Ra777: than you, Counterbalance ;)
dbdr: I know it from previous contests
dbdr: when legend was legend
dbdr: and I was a noob
aCat: oh, Smits on the run ;-)
DomiKo: Do you enjoy that game guys?
YannT: I'm just a few luclky wins away now :)
Astrobytes: when legend was legend. Long live legend.
VizGhar: I'm noob that cant reach legend yet :)
YannT: don't kick me down MSmits plz
VizGhar: My algo is too slow to search enough space and I can't write correct eval
YannT: almost there
VizGhar: YannT are you running BS too?
AllYourTrees: MSmits see i told you
Astrobytes: I never got legend in a contest. That doesn't matter though. What matters is that Legend is supposed to be pretty fking hard to get to.
YannT: MCTS
elderlybeginner: There is a serious battleground at the top :joy:
VizGhar: ah ok :O I probably just don't know how to write fast algos in kotlin (main language for me)
YannT: it boils down to don't no new in your algo
YannT: new is bad
YannT: pre-instanciate everything you can
VizGhar: hmm can you avoid garbage collection?
MSmits: damn, Harry is back
MSmits: VizGhar dont make garbage
ddreams: beating me too
YannT: if you pre-instanciate everything and then don't do new in your algo, you can avoid it mostly
YannT: GC is bad, it'll time you out
Nerchio: VizGhar what YannT said basically, use 1st turn to create most objects you need (like States/Nodes for search) and then never create new again
Nerchio: but even if you optimize everything as best as you can it will probably be 2x slower than C++ lol
VizGhar: this is completely different world then standard app development :D
MSmits: very true
YannT: it is
ClockSort: looks like you did it, @MSmits. congrats
Nerchio: true but maybe thats part of the fun
Astrobytes: Yes.
YannT: also don't do .filter or other things that will instantiate lists
YannT: very bad as well
MSmits: eh ClockSort, last time i was this far ahead in my submit, i was +2 rating vs boss and still lost
YannT: use .forEach and .repeat a lot instead and if (...) return@forEach/repeat
MSmits: at least battles are going fast
VizGhar: I understand the concepts behind all algos, but i like nice and readable code
ddreams: but the previous time I was winning against you, now you're better
YannT: bot coding contests are not the place for clean code :)
YannT: do it dirty
YannT: everything public
Astrobytes: ^
Nerchio: :scream:
VizGhar: too late :D I'll do it next contest
YannT: dirty code is faster :)
XeroOl: if you're in c++, you can just hide the dirty in a function
XeroOl: and let the compiler inline it
VizGhar: slowly getting there...
MSmits: damn harry :(
YannT: it's too crowded for my bot up there :(
AllYourTrees: anyone got advice for how to tell what you did wrong in a game
Astrobytes: If you lost, you need to...
MSmits: look at how completes are spaced
Astrobytes: FIX IT
MSmits: the yshould be mostly in the second half
MSmits: mostly spaced, not all at once, so you keep a steady income
ddreams: cut all your trees
MSmits: spread seeds so you dont shade your own trees
MSmits: sometimes i have a tree at the end of the game
MSmits: uncompleted
ddreams: yeah, was mainly a jab at his nick
Astrobytes: That can help in a tie MSmits
MSmits: but it's richness 1, so takes 4 suns, gives 1 score, not worth it
MSmits: right that too
Chainman: what's the strategy to beat silver boss? I'm still losing
Nerchio: it can be worth it msmits
MSmits: how
MSmits: in the last turn, if the tree gives 1 point
Nerchio: lets say you have 30 sunpoints
Nerchio: -4 you have 26
MSmits: ok
Astrobytes: if you still have more suns than opp but just need a point
Nerchio: equal on points
Nerchio: u have +2 additional
Nerchio: you can throw a seed now
Nerchio: and win a draw
Nerchio: or something
ddreams: ?
MSmits: mmh doesnt make sense Nerchio
ddreams: no
Nerchio: ok
Nerchio: you are right
Astrobytes: see my comment
MSmits: i dont see that happening Astrobytes, how is that possible
MSmits: 3 suns is 1 point
MSmits: you lose 4 points to complete
ddreams: ok, so my first implementation of the idea gives me 100% losses, nice
MSmits: 4 suns
MSmits: so you always lose at least a point
MSmits: and now you also lose a tree
Astrobytes: Yes, it's not likely
MSmits: so how is it ever worth completing a tree that gives 1 point
Nerchio: so how about you have 30 suns, sell 1 tree, you can upgrade now seed to level 1, throw seed for 0 and throw seed for 1
Nerchio: but its like a super corner case
Astrobytes: It would need extreme conditions in my case too
MSmits: why does selling a tree give you a chance to uprade a seed?
MSmits: you can do that anyway
Nerchio: because normally you have 30 sunpoints / 3 = 10 points
MSmits: yes
Nerchio: you will lose a point if you do anything
MSmits: sure
Nerchio: hmm yeah
Nerchio: that probably doesnt work
MSmits: give me 1 exact example,, i dont think it exists
Nerchio: well i coded in my bot to never sell these tbh
Nerchio: when nutrient 1 and richness 1
ddreams: so instead you have 1 extra tree point and 26 sunpoints, giving you one less tree, but 2 suns to play with hmm..
MSmits: last turn, big tree on richness 1, no nutrients, there exists no example where it is good to complete this
ddreams: right, you lose 2 points
ddreams: 2 sunpoints
ddreams: so never good
Nerchio: ok so if u have 32 sunpoints
MSmits: you will still lose 1 point and 1 tree
MSmits: and gain 1 point from complete
Nerchio: true
MSmits: so not worth it
Nerchio: ok you win
Nerchio: i always forget that u sell tree so you have 1 less now :D
dbdr: rhichness 1 is 2 points
MSmits: i dont win , am on loss streak :(
MSmits: richness 2 is 2 points
elderlybeginner: I wonder how you guys will fit in legend now
MSmits: i think
Nerchio: 1: +0 points.
dbdr: ok, I use base 0 :)
CameronWatt: CoC is so addicting
MSmits: wait nvm
jacek: richness 1 is 0
dbdr: 0 1 and 2
dbdr: and -1 for empty
DomiKo: so -1 base?
Astrobytes: lol
Nerchio: yeah i thought it was weird i actually noticed that richness doesnt equal points in gold or something
dbdr: well, those are just ignored :P
Nerchio: why doesn't it equal points? :P
MSmits: me vs harry 4-9 this hurts
MSmits: well he'll be gone soon
elderlybeginner: Volt is back :joy:
ClockSort: lol msmits why are you still golden
elderlybeginner: for a moment only
MSmits: an existential question I ask myself
dbdr: :popcorn:
Astrobytes: Goldsmits
elderlybeginner: yep, nice to watch that competitive fights :grin:
AntiSquid: because the game is balanced
VizGhar: hmm wasnt Voltmach above boss?
ddreams: for a little while, yes
ddreams: for beating me so much
ddreams: MSmits looking great, jinxy kinxy
MSmits: nooo dont jinx now :P
MSmits: 1 more battle
MSmits: yay!!!!
ItsNotABug: winner
ddreams: nice, gz :)
VizGhar: clap clap
YannT: ah time to repush, way seems to be clear :p
MSmits: there's still harry
MSmits: well not for long
YannT: he'll be gone by the time I get there
MSmits: true
YannT: 10/10 first batch
YannT: good omen :)
MSmits: 4/10 here :)
Chainman: how good is it to try to just place size 3 trees in the middle and not harvest till end of game
MSmits: its bad
ddreams: well, I'm going to resubmit too.. the discussion of richness point made me realize I can't read
BrunoFelthes: bad
Chainman: really?
MSmits: Chainman you shade yourself
YannT: well it's good to go silver, even gold
Chainman: but won't you shade more than yourself potentially
Chainman: Yeah I need to get gold for context haha
MSmits: yes, but with opponent the shade works both ways
YannT: make best trees worth most points in whatever manner, gets you to gold
Chainman: I'm trying just gold worthy strategies
MSmits: with you, it works doubly on yourself
MSmits: so dont shade yourself
MSmits: watch top games, you'll see thats the most important thing they do
YannT: yeah for gold just grow as many trees as you can on richest spots, harvest at the end
YannT: that should do it
MSmits: to get to gold
Chainman: oh
MSmits: not to get to legend
YannT: he wants to get gold no?
Chainman: yes
tobk: dang... my bot was Gold 400, now after resubmit only Gold 750... of course that last submit took 9 hours, so maybe there's just 300 guys who passed me in that time? :-/
MSmits: oh ok
Chainman: I want to get to gold first.
MSmits: tobk it could also be that the guys who were submitting were not at their proper rank
MSmits: it takes time to move up the ladder
tobk: MSmits Yes, that's what I thought after the 9h submit, but in fact I resubmitted since then and ended in pretty much the same spot. Maybe that first submit was just super-lucky...
XeroOl: how are people so far with java
MSmits: ah ok
XeroOl: java seems like it would be a bad language for contests
tobk: XeroOl Why?
ddreams: java has been a bad language for everything since C# came out :p
XeroOl: because you're forced to deal with the memory allocation stuff
XeroOl: they don't have structs, so the moment you try to abstract anything, your code runs slow
Astrobytes: You have to be clever about it
YannT: java is a decent language for contests
YannT: you just can't do whatever you want :)
MSmits: XeroOl you make all your objects in the first 1 second
XeroOl: unless you start by just allocating giant arrays of primitives
XeroOl: but that sounds painful
XeroOl: you might as well use C
MSmits: thats normal
ddreams: you can allocate everything you need
MSmits: I do this in C++ and C# also
YannT: no you don't do it for priimitives
ddreams: initially
Astrobytes: JVM warmup
YannT: primitives are already cacched in the VM
ItsNotABug: My bots capable of beating msmits and gold boss. But it still ranked 360th in gold :/ Too inconsistent to rank high :(
Astrobytes: allocate a bunch of objects instead
MSmits: ItsNotABug capable of beating msmits or capable of beating him 50+ out of 100
XeroOl: that just seems not ideal
XeroOl: hmm
YannT: "sim perfect"
MSmits: lol
MSmits: it makes the creature look kind of arrogant doesnt it
Astrobytes: Do some calculations in the first second. Helps with JVM warmup.
YannT: ddreams: plz promote I have a horrible winrate vs you
ddreams: Trying :p
ddreams: you can just push me over, and then resubmit !
ddreams: I'm at 49.13 and boss at 49.25 :p
ddreams: If only Harry would stop beating me down
Astrobytes: Submits slowed down again?
MSmits: harry is annoying\
theycallmedavid: How does the response time restriction work? I'm measuring my response and in the worst case it is 250+ms, but only once in the middle, and multiple times it is a bit over 100ms. It looks like it is not a problem.
YannT: ddreams: grah I was hoping you'd get pushed you were equal to t he boss :(
YannT: why u stiill here man
ddreams: Harry is at 97% tho
Nerchio: theycallmedavid start calculating time after first input and the end before output
ddreams: Harry is gone legendary
theycallmedavid: That is how I measure
theycallmedavid: Wait, no
theycallmedavid: Ok
ddreams: you timed before first read?
theycallmedavid: Yes
ddreams: then you time the opponent too
theycallmedavid: I see
theycallmedavid: Now it is much better
theycallmedavid: Thank you
ddreams: chat is helpful
Chainman: ahhh I can't beat silver boss
ddreams: post a replay, chainguy
Chainman: https://www.codingame.com/replay/556171635
ZarthaxX: gz MSmits!!!
AllYourTrees: did you get it???
ddreams: Chainman try to spend more of your sun in the middle game
MSmits: thanks ZarthaxX :)
AllYourTrees: gz MSmits
MSmits: ye i got it finally
MSmits: 3 submits where i went above boss, only last time i stayed
Astrobytes: grats MSmits!
Chainman: okay I"ll try
Dakkad: std.error is behaving odly for me for the first time... any known problem/bug?
MSmits: thanks astro
Astrobytes: Please state the nature of your stderr emergency Dakkad
Dakkad: I wouldn,t say emergency, but here goes:
Dakkad: I have a console.error('lol') on line 502 that shows up on every round, except for the round where line 503 throws an exception.
Dakkad: I usually can rely on the previous line of code to debug the following line of code.
ddreams: flush?
ddreams: dunno how js stderr works
Astrobytes: Flush issue I think yes
Dakkad: This time, when I get this exception thrown on line 503, no prior console.error() shows up
Dakkad: What is a flush issue :-) (something I can google?)
ddreams: often output is buffered for performance reasons
ddreams: flush means to make sure it's sent to the destination
Dakkad: What can I do about this problem?
ddreams: google
Astrobytes: What are you coding Dakkad?
ddreams: or wait until a js-person with knowledge chimes in
Westicles: How does printing lol give you more information than it not printing?
Dakkad: It's just to simplify the conversation with an irrelevant std.error Westicles.
Chainman: wooo finally won 135-124
Dakkad: i.e. this guarantees there's no error in my console.error() :p
Westicles: Oh, well the other option is the error was on line 502
Dakkad: @astrobytes, I'm coding the spring challenge 2021 in javascript
Astrobytes: Are you coding for the contest or something else Dakkad?
Astrobytes: Right, gotcha
Zenoscave: I got pushed!!!!
Astrobytes: Is there a relevant code snippet (NOT full code) you could post?
Twelve0: congrats
ddreams: Nice, gz Zeno
Astrobytes: PEWPEWPEW Leg End!
Zenoscave: Thanks :)
elderlybeginner: I think I'm loosing more when there are richness 0 cells. Any idea how to modify action scoring taking into account that factor?
Dakkad: @astrobytes: http://chat.codingame.com/pastebin/91447232-ffa9-4ba2-a7a2-79b4adf27011
ddreams: boss at 48.90, I'm at 48.85 :p
ymoukhli: congrats twelve
Twelve0: not yet
ddreams: don't jinx it ;)
YannT: I'm coming please be nice
ymoukhli: spoke too soon
Counterbalance: Dakkad what's the exception message?
Dakkad: TypeError: Cannot read property 'size' of undefined
Counterbalance: f.tree is undefined
Dakkad: Yes yes, that's obvious :-)
Twelve0: thank you anyway ymoukhli
Dakkad: I know what the error means, but I'm trying to use console.error() to find out what object it receives
Astrobytes: I hope that ` is a typo :P
YannT: ddreams my dude, don't knock me down
Dakkad: it was me trying to format for discored :-)
Dakkad: There are no backticks in the actual code
Astrobytes: :D
Twelve0: sorry YannT
Westicles: that ` is spider poop on your screen
Dakkad: every forest object has had a tree in this array for hundreds of runs... I messed up something about the array, but can't find out since console.error() doesn't work on that specific round.... :-(
YannT: it's okay I still have 50% and youi guys are gone soon :D
Astrobytes: It's on the other side of the room now Westicles
YannT: I seem to have a decent winrate on Valphyr too, that helps
Zenoscave: Man this is a dog fight in legend
Westicles: Oh good Astrobytes. Probably laying eggs
Zenoscave: Who was gold boss?
ddreams: Bernard_Lowe
Astrobytes: Westicles: Probably waiting to lay eggs in my ears or hair I'd say.
ddreams: lol, both me and the boss have 48.94
Astrobytes: Cats will get it if it dares to come down from up there though.
YannT: ahh finally
YannT: bye ddreams
YannT: won't miss ya :p
Astrobytes: nice one YannT
Chainman: I'm not always winning against silver boss :(
ddreams: yay
Counterbalance: Dakkad ok, because that error is an uncaught exception, the process exits and stdout/stderr are not flushed. You can try wrapping the code in a 'try {} catch () {}'
Chainman: I'll submit to see if it is good enough to win 50% against boss
ddreams: 10 losses 3 wins in legend
ddreams: well, time to submit a better bot
Chainman: :flushed:
Chainman: I don't think my bot is going to get in gold with this submit
Chainman: But I might be closer to top of silver
Dakkad: Hey @astrobytes, @counterbalance. I removed all my console.error() and kept the one I absolutely needed to debug this, and they now shows up
Dakkad: as if I had been logging too much for the following rounds to keep on logging
Dakkad: I tried with a try catch as well (with all the console.error()s and it didn't work)
Astrobytes: Ah. If you see a ... at some point, you logged too much to stderr
Astrobytes: It is limited
Dakkad: Yeah, that was the case, but I thought this "max" was cleared for each round
ddreams: it's not cleared for each round?
Chainman: https://www.codingame.com/replay/556191535 He's on top in silver, he is just putting all seeds as closest to middle in the game. But he also uses sun points better mid game.
Chainman: He is harvesting quite early
ClockSort: i've been stuck at Gold 8 to 15 for the past day. That's enough. I'm going for it with a new submit, wish me luck! New feature: reuse search tree from previous turn as a start for the next search.
ddreams: Chainman I think you have too many level 3 trees
Zenoscave: good luck ClockSort
ddreams: I should probably have submitted this bot earlier
MSmits: grats Zenoscave
ddreams: 116 in legend now
Zenoscave: thanks MSmits
Chainman: wow nice
Zenoscave: doing gradient descent optimization on constants now
Zenoscave: Param tuning
struct: grats both
Astrobytes: nice
DomiKo: Opponent fitting would be better
Zenoscave: DomiKo what do you mean by opponent fitting?
DomiKo: Choose one player and fit params to beat him and upset him: D
Zenoscave: Lol fair
Zenoscave: I'm too low for that to have much impact though
Chainman: hmmm okay yeah with that change ddreams I seem to beat ksaw at least once of 2.
Chainman: He is still beating me, but I win sometimes now.
DomiKo: Itf funny how even my bot know that top player won in day 12 :D
Chainman: He takes so much of board throwing seeds.
Chainman: I was trying to only throw seeds 1 at a time, and only do it 2 distance from another tree.
Astrobytes: Oh, YannT, I thought you promoted earlier, my mistake :(
YannT: no I've bneen knocking on the door
Astrobytes: You'll enter soon I'm sure
YannT: been above the boss, it's really down to good/bad luck at this point
Astrobytes: Always
Chainman: 29 in silver, I don't know if this bot will promote
Chainman: I have a better version to submit already though so I think I'll get in gold.
Astrobytes: Don't give up Chainman
Westicles: You can do it chinaman
KelvinAndHubbles: found a wee bit of difficulty getting out of Silver for some reason though, hit top 100 4 days ago, and haven't moved much despite changes, probably just my botched code :/
Chainman: chinaman lol
Chainman: :joy:
YannT: I'mma kovi this crap
YannT: tune constant by .05 and resubmit until legend
Chainman: cool
Chainman: coming for bossooot
cegprakash: depth = 1 eval doesn't work for depth = 5 and depth = 5 eval doesn't work for depth = 1 how am I supposed to debug this eval
reCurse: "kovi this crap" that's a new one
Queuebee: does anyone have a rough max amount of actions to make by a player in any given turn?
KelvinAndHubbles: 3 or 4 usually works for my bot
Chainman: 4 people in my way to beat boss
Queuebee: I meant what could be the possible max so i can set a Action[size] for that but nvm ill just use list lol
Numby: Does code in the IDE get throttled if you run it a bunch of times? I've noticed I'm getting time out if I run it a bunch but after taking a pause it runs fine (using replay in same conditions between runs)
Chainman: wait when you are one spot behind boss you don't battle him?
Chainman: oh I passed the boss
Chainman: but I didn't atually have an battles with him
ddreams: as long as you are ahead with 100% completed games you'll promote
Chainman: crap I'm swapping places with him
Chainman: I'm 2nd to him again
Chainman: okay now I'm above by about .5 trueskill
Chainman: that should be good
Chainman: bossoot in java
ClockSort: my MCTS just loves cutting trees too much. Around turn 15 it might cut everything if it thinks the opponent can't catch up. Any advice?
BrunoFelthes: heuristics, fix it to not cut
Queuebee: Your MCTS can tell you what it thinks? O_o
ClockSort: here's an example. the percentage in Message is its chance of winning. It thinks 90% chance of winning after the forest is gone. https://www.codingame.com/share-replay/556211617
Queuebee: So why does it think that ^^
ClockSort: it doesn't see that the opponent can hold trees for a very long time afterward. in the random rollouts, the opponent cuts randomly sooner than the very end of the game..
BrunoFelthes: the of in your mesage is the number of rollouts?
ddreams: more sims or prune the actions perhaps?
ClockSort: @bruno it's the rollouts for the chosen move
mybk: are you adding the sunPoints/3 when comparing the win/loose ?
BrunoFelthes: i have a rule that i always have at least 3 tree 3 until turn 18...
ClockSort: @mybk yes
BrunoFelthes: it works for me
mybk: ok
ClockSort: @BrunoFelthes 3 large trees?
BrunoFelthes: yes...
ClockSort: that would definitely shake things up.
ClockSort: my bot chops way sooner haha
Chainman: how could I get started with MCTS? hmmm
ClockSort: read wiki article
ClockSort: it was a long learning process for me
Chainman: how long have you been doing it?
ClockSort: i started at the start of this contest and it took 3 days to get it working enough to get out of bronze
ClockSort: because i'm a noob :D
Chainman: Where are you now?
Chainman: I am too
ClockSort: Gold #23
BrunoFelthes: try first only monte carlo... than bean search, then MCTS, I think that this order is nice
Chainman: I tried a beam search but scratched it, so I can advance with heuristics.
ClockSort: i think search algs are the best way to improve as a coder
Chainman: finally promoted to gold wooot
ddreams: not bad
Chainman: lol 1404, I'm in the bottom
ddreams: still gold
YannT: I think I've got it :)
Chainman: maybe I'll go back to my search algo now.
YannT: fixed another bug
aangairbender: my mcts with duct cant beat my beam search
aangairbender: sad
Chainman: yannT about to go to legend
BrunoFelthes: until wednesday, I had a bug at my code because I was using a signed byte to store the score... So, my bot was tring to do a max of 127 points...
ClockSort: lol
YannT: maybe, no jinx plz :D
Chainman: Okay so this boss is in python3.
ZarthaxX: F BrunoFelthes
KelvinAndHubbles: I have positive winrate against like everyone in the top 50, but randomly losing to neesh like 1700/1950 rated in silver :/
ddreams: sometimes it's a bit of rock/paper/scissors
KelvinAndHubbles: yea, I have a problem with the "hoarders" as I call them who don't sack until the end. It's kinda a coinflip in those games
Stormalix: how do i even begin with coding a bot
ClockSort: anyone else get random "-1" results? like 1 in 100 games?
KelvinAndHubbles: yea, that means timeout error or some error ClockSort
ClockSort: i'm wondering if it's timeout or crash, any way to tell?
KelvinAndHubbles: you can send to IDe, but I've had it change a million times on me
ClockSort: i load in IDE and no crash or timeout.
KelvinAndHubbles: yea, either you, opponent have random or Codingame is just playing tricks with your head
ddreams: if you are usually far away from the timeout, it's probably a crash
mybk: timeout -> maybe memory issue
ddreams: if you're sometimes around 100ms normally, probably timeout
pmor: the memory thing has bitten me a few times.
mybk: me too
pmor: I now have my own allocator that I use for most of my containers
mybk: what do you mean ?
YannT: I think I've got this this time :D
pmor: i get a block of memory from the system, and then I have my own memory management of that block
pmor: you can pass the allcoator into vector, unordered_map, etc
mybk: ah cool, advanced tricks, cool, maybe I need to learn c++ (properly) for the next context, right know I am using java
pmor: yeah, this is a c++ thing. Java is a whole other beast with the GC
mybk: yeah it's a big problem
Zandy156: I enjoy making fast heuristic bots work as well as I can.
mybk: yeah I missed that, I enjoy that also, next context I will spend more time doing heuristics
Zandy156: I barely got top 2k in the potion contest because of it.
Zandy156: hopefully the next one isn't like that one
cegprakash: is 5 a good seed for srand?
Zandy156: are there bad seeds?
cegprakash: no idea
pmor: there isn't a difference in what seed you use.
pmor: if you use the same seed though, you will have the same results come through rand
pmor: i don't think anyone here is able to take advantage of that
casmith789: you probably want to use the same seed if you want to debug the bot when it loses to see what is going on
pmor: yeah, its a pain if you use a changing seed. I keep to one
Littleyounes: Codingame c'est une dinguerie quand meme
Littleyounes: sous coté
pmor: oui
somnek: xqcL
Pick8chu: does anyone uses brutaltester on windows with javascript?
Pick8chu: I've been working on it since yesterday still hasn't work
Pick8chu: cause of the readline
ClockSort: what kind of buffer against the 100ms are you guys using?
ClockSort: I'm on 90ms and I'm not sure if that's the cause of my "failure to provide a move in time"
ddreams: gz YannT
ddreams: now you can come push me up again ;)
pmor: I currently have 70 due to my getvalidactions call can take over 20ms
pmor: I'm adding in pruning to reduce the cost, which should allow me to up it
YannT: ddreams: I'm in front of you tho :p
ddreams: hehe, three spots, perfect for fighting
ddreams: well, I guess we both are not in submission, so
ddreams: and I'm heading to sleep
Wontonimo: ClockSort, you don't need more than 1/2ms of buffer. Make sure you start your timer after the first line of input you read
reCurse: There are server hiccups you can hardly prevent
JohnCM: all right, reached gold!
ClockSort: @Wontonimo that's interesting, i'm starting it too late -- at the start of my searching.
Pick8chu: does anyone know how to execute js file with spider monkey?
JohnCM: is the gold bot a heuristic bot?
Zandy156: I think so
JohnCM: it's a huge upgrade over the silver one. i'm losing it by 60 pts
ClockSort: recurse so this might not be my fault? https://www.codingame.com/replay/556304651
reCurse: I don't know it's hard to tell
reCurse: If you print the time taken in debug and it's always the same time
reCurse: And on a random turn it jumps to +15-20ms
reCurse: Then yes that's it
JohnCM: the gold bot is blocking all my trees, no wonder my pts so little
Kellthazar: its pretty well implemented
Zandy156: that's its main strategy
ClockSort: Gold #2, COME ON COME ON
Wontonimo: go ClockSort go!
MyCCODE: Can you do aa favor to subscribe to my channel - https://www.youtube.com/channel/UCEHp4w8fTDoOcRUwG8qTypQ
Wontonimo: no
Wontonimo: (secretly opens up channel and looks at all videos)
Wontonimo: Cool, unity dev. Did you compete in this challenge using C#?
MyCCODE: Yes, sure
Wontonimo: I don't see you on the leaderboard. You know there is a competition going on right now right?
NotSureWhyThisWorks: its been 5 days of learning c++ and my beam search can finally beat my python heuristics bot.
cegprakash: NotSureWhyThisWorks awesome
Wontonimo: congrats!
cegprakash: but are u sure why it works NotSureWhyThisWorks
NotSureWhyThisWorks: I'm so happy lmao
NotSureWhyThisWorks: Nah not sure still
Wontonimo: yeah, it's a great feeling!
cegprakash: NotSureWhyThisWorks NotSureWhyThisWorks
Lokmene: any escape room
cegprakash: lol
Lokmene: ??
NotSureWhyThisWorks: wontonimo did your mcts get you into legend?
Wontonimo: no, i couldn't get it working as well as my MC-1ply bot which got me into legend.
NotSureWhyThisWorks: still impressive
Wontonimo: i think the MCTS i wrote would probably rank about 600 gold, maybe
NotSureWhyThisWorks: I actually don't mind the idea of a MC-1ply bot I didn't know that existed
NotSureWhyThisWorks: how many rollouts can you get within a single turn?
Wontonimo: i don't know if it has an official name, but that's a name that LazyMammal came up with, so i'm using it
NotSureWhyThisWorks: fair enough
Wontonimo: around 3k*24/(24-day)
Lokmene: https://escape.codingame.com/game-session/e8H-807-aUJ-l1R
Wontonimo: I think that's pretty low and with some profiling I should get that up to 10k if I put some sweat into it
NotSureWhyThisWorks: are you using c++?
Wontonimo: yeah c++. btw, everyones lang is posted beside their bot.
Anti-Matter: Just wondering but for the games we have to write everything in the one file, yeah? we don't have access to multiple files
Numby: Yeah but it can get a little unwieldy to do this locally. I used a python script to stitch all my files together into one
Anti-Matter: ooo wait you can code on your PC and then upload to codingame?
Whiskee: yeah Codingame Sync is a nice extension you can use
Whiskee: every time I build on my IDE it lanches a script that merges all my .cs files into the one that's synced here
Anti-Matter: naisu
mnechromancer: Hey, just wanted to say to everybody in chat that you guys are awesome. This is my first bot-coding competition and I'm working a lot more on it than I thought I would, and my last version made it into silver! Couldn't have done it without chat's help, and I still have so much to learn :grin:
Whiskee: the only part that's a little weird is that I have to put all using statements inside the namespace, because they can't be at the top level on a merged file :)
Anti-Matter: hmm, can you test the code you wrote locally or do you have to upload it first
Wontonimo: hope to see you around after the competition mnechromancer , there's lots of other challenges here between competitions
Whiskee: always nice to see some enthusiasm!
Whiskee: Anti-Matter there are ways to do that, they were discussing it earlier
Kellthazar: Anti-Matter you can use the source code of the game that CG provided on github.
Zaphus: finally sorted out why my brutaltester was failing every time. turns out the game referee was not supplying any possible moves on the first turn, causing my code to return nothing. Doesnt happen in the real game but multiple referees I tried all had the same issue. Just forced a wait on turns without moves and it works... I think. All the starters fail for the same reason :-(
Pick8chu: Wait
Pick8chu: so brutaltester is not working because of the refrees?
Numby: I feel like as soon as I solve my issue with unbound memory growth and stop timing out I'm gonna jump to top gold: https://www.codingame.com/replay/556323585
Zaphus: for me, yes - I tried the referee alone and it still failed with starter bots
Pick8chu: well mine doesn't work neither
Pick8chu: It's so frustrating
Zaphus: put a check in your bot for a turn with no possible turns supplied, and return "WAIT" on that one
Zaphus: see if it helps
Pick8chu: wdym?
Pick8chu: for starters, i always return WAIT
Enknable: Numby are you use recursion?
Pick8chu: but non of them works
Numby: I think the issue is I either go over the memory limit or garbage collections trips me up. When I run my bot locally it never not provides a turn
Zaphus: do you return "WAIT" - or do you return command[0] ?
FrancoisB: Anyone using rust-sourcebundler?
My output file does not contain the dependencies
Numby: No recursion but my search tree might be too big
Enknable: are you timing your game?
Numby: I'm thinking of trying a pool of nodes and reusing
Pick8chu: string action = get<0>(possible_actions[0]) + " " + to_string(get<1>(possible_actions[0])) + " " + to_string(get<2>(possible_actions[0])); // default
Pick8chu: this is what i return
Numby: Yeah, you can see I print out the time and it's consistently at 80-81 ms (limited at 80)
Pick8chu: so it should be "wait" since the first one is always WAIT
Zaphus: thats the problem, there are no possible actions - just force "WAIT" and see what happens
Pick8chu: Oh
Pick8chu: there's no possible actuons??
Numby: so it doesnt seem there's turn variability causing it go over. Imo it's because I keep allocating objects and eventually GC cleans them up and I time out on that turn
Zaphus: again, for me on windows this is what is happeneing, looks like a referee bug but that change fixed it for me
Pick8chu: Oh you're right
Pick8chu: they don't have any possible actions
Kellthazar: How deep are your simulations? 2 ~ 3 days?
Zaphus: if it is happening on other turns then it makes the referee useless - thats the next thing I want to check
Enknable: after your first stdin, start = chrono:time, then after every iteration check that (chrono:time.now < 75)
Numby: I added depth limit of 4 simultaneous actions
Enknable: then at least you won't timeout
Enknable: or w/e # works
Kellthazar: Hmm
Pick8chu: so i should code without having possible actuons?
Numby: Yeah I'm already doing timing myself after first input (System.nanoTime() in java equivalent_
Enknable: ok ok
Enknable: :)
Zaphus: @pick8chu looks like it only happens once from a quick check
Pick8chu: only once
ClockSort: what is the memory limit for spring challenge?
Wontonimo: the same
Wontonimo: it's in the faq ... what's that link?
Wontonimo: https://www.codingame.com/faq
ClockSort: 768mb
ClockSort: well that's a problem, i'm using 2gb. :D
Pick8chu: now my replay board is not shown lol
Pick8chu: says demo.js file is not there
pmor: I wish CG would tell you why you timed out...the memory issue plagued me in previous challenges.
pmor: and I believe I was getting a time out from a stackoverflow
pmor: for this challeng
Wontonimo: what is taking up all that space ClockSort?
pmor: at least I'm able to tell when a sigv happens
ClockSort: @wontonimo my brain
Wontonimo: well, it's swollen cuz it needs to fit in 768mb
ClockSort: working on it
ClockSort: ok, submitting a new bot with a smaller brain. thanks for the help
Numby: ClockSort i'm rooting for you I'm struggling with the same exact issue lol. Will need a bit longer to fix though
ClockSort: i sat on Gold #1 but the bot is better. nobody knows my pain :joy:
Wontonimo: i've been there. not on a contest though
pmor: at least youre gold ...
Wontonimo: can anyone tell me the GoldBoss score ?
mybk: 48
mybk: 48.43
Wontonimo: you curb stomped it reCurse ... geez man, go easy.
struct: holy
k4ng0u: http://cgstats.magusgeek.com/app/spring-challenge-2021/reCurse no mercy xD
JohnCM: recurse to legend soon
JohnCM: clocksort, you just need to sit your bot there, eventually some pro programs will push the elo of the bot down and you will rise
mybk: to top 10 soon ;)
struct: seems like rank 1 bot to me
JohnCM: yea it's like owning everyone
Wontonimo: HAHA
Wontonimo: there is a certain satisfaction from those kinds of submits
struct: his bot plays different from top from what I can tell at first
KelvinAndHubbles: should I do anything if I'm 1st, but not beating boss?
KelvinAndHubbles: Ik Smetly just kind of waited for two hours and he was above
mybk: yeah, ameliorate your bot, play against yourself, if you win 8 from 10 battles => submit
Wontonimo: i waited about 18 hours and I got moved from 35th into next league. It's a gamble though. it could be that you get pushed the opposite way
struct: Protect your bots if you are in legend
struct: duck
JRFerguson: my bot is afraid
CouscousAgha: yo who spawned the legend boss in gold?
struct: lol
struct: 100% win rate on legend?
struct: any bets?
CouscousAgha: hard
struct: its 50/50
casmith789: definitely not
k4ng0u: one loss against the gold boss!
k4ng0u: it's not invincible anymore :D
struct: time do put it in the bin
struct: to*
CouscousAgha: top 24 already
struct: uff it didnt play vs me
struct: im safe
KelvinAndHubbles: problem with it beating my bot 80% of the time is I don't know if thats just countering my bot or actually good
CouscousAgha: bro
CouscousAgha: this bot is madness
CouscousAgha: lmao i imagine the look on top legends' faces
CouscousAgha: whent hey wake up to this
struct: Imagine going to sleep and wake up to this :'(
CouscousAgha: god of ubisoft
CouscousAgha: imagine if it's made by an ubisoft hardcore dev
CouscousAgha: reCurse go easy on people
Zandy156: wow
Counterbalance: a non committal submit, immediately to rank 1
JRFerguson: insane
mybk: what do you think recurse use like algorithm ?
pmor: skynet
JRFerguson: saelyos beat him
struct: I think cg can save the servers and end the contest
Zandy156: its like its day 1 of contest and one person actually tried.
VincentBab: insane xD
JohnCM: recurse has won plenty of previous competitions. this one should be easy for him
JohnCM: but the sun shifting is really difficult to code for
CouscousAgha: theyw ere like meh i will do this in one saturday afternoon and submit it
struct: japan is already reacting to recurse :(
Wontonimo: this has reminded me it's high time I code my NN harness for CG competitions
struct: This reminded me that is time for me to stop coding
mybk: isn't the limitation for NN in CG too high ? 100ms/100kb file size ?
struct: yes
struct: there are plenty of NNs though
mybk: so how we can do NN
struct: compress the weights
UndercoverToad: reCurse wtf
struct: toad woke up
dbdr: yeah, just saw :D
struct: Throw the towel
Marchete: +5 points euler
Counterbalance: 6 point lead
dbdr: gg reCurse
Marchete: http://cgstats.magusgeek.com/app/spring-challenge-2021/reCurse
pmor: you missed the carnage as his bot took the throne
Zenoscave: I broke my bot
struct: Everyone had alarm to when recurs e submited
struct: lol
UndercoverToad: let me guess: the almighty NN
struct: I would say yes
Wontonimo: and alphaZero enters the battlefield
Zenoscave: cool Good to know My eval is super fragile
UndercoverToad: so he has a winrate as high as struct when bronze opened?
struct: lower
struct: i had 100%
Zenoscave: slightly lower
dbdr: noob
dbdr: ;)
ZarthaxX: gg lol
Wontonimo: well, i don't think truskill really works for more than 8 points difference no matter how one-sided the battles are
Marchete: you lose points by winning
struct: yeah
Pick8chu: is the refree working now?
Pick8chu: I'm trying to code possible_actions but it's tricky
Zenoscave: I have the same losses as recurse this submite
Zenoscave: so far
Zenoscave: just less wins
struct: he lost 1 match in gold
Zenoscave: Jesus
struct: vs the boss
Zenoscave: Nevermind
struct: https://www.codingame.com/replay/556355349
UndercoverToad: that's why he already submitted today: was scared of the boss
Zenoscave: however fixing my coeffs jumped 50 places so far
Marchete: was scared of CG servers going slow
Zenoscave: Is it very time dependant his solution? or is it nearly instant
Marchete: I'll wait his PM :D
UndercoverToad: play him vs him to see how fast it goes
Marchete: I like to read PM a lot
Marchete: they are interesting
UndercoverToad: the only thing that can stop him now is CG seeing obfuscation in his NN encoding
Marchete: but are you sure is NN?
Zenoscave: They wouldn't do that would they?
Marchete: It's hard to do it here
UndercoverToad: of course i'm not sure
Marchete: too many different bits
Marchete: with one hot encoding
Zenoscave: I'm sure it's just weights and they've stood on the fact of weights are fine i thought
Marchete: 37*4 *2
Marchete: just trees
UndercoverToad: 37 for richness 37 for owner 37*4 for size
UndercoverToad: plus points and suns
kovi: that hurt...
kovi: but better now than tomorrow
Marchete: I'd say owner*2
struct: Just wait
Wontonimo: i'd instead do 9 for richness, 9 for owner, 9*4 for size and convolve it ...but i haven't tried NN on CG, so i'll let you know how it goes
struct: Ive been saving the bot that I made during testing for this
Marchete: convolve seems good here
Marchete: because of rotations and such
Marchete: and it needs less weight but more CPU time
Counterbalance: isn't nn nearly instantaneous once trained? it's just fixed number of multiplications and additions.. his bot seemed to squeeze every millisecond out of each turn..
Zenoscave: maybe training inline on top of?
Marchete: convolutional isn't
Marchete: convolutional is like taking a point and its neighbours
Marchete: and calculate
kovi: its still mts (duct?)
Wontonimo: *caugh* alphaZero *caugh*
Marchete: so guess how many calcs
kovi: (i guess)
kovi: mcts
Marchete: robosta_c also has very good A0 bots around there
Marchete: it shouldn't be that easy to pull it a working NN here (due to restrictions)
Marchete: CG*
dbdr: local training
Marchete: I mean size restrictions
Marchete: and CPU time
Counterbalance: it's probably just a search with 10 million sims :)
ZarthaxX: LOL
dbdr: training is CPU intensive
struct: heuristics maybe?
Wontonimo: if (losing) { win }
Wontonimo: i should have thought of that
mybk: System.makeMeWin()
mybk: No SAO Fun here ?
Wontonimo: well reCurse, if you are listening, Good on you! Really, cheers for showing what can be.
Smelty: when's the contest over
struct: now
Smelty: ah 1 day
Smelty: hi mr. str0ct
dbdr: lol struct, was gonna say the same :)
Smelty: lol
Smelty: hi mr mod
struct: hi Smelty
Smelty: 7 mods in a room what do they do https://snipboard.io/7zJ2WV.jpg
Smelty: hi
UndercoverToad: they intimidate
Kellthazar: trying to simplify my evaluation heuristc...
Kellthazar: my bot stop to plant seeds after a while. XD
Hoyer: I keep getting `An error occurred (#407): "You reached the limit of plays for a period of time."`
Hoyer: whats the limit
Wontonimo: how many did you do?
Hoyer: idk a few, I was adjusting parameters
struct: its very low atm
Wontonimo: (joke aside, i think you only need to wait 5 min)
struct: they lowered it
Wontonimo: (says in deeper voice) only 5 min
Wontonimo: oh, you mean numerically ... okay
UndercoverToad: "atm" these restrictions will remain the way they are
ALDERS0N: you guys are so annoying
struct: thanks
Zenoscave: :+1:
Kellthazar: o__o
JRFerguson: https://www.codingame.com/replay/556374400
dbdr: https://www.codingame.com/replay/556374058 recurse, fix it first match I tried, disappointed
Zenoscave: could be hiding against ide?
JRFerguson: dbdr, we both have 100% win rate against reCurse
struct: you wasted your luck on the ide
struct: rip
dbdr: dunno. I thought I just got lucky
struct: he has an if
dbdr: JRFerguson :+1: :D
struct: random chance to lose
dbdr: I don't think reCurse needs hiding
struct: nah
struct: nobody can adapt their bot to beat this
Zenoscave: not even dapo?
ZarthaxX: lol
ZarthaxX: clash of titans
struct: maybe pb4 is coming
ZarthaxX: would be amazing
Zenoscave: would be scary
Zenoscave: ftfy
Wontonimo: that would be fun to watch
JRFerguson: or Agade
Marchete: I'm not sure this is easy for a NN, also the possible moves are big
Marchete: 37 * grow sizes
Marchete: 37*20? seeds
Marchete: etc etc
ClockSort: should be easy to prune seeding, or is that not true for NN?
Wontonimo: i had a hard enough time figuring out how to queue the moves in MCTS Marchete, the though of how to do that with NN made me remember I want to go to sleep soon.
Zenoscave: depends on how you encode/decode actions
BrunoFelthes: anyone are processing moves during opponent turn?
ClockSort: you don't get cycles during opponent turn
Marchete: or maybe he just use it for value
Zenoscave: *shouldn't* get cycles
Marchete: in that case is just a number
struct: What do you mean BrunoFelthes?
Marchete: pondering
Marchete: and it doesn't work here
BrunoFelthes: while you are waiting for your next input, maybe your opponent is doing some move... you can use this time...
struct: no
Wontonimo: your instance is paused
struct: Its gets paused
Marchete: paused it gets
Zenoscave: They fixed that?
BrunoFelthes: how they do it?
Marchete: linux AAA knowledge
Zenoscave: lol
Counterbalance: cgroups
BrunoFelthes: yeah, it is not a 100% for sure... but i already notice that my process is slower during opponent turns...
Marchete: you'll get a10% or so
Marchete: it's the referee time I guess
Marchete: at some point you are "allowed", but not in enemy time
Marchete: it's some another time inbetween
BrunoFelthes: cool
ClockSort: 100ms is already a lot of time
Marchete: so tldr, just don't waste time in pondering
Marchete: isn't worth it
ClockSort: it's enough time for my MCTS to decide on its bad move. more time would just make it more confident in the bad move
BrunoFelthes: not for my Garbage colletor
Marchete: or trying to use more than 1 vCPU
Marchete: ahh
Marchete: GC
Marchete: then yes :D
ClockSort: GC :grimacing:
Marchete: but the best way to improve CG is never using it
Wontonimo: don't have garbage
Zenoscave: I should look into my GC in my bot
Zenoscave: Haven't done that yet
BrunoFelthes: yeah... but I changed now to MCTS, it is very hard to not use GC on it...
Wontonimo: what Marchete said
Marchete: class NNN{
Wontonimo: (oh sorry, bad timing)
Marchete: static NNN[];
Zenoscave: in C#?
Marchete: static List<free>
Marchete: static List<used>
Zenoscave: this works?
Marchete: NNN getNew()
Marchete: void setFree(NNN)
Marchete: setFreeAll()
Marchete: ofc, both Java and C#
Zenoscave: so write maloc free and clear
Marchete: if you want to keep "new"