jacek: happy Caturday
dbdr: happy Raturday
eulerscheZahl: oh, i got beaten up during the night :(
eulerscheZahl: and someone gave 1 star to jacek's paper soccer
dbdr: monitoring multi ratings eulerscheZahl?
dbdr: who's beating you up?
eulerscheZahl: from time to time. several games were downvoted
struct: yeah :(
eulerscheZahl: legend is beating me up
struct: I wonder if I even should finish yinsh tbh
struct: so many multies now
struct: even on pending list
struct: is like 4
eulerscheZahl: yeah, impressive
eulerscheZahl: moderating them is quite some effort
struct: especially during contest
eulerscheZahl: did anyone try codejam last night?
struct: No :(
eulerscheZahl: me neither, i decided to sleep
struct: has yurkov joined ooc long ago?
struct: only saw him now
eulerscheZahl: i think i saw him before
eulerscheZahl: https://github.com/dbdr/codingame-leaderboards/blob/master/challenge/ocean-of-code.tsv
struct: he is top 50 now
dbdr: that URL pings me :D
eulerscheZahl: guess how i feel with my github and herokuapp
dbdr: :nerd:
eulerscheZahl: i can even ping myself when i have multiple tabs open
dbdr: do they come up often?
eulerscheZahl: yes, when i share them :D
dbdr: :D
struct: dbdr you store leaderboards everyday?
dbdr: yes
struct: Now just need a website to choose dates and contest
dbdr: sure, this is the raw data, not user friendly
eulerscheZahl: there was such a website from adrien
dbdr: though github does a decent job of displaying tsv as a table
eulerscheZahl: but he stopped indexing
struct: Good job displaying yes
struct: but navigating is a bit meh
dbdr: I think it's the right architecture, separate the data from the visualization
dbdr: sure, it's not built for that
eulerscheZahl: https://www.adrienvannson.fr/cg-ranking/
dbdr: before the ranking v2?
eulerscheZahl: yeah, old one
eulerscheZahl: but it tracked your contest rating every 2h
dbdr: anyone could build a visualization on top of the github data, any good frontender?
eulerscheZahl: helped finding the "good one" again :D
dbdr: with agentid?
eulerscheZahl: your github data is 16 day old
dbdr: not anymore
eulerscheZahl: yes, agentId in the tooltips
dbdr: only the push is not automated yet
dbdr: should do it
eulerscheZahl: afk breakfast
struct: you want to just display leaderboard and navigation
struct: or something else?
dbdr: struct: a lot of things could be done with this data I think
struct: yeah
dbdr: just display the lb is fine but already on CG
dbdr: but graphing over time
struct: Can also check last agent id
dbdr: dayly/weekly/month changes, ...
struct: and see what changed
dbdr: right
kovi: uh, variance is still high
Marcose: anyone?
dbdr: kovi same or a bit less?
kovi: same code 7th and now 1st (3points diff)
dbdr: at least there's more spread in points
struct: I havent used javascript in a while
struct: but i can give it a try
dbdr: if you feel like it, very cool struct!
eulerscheZahl: but don't forget about the contest
eulerscheZahl: i don't want you to delete your account again
struct: Im not doing the contest :/
eulerscheZahl: you said legend or delete
struct: :sweat:
kovi: legend and delete username is already taken. "not doing the contest" is accessible
dbdr: https://www.codingame.com/replay/447659206
dbdr: shows 30.73 for me. wonder where that number comes from
eulerscheZahl: gold
dbdr: looks like it
dbdr: but it gold it was updating
eulerscheZahl: that's your old bot
dbdr: they keep one score per league?
dbdr: forever?
eulerscheZahl: you need a replay with your most recent agent id
eulerscheZahl: the score will remain the same, it doesn't show a rank for old replays
dbdr: they keep the score of all agents?
eulerscheZahl: yes
dbdr: whatever for?
eulerscheZahl: bots are bound to an agent id, not to a user
dbdr: but past bots will never be shown
dbdr: keeping the source I understand
eulerscheZahl: fun fact: you can manually set the agent id of your opponent. and then play against kovis tower defense bot in OOC and make him believe that he's crashing :D
dbdr: :D
dbdr: build a universal bot
dbdr: detect which game it is playing based on the input
kovi: :)
dbdr: fit it in 100K too ...
eulerscheZahl: oh, kovi saw my idea now :(
dbdr: the meta multi
eulerscheZahl: well, still some other users to fool
dbdr: we should definitely build it
kovi: the ultimate leaderboard
dbdr: yes
dbdr: oh, eulerscheZahl has a trick, need to reverse it :)
eulerscheZahl: what trick?
dbdr: "kovi saw my idea"
dbdr: I thought that meant you implemented a cheese
eulerscheZahl: no, the bot from another game
dbdr: ah right
dbdr: hum, we will need to build a meta referee too
dbdr: any size limit?
eulerscheZahl: 10MB
eulerscheZahl: including graphics
dbdr: just copy all the referee in one and rand() to pick the game first :)
eulerscheZahl: and downscale the graphics
dbdr: yeah, worth it
dbdr: the variance on that metamulti :D
dbdr: yeah, I got lucky, playing PCR in this match
kovi: there are generic solvers...
eg. one smitsimax with game specific eval
eulerscheZahl: will we tell players which game they play? or do they have to reverse the input format?
dbdr: right, a good motivation to factor some generic code
kovi: yeah reverse
dbdr: reverse
dbdr: more fun
dbdr: some might be amgiguous
dbdr: need to play both for some time
dbdr: inspired by OOC
dbdr: uncertainty to the next level
eulerscheZahl: isn't aCat working on that kind of generic bots?
dbdr: ah, can we send the referee source code as the first input? ;)
kovi: yeah, i remember he talked about that at locam
dbdr: yavalath is part of a generic game framework I think
dbdr: where the rules are data
jacek: ludi?
dbdr: yeah that one
jacek: youre talking about implementing general game playing?
dbdr: not general, just all the CG multis
eulerscheZahl: with regular updates to add new games
dbdr: yes
dbdr: keep up or you will timeout more
navin89098909890: Hi
dbdr: a wait bot in everything would probably do quite well :)
eulerscheZahl: just don't crash
dbdr: xactly
dbdr: https://github.com/dbdr/codingame-leaderboards/blob/master/challenge/ocean-of-code.tsv
why did I called the directory challenge? should be contest
dbdr: *call
AntiSquid: hi navin89098909890 join the contest
Neozero: I fcking did it Hahahaha
Neozero: :D
Nerchio: guys before i fell asleep yesterday I had a very important question pop in my mind
Neozero: I beat the TAN Network finally
Nerchio: in terms of strategy what is better - to start in the middle or on the outside in the ooc
eulerscheZahl: in the largest lake
Nerchio: that narrows it down for sure :D
AntiSquid: is there any reason pragmas could cause bugs? (C++) deactivating them makes my bot run just fine :/
AntiSquid: #pragma GCC optimize "O3,mavx2,omit-frame-pointer,inline,unroll-loops"
eulerscheZahl: the optim level might reveal existing bugs like not properly initialized variables
eulerscheZahl: memory layout might change too. so an index out of range for an array can have different behaviour
AntiSquid: are some bugs probably caused by lack of memory, doubt i reached the limit, but who knows
AntiSquid: any way to check how much i use from the total ?
eulerscheZahl: seems unlikely to me
AntiSquid: ok will check my variables then, thanks
eulerscheZahl: google, sorry i don't know it either
kovi: what euler said. +add bunch of logs and run the same code with/wo. check difference
pb4: AntiSquid : like eulerscheZahl said, O3 will not initialize variables to 0 by default
pb4: O0 will
pb4: hence some bugs appear with O3 only
AntiSquid: i use array_something = {0} to initialize
wlesavo: im trying to use java cg benchmark, the login is successful and after trying to start a game im getting this error [ wlesavo ][ SEED 0 ] ERROR internal error
wlesavo: can you advice what is wrong?
AntiSquid: which seed are you using
dbdr: wlesavo can you get the stack trace of the error? might put in the right direction
wlesavo: seed was random, but with list getting the same error
wlesavo: stack trace?
dbdr: "internal error" is a terrible message. should give some info about what went wrong and where
wlesavo: oh, maybe its because i use precompiled jar, ill try to rebuild, and then probably will comeback :slight_smile:
dbdr: does not sound likely to me
dbdr: maybe there's a configuration about how verbose the logs are?
dbdr: sorry I don't use cgbenchmark so I don't kno the specifics
wlesavo: hm
AntiSquid: did you use correct agentID ?
AntiSquid: i could send you my configuration file, if i find it, not sure where i placed it @_@
wlesavo: ok, it was due to space in Python 3, nwm :slight_smile:
Nagatwin: Antisquit I had an issue when using a pragma on my dfs, still haven't investigated why
Nagatwin: AntiSquid *
AntiSquid: well i found out the bug is related to eliminating possible enemy positions after torpedo shots
AntiSquid: i don't clean up the positions properly somewhere ... so it tries to access indexes it shouldn't
Nagatwin: uh rip
Nagatwin: xD
Nagatwin: https://www.codingame.com/share-replay/447931600
Nagatwin: "steering" stuff isn't for me i guess
YannT: ahah you made the default stub AI so much worse :D
dbdr: playing the same person several times in the first ten
dbdr: CG matchup algorithm could be improved...
AntiSquid: finally \o/
AntiSquid: too many variables Q_Q
dbdr: what did you achieve AntiSquid?
AntiSquid: i can correctly eliminate positions based on damage/explosions
dbdr: nice
eulerscheZahl: does that include path backtracking for mine placements on TRIGGER?
dbdr: "based on damage/explosions"
AntiSquid: no, just torpedo, but the new bot will have the trigger bit as well
dbdr: not part of the requirements ;)
eulerscheZahl: TRIGGER is an explosion too
AntiSquid: with the current arena bot i cut a lot of corners and do some very superficial checks
dbdr: true
eulerscheZahl: Illedan lat me win C# for once :/
AntiSquid: but will add mines too
Illedan: No.
AntiSquid: just get 1st place in contest euler, problem solved
Nagatwin: ah
Nagatwin: thats the 3v
eulerscheZahl: exactly :D
Nagatwin: rofl
eulerscheZahl: now you have the secret algo for gold
Nagatwin: yay
Nagatwin: finally unlicked the full statement
eulerscheZahl: you can still lick the max thrust, will be 200 in legend
Nagatwin: unlocked
dbdr: be careful what you lick these days
Nagatwin: tuning the thrust is enough gor legend ?
eulerscheZahl: no
Nagatwin: oh
eulerscheZahl: i wrote a sim in bronze, had to rewrite for changed gold input
Nagatwin: ye
eulerscheZahl: completely killed my motivation for that game
Nagatwin: thats why I wanted to get that strat before writing anything
JasonTruter: How do I use my sonar?
Nagatwin: SONAR 5
Nagatwin: replace 5 by the sector you want
AntiSquid: Nagatwin you could tweak speed / angle turning point variables for legend
Nagatwin: angle turning points ? :o
eulerscheZahl: did you give up on OOC Nagatwin?
AntiSquid: and have a blocker pod and one runner pod
JasonTruter: What is a sector?
AntiSquid: the big 5x5 squares
eulerscheZahl: watch a replay JasonTruter
do you see these numbers in the squares?
Nagatwin: eulerscheZahl can't get that code any higher, I have to write a full sim and i'm not godd enough at these thing
Nagatwin: good
AntiSquid: you can get legend without sim
eulerscheZahl: he did
Nagatwin: im legend
AntiSquid: ah nvm
dbdr: in OOC
Nagatwin: I really have no idea on how to write a good simu, especially with several actions in the same turn
Nagatwin: So far I have a small path simu and killmove simu
eulerscheZahl: and the incomplete information
Nagatwin: thats not my issue eulerscheZahl
AntiSquid: you got legend in OOC without sim?
Nagatwin: Yeah
AntiSquid: wow . what did you use lol
Nagatwin: good pathfinding
dbdr: LeChuck is not simming either I think
AntiSquid: that's it ?
Nagatwin: killing simu
eulerscheZahl: https://www.codingame.com/replay/447873049
val growth has a kill avoidance at frame 59
Nagatwin: like if hp < 4, I kill
JasonTruter: so an example of a sonar would be the following "MOVE N | SONAR 5"?
Nagatwin: and perfect detection AntiSquid
AntiSquid: jeez ... and i am now writing a sim to get to legend, maybe i should tweak my if-else instead?
eulerscheZahl: yeah but you can still charge in your MOVE JasonTruter
JasonTruter: charge?
eulerscheZahl: MOVE N TORPEDO | SONAR 5
JasonTruter: I see, for some reason my sonar result comes back as NA and cooldown still -1
JasonTruter: and breaks
eulerscheZahl: -1 means that it's not unlocked in your league
eulerscheZahl: are you in wood2?
JasonTruter: yes
Nagatwin: AntiSquid Idk haha I still have a lot involved on my pathfinding that makes it a simu partially. But I only assume I do 1 move per turn (thats why I always silence 1 distance)
JasonTruter: I've got basic movement down, where it decides on moves based off islands and visited right now
Nagatwin: AntiSquid and torpedo/mine are not any part of the pathfinding
eulerscheZahl: not always Nagatwin https://imgur.com/a/zHjgB2X
Nagatwin: eulerscheZahl that's the killing move
Nagatwin: but yeah I agree, I sometimes use silence 4 to kill the opponent
eulerscheZahl: you always go by 4 on killing?
eulerscheZahl: no 2 or 3?
Nagatwin: and the next line is return 0;
Nagatwin: eulerscheZahl 0 to 4
Nagatwin: the simu is complete for the killing move
eulerscheZahl: maybe i need more replays of your bot then
eulerscheZahl: only analyzed 22
Nagatwin: trigger and torpedo included
dbdr: Nagatwin return 0 that's confident, nice :)
Nagatwin: dbdr haha yes
dbdr: I'm done here :D
Nagatwin: when I do that I do not compute any update on my paths
dbdr: saves CPU cycles
Nagatwin: So if I do not kill with this part, my bot would be screwed up
Nagatwin: so yeah just return 0 for debugging
dbdr: that's good. better to crash and investigate than hide the issue
Nagatwin: http://chat.codingame.com/pastebin/2cc8888f-65c3-46dc-9a65-be87d9a4fe99
Nagatwin: inside 8 nested for loops :p
dbdr: :upside_down:
eulerscheZahl: my software engineering prof would hate you
dbdr: that's why it does not scale to a sim
Nagatwin: I have no sanity checks for my path arrays :p
dbdr: you just sim a special case
Nagatwin: Yes
Nagatwin: But sims are slows as fuck
dbdr: does not have to be
Nagatwin: I couldn't manage to simulate several turns with this game
dbdr: essentially the same work as your 8 loops
Nagatwin: With a proper one, not the bruteforcing i'm doing fot the killing move
dbdr: right, if you go deeper there is more work, exponentially
Nagatwin: essentially sampling some moves for each turn and applying them
Nagatwin: But the update of my possible paths killed the simu count
Nagatwin: Otherwise I could make a simu without considering the possible paths at the end but that seem wrong for this game
Nagatwin: and it basically is the same as my current pathfinding part
JasonTruter: Wouldn't you need sonar to detect within the path?
JasonTruter: So you can't really use BFS or DFS as it would be required that you visit the node?
JasonTruter: and then go back to the root node, unless you surface everytime you need to go back
Nagatwin: I did that on my first version
JasonTruter: Not very efficient?
Nagatwin: randomized dfs, take the deepest branch as base path and then add some path length using heuristics on 2x2 blocs
Nagatwin: JasonTruter that's a way to fill an area
JasonTruter: Yeah, I'm considering finding the quickest path between each sector
Nagatwin: but let's say that we have to consider more short-term pathing strategy for legend that involve aescaping mines, ...
JasonTruter: to use my sonar on each sector
Nagatwin: I'm not sure to understand what you mean
Nagatwin: sonar is supposed to tell you if the opponent is in a given sector
JasonTruter: I'm on wood 2, what would the minimum I need to do right now be?
Nagatwin: start on a place with water
JasonTruter: Yeah I've done that
Nagatwin: and move around without going into islands/visited cells
Nagatwin: when you are out of visited cells, you can surface
Nagatwin: generating a long path is up to you
JasonTruter: Yeah so I've done that too except for surfacing when no possible moves left
JasonTruter: So how would I get to the next league?
Nagatwin: If you have a path long enough and ensure you are not producting invalid moves, you should be able to get to wood 1
JasonTruter: I'm in wood 2
Nagatwin: I know
Nagatwin: when you reach rank 1 in wood 2 you get promoted to wood 1
Nagatwin: Otherwise you can try using torpedos randomly, this may touch your opponent
SabertheLost: my algo sucks :) it only fins the furthest point, but does not take into consideration what it might block ;)
SabertheLost: finds*
JasonTruter: Whats a MSG?
Nagatwin: MSG is something you display each turn
SabertheLost: as in message I think
Nagatwin: https://www.codingame.com/share-replay/447962016
Scarfield: for debugging or silly comments
Nagatwin: here below my nickname you can see the message I display each turn
JasonTruter: Oh okay cool
MadJetski_58cc: nice
SabertheLost: Ok, so I changed from BFS to DFS, then limited to 50 steps. I went from rank 240 to rank 16 in wood 2 with only moving...
SabertheLost: ok rank 3 now.
JasonTruter: where do you see rank
SabertheLost: At the top, where you see the league your in and time left.
SabertheLost: ok, i´m actully in wood 1 not 2.
SabertheLost: And there i will be promoted by only moving?
jrke: guys i am in eight standard i dint know trignometry can you help me for coders strike back gold league that how to find angle in it
jrke: please
jrke: :pray:
wlesavo: missed conversation, but dont use simus as well
blasterpoard: jrke https://www.khanacademy.org/math/trigonometry should be a good source
jrke: can you tell me which function to use
blasterpoard: nope
jrke: like acos
jrke: asin
tranculent: are we supposed to handle initial "Invalid coordinates" somehow?
jrke: or atan2
blasterpoard: jrke it's explained in there
jrke: ok
blasterpoard: which function is used for what
tranculent: to find an angle, I use Acos @jrke
jrke: ok
Scarfield: take note that the angle input of your pods are measured differently than the standard for sin/cos calculations. and the coordinate system has y axis inverted as well
SabertheLost: rank 68 in bronze, maybe time to start using mines or torpedoes
SabertheLost: 43 now
jrke: http://chat.codingame.com/pastebin/daf2cd30-b498-4a46-bbcc-deedb0b2de8d
jrke: can you check it is ok
blasterpoard: jrke if you want people to check something, it should be clear what were you trying to achieve
blasterpoard: I have no idea what do you want to have in those 1letter variables
jrke: can you share your code in private
elsinnombre2d2: lmao
blasterpoard: I can't
jrke: just for angle
jrke: only
pb4: You should first answer the questions you were asked jrke
jrke: what?
pb4: read
jrke: which question
struct: jrke preety much all you need to write a sim for CSB http://files.magusgeek.com/csb/csb_en.html
MadKnight: ezpz
ThienPhu: give
MadKnight: what
AntiSquid: kick
kovi: top4 is balanced
Illedan: Let's try to fix that :P
Famout: Guh, been far too long since I did stuff with actual pathfinding. Once this is ready to run though it's gonna be fun seeing how much I have mucked it up.
Famout: And forgot 2D arrays don't like to be passed without their size fixed in stone.. Anyone know if the Ocean of Code challenge ever changes the map size?
Counterbalance: it's constant
Famout: That's great, used to lots of things changing in a challenge and was keeping size flexible, that makes it a lot easier.
Counterbalance: whoops https://www.codingame.com/share-replay/448038500
eulerscheZahl: scared of mines? my bot does the same nonsense :(
Famout: The fact the mine blew up and he still didn't leave I think was the larger issue, big fish small pond.
Counterbalance: still too many possible mine locations.. Pre detonation:
Counterbalance: opp mine blast area: http://chat.codingame.com/pastebin/eee9b491-353f-4574-94bd-0377def767fb
Counterbalance: post:
Counterbalance: opp mine blast area: http://chat.codingame.com/pastebin/b21a5c48-d0b9-46fd-907a-065866944cfb
Famout: Ah, fair enough, fear is the sub killer after all.
Nerchio: time to add mines to my bot :D
Famout: Moment of truth, time to see if I correctly coded full map pathfinding correctly the first time.
Nerchio: probably not :D
Famout: The fact it just ran into an Island tells me no....
Famout: Huh, bugs in brand new untested code. strange
Famout: On the plus side, it did move at least.
Nerchio: is there a better way of checking your 8 neighbouring squares than 8 ifs?
NormantasStankevicius: wdym?
NormantasStankevicius: 8?
Nerchio: when you are at 5,5 you have 8 squares around you
Nerchio: or wherever else
Valdemar: http://chat.codingame.com/pastebin/8e32e889-f42e-4296-9a38-866005493d80
NormantasStankevicius: well, if you need to check
NormantasStankevicius: if something is at range
NormantasStankevicius: you can do some range calculations
NormantasStankevicius: let's say
NormantasStankevicius: if I target cordinate is in those squares around me
Nerchio: valdemar its a vector?
Nerchio: i will be checking how many island squares there are around x,y
Nerchio: i want my mines to be placed with maximum area open
Nerchio: its simple
Valdemar: vec2i is x,y structure
Nerchio: but i am asking if there is anything better or nicer looking than 8 ifs
Nerchio: i am writing in java
Valdemar: idea is to use array of shifts
JasonTruter: Whats the benefit of using different traversals in DFS?
Famout: Hmmm, the response time of 50ms might be an issue...
Famout: Great, I kinda need to check how much can be done in that time, to see just how badly I am going over it.
Counterbalance: what language, Famout?
Famout: c++
Counterbalance: do you have the pragma?
Famout: I have never done anything with that before.
Counterbalance: put this at the first line of your source file:
- pragma GCC optimize("Ofast","unroll-loops","omit-frame-pointer","inline")
Counterbalance: that'll turn on compiler optimisations so it will run 10x faster or so.
Famout: Ah, that's why! Used to Visual studio where I handle all that stuff via the GUI
Famout: Many thanks
D4yBr3ak: ofast = o3 ?
Counterbalance: O2
Counterbalance: more or less
D4yBr3ak: does it work for C too ?
Counterbalance: It should, yes
D4yBr3ak: thank you
Famout: OK, much improved, got to turn 115, before I tried to doubleback for some reason.
Famout: Admittedly did have to also put a limiter on the pathfinding, but I think part of that is some nodes doubling up over themselves.
Famout: Huh, it doesn't seem to be respecting the closed node list, that would screw things up alright.
Counterbalance: What are you using the DFS for?
JasonTruter: Pathfinding
Nerchio: i use bfs for targeting
kovi: nice illedan. not top4, but definitely closer
AntiSquid: how much rock-paper-scissors is there in legend kovi?
Illedan: Thx, let's see how we can fix that ;)
kovi: antisquid, there is definitely some
kovi: but most of the time i cant tell the reason
NormantasStankevicius: I was around 116th in wood1, left wood and played the same code, and now in 5th/300~ in bronze
YannT: I can't figure out why I consistently have bad winrates against some bots and good ones against others
YannT: there's seemingly no strategy difference or whatever
kovi: yeah. even though i cant tell what is under the hood
revo1: it's works but what the hill is this game
Famout: Something is off with my code still.... but not crashing, and just won twice agaisnt the test bot... and I am way overdue for bed.
Famout: Screw it SUBMIT
Famout: Zero surprise, so far beating folks who run into islands/off the board....
Ran_42: can someone help me with python binascii
MadKnight: what's the problem ?
Ran_42: i dont understand how to convert a ascii to binary
AntiSquid: google it
Ran_42: it converts to something weird
MadKnight: ord('x') returns u the code of a char
AntiSquid: that's what i do
Ran_42: i did google it
MadKnight: then subtract ord('a') from it
MadKnight: and u will get the alphabetical index of x
AntiSquid: ah wait you mean the chuck norris puzzle?
Ran_42: ye
**Famout thinks making a passive aggressive sub that takes 600+ turns to win might be making the test battles take some time.
MadKnight: then convert to binary
MadKnight: index
Ran_42: it convert to this b'\x00\x00
Ran_42: instead of 0s and 1s
MadKnight: \x00 == 0
MadKnight: \x01 == 1
AntiSquid: build an array of 7 chars or something and then convert that
MadKnight: just replace this stuff
Ran_42: but everything is 0 then its worng
MadKnight: but it feels like u did something wrong
MadKnight: show your code
Ran_42: ye probably
JasonTruter: How do I progress to wood 1? I've got movement going where it resets surface when theres no other moves.
Ran_42: message = input()
message = b.a2b_uu(message)
Ran_42: this too import binascii as b
AntiSquid: JasonTruter you need to track opponent and shoot them
Famout: JasonTruter or have REALLY good movement, a mix of both would be best.
MadKnight: no u used the wrong thing
MadKnight: u gotta convert ord(message[i]) to binary
AntiSquid: also use bfs to save some space when going around the map so you surface less and then you definitely get wood 2 JasonTruter
JasonTruter: I'm on wood 2 right now
AntiSquid: i mean wood 1
JasonTruter: cool
JasonTruter: Going to start by firing aimlessly
AntiSquid: no
Nerchio: everybody has different tips haha
AntiSquid: that's a bad idea
Famout: I did a bit of overkill and set up pathfinding to dig out a path that uses most of the screen before surfacing, no code for firing torps
JasonTruter: Is it
Nerchio: i did random shooting and random movement until bronze i think
AntiSquid: save shots for when the opponent location is known
Ran_42: thanks mad knight i got one more question
Famout: With %50 of the battles done, it is ranking me above the king bot.
Ran_42: why does it says 0b100... instead of 100...?
AntiSquid: Nerchio the must have overnerfed the bosses then or you submitted when the boss was broken first hour or so ?
JasonTruter: So for pathfinding maybe use dfs so that Its faster to get to my opponent?
AntiSquid: whatever you like
JasonTruter: As BFS can take quite some time, I'll use dfs once I know location
Nerchio: no i mean random shooting and when you hit something then it narrows down their position :D
Counterbalance: bfs can be much faster than dfs
AntiSquid: that's a different story Nerchio
JasonTruter: bfs would search through layers, that would take longer?
JasonTruter: I guess I could slowly move toward my opponent with bfs
Counterbalance: you only visit each node once in bfs
JasonTruter: This game really got me working on algorithms lol, I normally work with high level code
AntiSquid: have a function with bfs and one with dfs and switch your bot to use either as you see necessary, you will then find out what's best
JasonTruter: for building webapps etc
Famout: Oh wow, only now I discovered I screwed up using rand and picking my spawn point.
Famout: Also, woo! Wood1
AntiSquid: which languages JasonTruter?
JasonTruter: Writing in Java right now, but I write in python and javascript at work
JasonTruter: Do alot of AWS stuff
AntiSquid: nice
AntiSquid: why don't you get an avatar?
JasonTruter: I should hey
JasonTruter: This game has got me hooked right now
AntiSquid: the ranks on the leaderboard are like ocean waves, back and forth :D
AntiSquid: going *
Nerchio: for me its more like a sinking ship but okay
NormantasStankevicius: okay, my wood1 code put my in silver
JasonTruter: Don't know why my torpedo isn't firing
AntiSquid: check the console log in bottom left JasonTruter
Nerchio: this gold bot with MSG FATALITY is so annoying :D
JasonTruter: Torpedo is working
MadKnight: hey guys
MadKnight: are there any websites for artists where they can find a contract job ?
eulerscheZahl: i'm not an artist, just remembering former CG artists showcasing their skills here: https://www.artstation.com/
Famout: OK then, added in extremely limited use of silence, lets see how well I do in wood 2
Famout: wood 1*
Nerchio: damn this gold bot is really good
eulerscheZahl: reaching legend is never easy
eulerscheZahl: except for Tron
Famout: Yep, bedtime now, but managed to reach rank 50 on wood 2 with just a pathfinding bot, good seafairing!
Famout: wood1 dangit!
Nerchio: i guess i will submit and see how i do vs others i dont win too much vs this bot
Famout: Ah, wow, 50 outta 596, feeds good
JasonTruter: How would you know if opponent surfaced?
Counterbalance: it's given in the input
JasonTruter: sonarresult?
Counterbalance: the one after that
Counterbalance: line 3, opponentOrders
JasonTruter: Ah ok, thanks!
kovi: wow, almost 30, got pushed up
AntiSquid: with great power come great bugs!
JasonTruter: so true
JasonTruter: Fixed up all my bugs now
JasonTruter: Won with just movement :)
JasonTruter: But only ranked 300 lol
JasonTruter: need to improve
AntiSquid: anyone has a seed with almost no islands? please
Quidome: 0?
AntiSquid: 0 or close to 0, either
blasterpoard: seed=-430391657 AntiSquid
AntiSquid: thx
Nerchio: haha
Nerchio: opponent bot broke when i hit hit with torpedo and he used surface 5 times
Nerchio: so funny
JasonTruter: how would I use bfs without having to surface to get back to the root node
Swagboy: Why do you want to get back ?
Swagboy: If so, that's np hard i'd say
AndrejVelichkovski: @JasonTruter can you write the whole thing you are trying to do with bfs?
jrke: any suggetion coders strike gold league how to find angle
JasonTruter: trig
Uljahn: atan2
jrke: can u tell me algo
jrke: because i am in 8th standard so i didn't learn trig
jrke: plz
Uljahn: jeez, just go play fortnite
jrke: plz tell me
Astrobytes: plz google it
jrke: i can't understand it
Astrobytes: Fix it
jrke: please just tell me how to use and where to use
Astrobytes: No. You should try to learn.
AndrejVelichkovski: the thing is, if they tell you, you won't learn anything
AntiSquid: that's more than a line in chat worth of learning material
jrke: ok
jrke: you are right
jrke: i will learn
AntiSquid: no, he will learn, but the chat is too small for that kind of classroom
jrke: thanks
AntiSquid: i would suggest playing a grid based game instead, no trigonometry
jrke: ok
Nerchio: I just realised that triggering mines gives out a lot of information as well
Nerchio: if your enemies can track it well
Counterbalance: yup
Nerchio: they can remove all the possible positions that weren't there at that time to place the mine at X Y
AntiSquid: ya but if you spin in circles and lay a mine in a similar rea?
AntiSquid: area *
Nerchio: you can remove some not all :P
JasonTruter: I have to rewrite my movement :(
Quidome: Zenoscave, are you online?
Counterbalance: There's a little icon,
------------v here that tells you that
Quidome: yes, but it doesn't match the green dot in the leaderboard
Quidome: so i thought i give it a try :)
Counterbalance: Sure :)
Quidome: So you can be online and not participating. Meaning you closed your chatr window?
Counterbalance: yes
Quidome: aha, well, then he won't see it :p
Counterbalance: appending something like ?disableChat to the url does that
Quidome: is there another way to reach out?
Counterbalance: hmm.. I don't think so. You can send a private message, I think that also sends a notification, but you'd have to be able to click on their username in chat for that, or already have a private chat going
Counterbalance: ahh, /msg
Counterbalance: see /help
Quidome: where is the help
Counterbalance: just type /help here
Quidome: Nice
Quidome: hm, doesn't work either, you have to be on the chat...
Quidome: well, a pitty
Counterbalance: you can follow him
Quidome: I am going to try that
Counterbalance: then on your profile page, there's a friends tab, and there's a private message button there. That'll probably work
Quidome: After he accepts my follow request, we will see
Counterbalance: AFAIK acceptance is not needed
eulerscheZahl: to send someone a private message, that person has to follow you, not the other way
eulerscheZahl: i recommend sending a private message on the forum
eulerscheZahl: triggers an email notification
Quidome: :thumbsup:
Quidome: He euler, you already in! Congratulations!
eulerscheZahl: for a day already :P
eulerscheZahl: thanks
Quidome: Yeah watched you junoing within the hour or so, nice
Quidome: jumping
eulerscheZahl: deactivated my mine avoidance to get promoted
eulerscheZahl: then activated it again :D
JasonTruter: Thinking of storing a node that has 4 different neightbours
JasonTruter: What do you guys think?
JasonTruter: Would it make it eaier for me to traverse using bfs?
eulerscheZahl: somehow avoiding mines performs better in legend but did worse against the boss
eulerscheZahl: JasonTruter that's what I do
Quidome: interesting.
eulerscheZahl: http://chat.codingame.com/pastebin/e05707fd-9a50-49f0-94e9-18079bf7cf19
Quidome: Now i lost my debugging possibility and am hunting Zenoscave
eulerscheZahl: i'm sure others print their stats as well
Quidome: OK
Quidome: Does anyone else print their remaining locations count
Quidome: ???
Counterbalance: my bot prints them too - similar rank as zenoscave
eulerscheZahl: Risen too
eulerscheZahl: don't ask me how accurate these numbers are
Quidome: Great, going to compare them with mine, thnxs
AntiSquid: i print an useful assessment of the enemy bot
Quidome: great thnx
eulerscheZahl: https://www.codingame.com/share-replay/448160974
i officially hate this game now
eulerscheZahl: so much rock paper scissors
Nerchio: why?
Quidome: Muda
eulerscheZahl: (or maybe my bot is just bad)
Nerchio: your bot moved into enemy territory
eulerscheZahl: i'm allowed to move whereever i want
Nerchio: :D
AntiSquid: my bot is full of bugs
Nerchio: you are allowed
Nerchio: but don't be surprised when you get shot
Counterbalance: too bad we dont get opp's MSG or the bots could communicate
dbdr: he didn't get shot by mines
dbdr: so territory is not relevant
AntiSquid: not sure if i told you but it's these bugs that give it +50 ranks
my new rewritten bot will hopefully get legend though, and no bugs ihope
eulerscheZahl: just me fighting an invisible opponent while my location is clear
AntiSquid: ya in that replay you give too much info away
AntiSquid: also my output is helpful
Counterbalance: you ignore silence?
eulerscheZahl: https://www.codingame.com/replay/448165415 maybe i should use it more often?
Quidome: @counterbalance, are you pruning remaining enemy positions when to high?
Counterbalance: yes, but only > 10k
Counterbalance: I have no idea how to prune effectively :(
Nerchio: antisquid silence spammer :O
Zenoscave: Quidome how did you get my email?
Counterbalance: it works! :)
eulerscheZahl: i told him how Zenoscave
darkhorse64: nagatwin removes duplicates but I do not know ho
eulerscheZahl: from your github account
darkhorse64: *how*
Zenoscave: Ah smart
Zenoscave: good thinking
eulerscheZahl: and also a lie
eulerscheZahl: private message on cg forum
Zenoscave: yeah i didn't get any notification from that hahah
Astrobytes: Eulersche Troll
wlesavo: not a lot of promotions today huh
eulerscheZahl: 0 to be precise
wlesavo: i expected legend to grow faster
eulerscheZahl: all your fault!
dbdr: eulerscheZahl and Illedan, please keep it up!
https://i.snipboard.io/WZS6c9.jpg
eulerscheZahl: but still a week to go and we don't want a huge legend
wlesavo: lul
YannT: well yesterday people killed the boss so it lowered the bar temporarily, I think he's quite high now
eulerscheZahl: Illedan just submitted too
YannT: until someone can hurt it proper it's going to be hard to get promoted
eulerscheZahl: yeah, we destroyed the boss yesterday
eulerscheZahl: but then it recovered
dbdr: :innocent:
wlesavo: got raided
dbdr: great team work
eulerscheZahl: you only won 4-2 dbdr
dbdr: the straw that broke the camel's back
eulerscheZahl: and then this happened: http://cgstats.magusgeek.com/app/ocean-of-code/nagrarok
ZarthaxX: boss currently is 31.28 and below 29.97
ZarthaxX: it's a bit high
Zenoscave: https://www.codingame.com/replay/448172409
Zenoscave: I still timeout
eulerscheZahl: directly after SILENCE, suspicious
eulerscheZahl: your tree explodes
Zenoscave: Yeah
Zenoscave: I don't have good tree pruning
Zenoscave: I consider every silence option
Zenoscave: How often is SILENCE N 0 used?
eulerscheZahl: often
Zenoscave: Might have to rework my bot to detect it
ZarthaxX: how do u detectthat
Zenoscave: is that a worthwhile task?
eulerscheZahl: top4: https://imgur.com/a/YfmRduu
Zenoscave: back pruning
Zenoscave: Would just ignoring silence be good? and resetting if you loose them?
eulerscheZahl: sounds like a bad idea to me, losing too much information
Zenoscave: hmmmmm
eulerscheZahl: maybe considering some states more likely than others could help
Zenoscave: ok
eulerscheZahl: right now i assume equal probability for each
wlesavo: Zenoscave if you already considering such a drastic measures just merge
Zenoscave: merge?
Zenoscave: I don't understand
SavinVadim: Did anyone implement a kinda of circular movement?
Zenoscave: merge paths?
SavinVadim: I think that my movement (that goes from one side to another) gives too much info and is easy to predict
wlesavo: if to paths coming to the same point, take the common part of them and summ up the mines
eulerscheZahl: be the spiral SavinVadim
wlesavo: you lose information, but not too much
Zenoscave: interesting
Zenoscave: I'll try that
AntiSquid: walk into your garden and spiral around
wlesavo: this is what im doing since wood 1 i believe
blasterpoard: wlesavo I have something similar, but I only join pairs of paths
eulerscheZahl: including the mines that were added on bronze ;)
wlesavo: also only pairs
blasterpoard: basically divide the amount by two until it gets under a threshold
wlesavo: eulerscheZahl well, most of mine tracking for me was in silver, but the basic concept the same
pb4: eulerscheZahl : What are those graphs ?
eulerscheZahl: there's a text above them
pb4: and when did you download the replays ?
Zenoscave: top 4 player silence dists
eulerscheZahl: thank you Zenoscave
Zenoscave: :+1:
pb4: one graph = one player ?
Zenoscave: yes
eulerscheZahl: yes
Zenoscave: ;)
pb4: Which one am I ? :D
Zenoscave: 2
pb4: nice graphs by the way
eulerscheZahl: but the number of games i used varies a lot
eulerscheZahl: pb4 is the 2nd indeed, 32 matches analyzed
eulerscheZahl: i just downloaded #1 + me
so i got some of yours as well
pb4: wow, I surface 0 a lot of the time then !
eulerscheZahl: i didn't share your surface stats
pb4: silence*
eulerscheZahl: but you do 2x surface per game
eulerscheZahl: i remove SURFACE | KILL here
pb4: uh ?
pb4: I do less than that I think
pb4: 1.4
pb4: Your sample might be too small
eulerscheZahl: 2x per game, 1.75x per 100 turns of your bot
eulerscheZahl: ok, i'll check...
kovi: which one was mine?
eulerscheZahl: 3
eulerscheZahl: pb4 1.56 surface in 313 matches
eulerscheZahl: and charts for top players are converging towards each other it seems
eulerscheZahl: some days ago they looked really different
pb4: Agreed
eulerscheZahl: kovi even with a bit of sonar now
eulerscheZahl: still less than the other players
kovi: wow, you track me hard
YannT: I assume you didn't do me since I'm lower? :)
pb4: who was #1 ?
eulerscheZahl: jolin
pb4: and #4 ?
pb4: siman ?
AntiSquid: 2000 participants
dbdr: \o/
Illedan: 2k \o/
ZarthaxX: nice :O
eulerscheZahl: of course i got stats for you YannT
eulerscheZahl: you haven't submitted in ages, for you i got the most battles
eulerscheZahl: and that:
eulerscheZahl: crashes: 448128749 447871204 447869471 447863277 447861340 447860809 447859358 447858637 447857329 447855358 447854224 447853683 447853254 447799136 447797590 447782229 447663750 447653728 447648577 447646966 447646750 447641545 447638629 447633505 447632842 447629881 447627069 447627057 447624486 447624038
YannT: yeah :|
YannT: next version shouldn't have this
Illedan: Hiding?
YannT: no, benchmarking
YannT: I submited yesterday y'know
YannT: so not "ages" :p
Illedan: xD
eulerscheZahl: compared to the other spam submitters there's a lot of dust on your yellow button
AntiSquid: if you keep it up you'll suffer submit withdrawal symptoms
YannT: it's true I like to be somewhat sure before pushing
Zenoscave: 999 submits
Zenoscave: ...
YannT: benchmarking against top5 is more relevant than just pushing anyway in my estimation
Illedan: :D
eulerscheZahl: i have 721 matches for you YannT
eulerscheZahl: (your most recent bot that is)
AntiSquid: i am surprised i even have 292 submits
YannT: also I like that I can "freeze" agentIds used, so I can actually compare results
wlesavo: im benchmarking top 6 =)
blasterpoard: I have 49 ^^
eulerscheZahl: 141 C# submits + 1 Java here
blasterpoard: submits
YannT: otherwise those guy change their bot 10 times per minute so you can't compare anything
DEIZ21: hello, i have a problem with CG Spunk it doesnt work for me a,y idea?
eulerscheZahl: that tool broke years ago
YannT: I think that's deprecated
YannT: you want cgbenchmark
eulerscheZahl: and CG Enhancer
DEIZ21: ah, i need to run battle in batch mode
eulerscheZahl: cg benchmark
DEIZ21: i can search for it in CG?
eulerscheZahl: or brutaltester to play against yourself offline
YannT: (brutaltester I haven't found to be very pertinent with this game)
eulerscheZahl: https://github.com/s-vivien/CGBenchmark
DEIZ21: cool, thx
YannT: I have a version that seems to have a 50% winrate against top5
YannT: my rule is to do better than 50% before pushing tho :p
eulerscheZahl: my rule is to submit when i'm in a mood to do so
eulerscheZahl: or crying in desperation
dbdr: yolo
Illedan: I'm full on yolo
YannT: I can't for the life of me figure out why kovi is my best winrate and pb4 / siman are my worst ones
YannT: there's no strategy difference to explain it
eulerscheZahl: Illedan climbing high
i can't tolerate that, will fight back tomorrow
eulerscheZahl: my old enemy...
wlesavo: speaking about 50 % https://imgur.com/yVoTQYV
Illedan: :D
dbdr: battle for #1 C# is full on
Illedan: I'm going for 1st
YannT: I think I've won #1 kotlin already :joy:
dbdr: well done YannT, you can rest now
eulerscheZahl: doesn't seem to be a hard fight
YannT: ahah no, not really
eulerscheZahl: in the past there were tshirts for the first in each language
YannT: it's such a sweet language tho, I don't know why you'd do Java over Kotlin ever
eulerscheZahl: but that was long ago
eulerscheZahl: i used kotlin recently
eulerscheZahl: for kotlin heroes on codeforces
YannT: it's so friendly about everything I love it
eulerscheZahl: and maybe intelliJ helped a bit to convert java into kotlin :innocent:
dbdr: fully automated?
YannT: whenever you think "I should be able to write something like this", it turns out you can
eulerscheZahl: there's an intelliJ plugin for that
eulerscheZahl: still a few minor things (not-null !!)
eulerscheZahl: but did a great job overall
eulerscheZahl: and that !! isn't a bug, just kotlin having other requirements than java
YannT: null checking is actually a sweet feature of the language
YannT: you can write stuff like
eulerscheZahl: yeah
YannT: maybeNullVar?.let { ...
YannT: and the let will only be done if not null
YannT: you never get a NPE again
eulerscheZahl: one thing that sometimes annoys me about java: catch every damn exception
eulerscheZahl: i understand that you should really do it for an enterprise project
eulerscheZahl: but when i code my 10 lines for a small private project, i don't care about exceptions
dbdr: throws Exception
YannT: kotlin is good for that too
YannT: val a: Int? = try { parseInt(input) } catch (e: NumberFormatException) { null }
YannT: it's a convenient one liner
eulerscheZahl: and add that throws to each function in the call chain up to main
dbdr: yes, or script it ;)
eulerscheZahl: that wasn't the intention of java inventors
dbdr: or catch and wrap in RuntimeException
dbdr: sure
YannT: actually the "Int?." part here is even for clarity, it's not necessary
dbdr: the intention is to have robust code
YannT: val i = try { parseInt(input) } catch (e: NumberFormatException) { null }
dbdr: that's why python can be good for protoyping
dbdr: YannT, eulerscheZahl complains about having to catch exceptions, and you show Kotlin can catch exceptions ;)
YannT: except in a very compact way
eulerscheZahl: except: pass
eulerscheZahl: python style
Zenoscave: [empty everything]
Zenoscave: js style
YannT: oh then
YannT: "Kotlin does not have checked exceptions"
YannT: :p
Astrobytes: noexcept, it either runs or it doesn't
dbdr: how do you guarantee you handle them then?
YannT: the kotlin standard library abstracts the checked exceptions of JRE
eulerscheZahl: JS on CG https://imgur.com/a/QjCX3rJ
YannT: you get nulls and stuff
dbdr: I mean, if it's not checke it means you program might crash at runtime
YannT: no
dbdr: potentially in rare cases
YannT: it turns checked ezxceptions into nullables, which in kotlin have to be handled
YannT: so you might have something like
dbdr: null is just one case
YannT: val i = parseInt(blah) ?: 0
dbdr: what about IO errors?
aCat: hey -when gold boss became nondeterministic :/
Zenoscave: The whole time
aCat: that makes testing against him useless
YannT: val bufferedReader: BufferedReader = File("example.txt").bufferedReader()
val inputString = bufferedReader.use { it.readText() }
println(inputString)
aCat: hmm, why before I had imprssion it is otherwise..
dbdr: and if the read fails?
YannT: you don't get anything
dbdr: empty string?
YannT: no, unchecked exception
YannT: you can handle it if you want
dbdr: and i you don't you crash
YannT: otherwise you just get a null string
YannT: no
YannT: .use has built in exception handling
dbdr: ok, then it just does not go into that branch?
YannT: yep
YannT: that's the idea with kotlin
dbdr: sometimes that's fine
dbdr: sometimes it's not
dbdr: and this could be hiding the issue
YannT: well you can catch if you want, or you can use the kotlin abstraction and never have to catch anything , exceptions turn into nullables, so you just handle nullability of things (and the compiler makes you)
dbdr: everything looks fine, no error, but it does not do what you wanted, and no indication why
dbdr: ok, in that case it goes into the null handling
YannT: yes
YannT: if you want to handle a specific underlying exception, you can do that, and if you don't wanna it's just nullability handling, which in kotlin is the ? operation
dbdr: so that's fine, but it's roughly equivalent to handling exceptions, i.e. you have to handle it someho
YannT: operator*
YannT: yes, you have to add a ? somewhere
Astrobytes: Yeah but it's not as annoying as having to handle them in Java
dbdr: only if you want to ignore that case
dbdr: which you do sometimes but not always
YannT: path?.contains(this)
YannT: if path is null, no exception
YannT: that returns a Boolean? type tho, not Boolean, so you have to write
dbdr: sure, but if it's a logcal error not to have a path, it's bad to get false
YannT: path?.contains(this) == true
YannT: no, a Boolean? can be true false or null
YannT: basically you cannot get a nullpointer, compiler won't let you
dbdr: I know that
YannT: Boolean? is not Boolean
dbdr: i'm saying a crash is better than incorrect semantics
YannT: then you can do
YannT: path!!.contains(this)
dbdr: right
YannT: you assert that path is not null and it will crash if it is
YannT: you *can* get nullpointers if you specifically want to :)
dbdr: :)
elsinnombre2d2: which heuristics do u use to get a good path?
blasterpoard: elsinnombre2d2 well, I have a 600-line "heuristic" for that that runs for 20 milliseconds
blasterpoard: I don't think that there's a knwon, simple algorithm that gives you a "good path" in this game
blasterpoard: at least, not really good
elsinnombre2d2: ofc, but u use some rules
elsinnombre2d2: maybe like choose a cell that is near to one visited
blasterpoard: uhm... maybe I use an evaluation
blasterpoard: of a long path
elsinnombre2d2: and why all use c++? xD
blasterpoard: not just the next cell
YannT: a good starting point is using floodfill to calculate the number of available cells
blasterpoard: elsinnombre2d2 it's fast
elsinnombre2d2: than java too?
blasterpoard: java is not quite as fast
elsinnombre2d2: hmmm
Nerchio: im pretty sure you can get to the top with jkava
elsinnombre2d2: using floodfill, which depth do u get?
Nerchio: i am using floodfill for the whole map
Nerchio: for each of my possible moves
Nerchio: dont worry about this i think, you will have more problems with tracking the enemy
Nerchio: and potentially running out of time
AntiSquid: you can get to the top with python 3, but java is both uglier and slower than c++
blasterpoard: and you'll also hit the character limit faster with java
blasterpoard: that matters in this game at the top too
Nerchio: in intellij i see i have 36k chars
Nerchio: how much is the limit?
mlomb: 100kB
YannT: my bot is like 500 lines :p
pb4: -_-
Rbee: Is there any way to view the past questions on clash of code?
elsinnombre2d2: nah, my problem is finding the path, my tracking algorithm is not really expensive at this moment
Nerchio: it means it's not good enough haha :D
AntiSquid: golfed? YannT
YannT: nope, kotlined :p
Illedan: I'm at 60k chars
elsinnombre2d2: why floodfill? i cant see the utility
YannT: 561 lines, I just added a thing to test
YannT: :p
aCat: YannT impressed ;-)
Illedan: 2k lines
aCat: but I believe the lines are 'write once' never read
Illedan: kovi, close enough to top 4 now? ;)
aCat: haskell-style ;p
YannT: it's really the same kind of thing yes :)
kovi: nice illedan!
Illedan: dbdr and pb4 inc to ruin my party :(
wlesavo: wow
Nerchio: who are the guys writing Cpt. bots for this game
dbdr: wlesavo wrote LeChuck
wlesavo: everyone and noone
Astrobytes: The Wood bosses are written by the game authors and the other bosses are user-bots chosen by CG
Nerchio: so when legend opened the bot was chosen from one of the top at that time?
dbdr: yes
Astrobytes: Yes, w lesavo's bot (from that time) is the gold boss
Nerchio: it's too strong :(
Nerchio: i know who to blame now
YannT: consider lucky you didn't get his current bot as boss :p
wlesavo: it differs a little codewise, but i actully broke smth in a boss submit
blasterpoard: Nerchio we need a strong boss, because otherwise we'd need more games to climb in legend
Nerchio: yea makes sense
dbdr: also legend should be legendary ;)
Nerchio: it's cool that potential kill shots can be done from like 10 squares away
blasterpoard: ...and it's not *that* strong
Nerchio: i am working on implementing kill shots like the gold boss who teleports across the map to finish me :D
dbdr: TELEPORT X Y
blasterpoard: I don't have that in my bot
Nerchio: i just died to the cpt
Nerchio: imagine this
Nerchio: surface -> move and load 1 more power so you can use silence -> torpedo -> trigger mine
Nerchio: this was the kill shot from 3 hp
Nerchio: :(
Nerchio: feelsbad
blasterpoard: quite common
wlesavo: only smitsi teleports
AntiSquid: stronger boss doesn't solve slow submits
but it's good to have a strong boss since it will be a multi anyway and it goes straight into multi after this
AntiSquid: meaning most people stay active improve and flood legend
AntiSquid: maybe
YannT: multi bosses won't necessarily be the same though?
AntiSquid: i am sure they will, since the ladder doesn't change
AntiSquid: you go straight into same league as you weere in the contest
blasterpoard: YannT I think in crystal rush they are
kovi: gratz pb4
YannT: ahah 33 :D
JasonTruter: I'm in wood 1 now
Astrobytes: gratz JasonTruter
Nerchio: thats not bad congrats
JasonTruter: Just movement though
Nerchio: that's how most of us start
YannT: gratz JasonTruter :)
JasonTruter: Thanks guys
YannT: now time to fire some torpedoes
JasonTruter: Yeah
JasonTruter: I have silence and sonar now
Nerchio: i used sonar+movement+torpedoes until gold
JasonTruter: I've written my code to be more manageable now
JasonTruter: hated all the ifs
Nerchio: i live and breath if skyscrapers
icecream17: 8th place out of 6000, wow
JasonTruter: SHWEET
icecream17: I used to have a super complicated thing for CSB, but I replaced it with a single if statement. wow
Nerchio: what is csb
icecream17: coders strike back (i don't feel like doing ocean of code anymore)
Nerchio: :(
icecream17: I'm in silver, but can't make anymore progress. Every improvement makes it worse somehow
Nerchio: maybe you will come back with fresh mind :D
wlesavo: so pb4 was it just lucky submit or real improvements?
AntiSquid: go to discord and watch the streams, maybe you'll figure something out icecream17
or save the bot and start over and try something new
pb4: wlesavo : real improvements
JasonTruter: Question, should you use all the actions?
JasonTruter: just because they are there?
JasonTruter: I intend on just using sonar and torpedo
rwilson: JasonTruter torpedo too early & you give away your position. and if you don't use Silence, you won't be able to re-hide after your position is known.
JasonTruter: I see
pb4: JasonTruter : early on focus on torpedo/silence, then add mines
JasonTruter: When you fire a torpedo do you give away positon?
pb4: somewhat
JasonTruter: Oh using your life?
rwilson: early leagues easy to pass with torpedo and silence though ... silver & gold seem to be where early torpedos mess you over too hard to pass on.
pb4: You show you're somewhere in range of where you shot
JasonTruter: I really need to get some decent pathfinding going
JasonTruter: I think it's holding me back from moving towards a position
JasonTruter: Example if I need to move towards my opponent if I know he's in a specific location, or moving to sectors
NormantasStankevicius: Just curious guys, what do you normally do, when your opponent uses silence?
JasonTruter: Is it worth rewriting everything?
YannT: probably not, optimizing your path to stay underwater longer is generally more important than trying to get to the ennemy
Astrobytes: JasonTruter are you floodflilling from each possible direction?
Astrobytes: *floodfilling
JasonTruter: floodfilling?
Astrobytes: Yeah, just checking the most accesible cells from each position
Astrobytes: *accessible
JasonTruter: Ah I see, no I'm not flood filling right now
JasonTruter: I have my own algorithm that uses a bunch of different inputs to detect the area
Astrobytes: Flooding with a BFS is useful
JasonTruter: Yeah, then Im planning on BFS to location while tracking every move my opponent does and updating my destination everytime
Astrobytes: If you go in the direction with the most available accessible cells you'll maximise your tiem under water, for now
Astrobytes: *not correcting any more of my typos
JasonTruter: Maximizing time just reduces being seen right?
JasonTruter: time under water*
Astrobytes: Yeah, unless sonar/tracking etc, but it's a good start
JasonTruter: A tactic I could use is reset on edge of map, then BFS to opponent
JasonTruter: So I have a clear path
JasonTruter: Surface on edge of map *
Astrobytes: You don't wanna surface too much, you lose life remember
JasonTruter: Oh yeah forgot about that. But possibly do it once just to get a clear path
Astrobytes: Try going to cell with most available flooded cells and surface when you run out.
rwilson: if you watch the higher league replays, you'll see that most of the time anyone uses SURFACE is to be able to go back over their own path - because they don't want to keep going ( usually a mine field, or face check an enemy that knows their position )
JasonTruter: I see
JasonTruter: So best advice is find a path where you don't surface at all
rwilson: simple flood fill to avoid surfacing should be enough to help get up to bronze/silver ( with a bit extra)
Astrobytes: ^^
JasonTruter: I'm fairly new to algorithms :) so learning along the way
Astrobytes: Good time to learn dude!
JasonTruter: I work with high level code mostly, and understand data structures well
JasonTruter: All the algo's were built already
JasonTruter: So now I can actually think
rwilson: ;) and a fun way to do it... ;) so much easier to go learn BFS if you'lre make that next League ( versus .. just because :) )
JasonTruter: Yeah dude
JasonTruter: I'm actually hooked to this game
Astrobytes: You'll get hooked to the others too, I guarantee it :D
Astrobytes: Anyway, I'm out for tonight
Astrobytes: cya's later
JasonTruter: cheers!
YannT: it's actually not a bad idea to go to leaderboards and watch some high rank replays to get an iidea of what you probably should be doing ;)
YannT: https://www.codingame.com/contests/ocean-of-code/leaderboard
YannT: "view last battles"
YannT: although you don't have mines yet
Nerchio: how do you guys tackle life lost?
Nerchio: i mean people can surface/get hit by their/your mine your/their rocket etc.
rwilson: ;) life lost plays into a number of scenarios .. if you're careful, you can use most of them to gain more information
rwilson: the 2 main ones being, 1. they SHOULD know where you are down to at least 9 squares, unless it was a surface --- so you might want to run.
rwilson: 2. if they got hurt by your stuff, you can narrow your knowledge of where they are
Nerchio: i know i do most of this stuff
Nerchio: but what i am saying is
Nerchio: sometimes you get false positives
rwilson: ;) only if you don't code it to account..
Nerchio: like i hit a guy with direct rocket + a mine so 3hp
Nerchio: but i usually think that direct rocket is 2hp
rwilson: i.e. if you assume they lost life from your shot, but they also surfaced... then you've coded it wrong ;)
Nerchio: so if i code it for 3hp when they combo
Nerchio: what if i shot him 2 hp and he shot himself for 1hp
Nerchio: then i have false positive and my tracking my fall apart
rwilson: ;) it's only a false positive if you don't take into account the other information you have..
rwilson: i.e. you always know if he surfaced, you always know when he triggers, and you always know when they fire a torpedo ..
Nerchio: hmm
rwilson: if any of those comibnations could 'confuse' your knowledge of 'did i hit him?' ... then you must assume you did not
Nerchio: i should trust my stats more i guess
Nerchio: I do assume i did not and thats why i lost some games
rwilson: .... ;) or split it .. and record it probabilisitically , like MAYBE .. but can't use it as 100% knowledge like you can while figuring out path movements..
Nerchio: i already have enough splits because of silence :D
Nerchio: but actually you gave me a nice idea
rwilson: ;) might be able to improve that a bit... most of the time in my games, it looks like silence only hides for usually 1 silence ( meaning, by the time you can silence again, your position & prior path has converged
Nerchio: true
Nerchio: but i should consider 3hp as 2+1 so direct hit + something
Nerchio: i mean there could be stuff like 1+1+1 but that would be crazy :D
Nerchio: maybe not worth taking that into account
rwilson: my mine, their mine & surface, or a missed torp
Nerchio: i mean surface i basically ignore
Nerchio: in terms of hp lost
Nerchio: so its more like rocket close/rocket close/ mine or something
Nerchio: instead of direct rocket + something
Nerchio: i guess still lots of possibilities
rwilson: ;) unless you thought your mine detonation caused it ... ;) i watch to make sure i don't accidentally narrow location due to dmg.. when they actually just surfaced..
Nerchio: i did narrow down a few times too much and i lost track of everything thats why i want to be careful with this now
Nerchio: I need to remember if I shot 100% shot last turn or not
rwilson: ;) gets a bit trickier if you fire a torp and mine ... he takes 1 dmg.. you can still narrow - but you have to be careful with how you do it..
Nerchio: i can't make a mistake a direct shot for 2 close-torpedos for 1hp
Nerchio: yea
Nerchio: the one you are talking about
Nerchio: i mean
rwilson: ;) you don't really know / care if you THOUGHT it was 100% ... you just know the result ... use that to correct your 'belief' of where they are.
Nerchio: if torpedo and mine don't overlap you can't do much i guess
Nerchio: i will need to reconsider this part of my bot
Nerchio: still have some mistakes
rwilson: :) 225 cells max, torp and mine cover at most 18 in a turn .. even if they don't overlap , you a 1ph hit would confirm their in one of those 18 ( just don't get confuesd by surface, or enemy self damage )
Nerchio: that's what i mean they can hit themselves and unless you know their exact position you can't know
rwilson: lots of info ... and it does look like many of the top guys have that coded in already .. but it doesnt' look like thats the MAIN thing
Nerchio: sure but every turn counts so properly assessing this information is important
Nerchio: in some cases you probably just have to accept that you can't use the knowledge from hp lost
rwilson: ;) it's a giant pile of VEN Diagrams ;) you can always glean SOME information out of it ... it may not always be 'THEY ARE HERE "... but it's always ( are they inside or outside this set of spaces )
Nerchio: okay thanks i will think about this tomorrow but maybe focus first on fixing some other stuff
YannT: the only really confusing information is if they hit themselves
pb4: Why would it be confusing ?
rwilson: ;) was thinking the same thing...
rwilson: but i totally get what YannT is getting at ... you'd assume MOST of the time they wouldn't shoot themselves
YannT: also if they shoot themselves, you don't know when that happened in their turn (it could be before or after moving)
rwilson: but i coded in a special allowance to allow self damage if it wins... ( and i see a lot of top bots do MORE self dmg , even if it's a couple turns till it wins )
YannT: you know your torpedo hits before they move, you know when they fired theirs (before or after moving), but you don't know what hit what really
rwilson: ;) right, the area of uncertainty is larger than normal.. but shouldn't be too confusing ;)
YannT: so if they hit themselves, it can be confusing
YannT: you actually miss the information
Nerchio: that's what i am getting at
YannT: you fire a torpedo, they fire a torpedo, they lose life, you don't know which hit what
Nerchio: if you do damage and you expect it it's all good
Nerchio: but if they also fire a torpedo and hit themselves for 1hp or something it gets tricky
YannT: there's actually missing information, you can't know
rwilson: ... but you know the damage was done whilst inside one of those blast radius ... so .. keep all potentials inside.. and prune all outside ..
YannT: yep
YannT: still, it's confusing :p
Nerchio: yeah thats obvious
Nerchio: but ideally i would like to say, i see 3 positions i shoot a rocket for '2' hp so I would like to think its a direct hit
Nerchio: but if they also shot a rocket, was it really a direct hit?
rwilson: ;) and that's exactly what you learn :) .. you only get a direct hit, if you fire , do no other dmg, they don't sufrace , trigger or torp.. THEN you know it's a direct hit ;)
rwilson: SOMETIMES you 'KNOW' ;) but sometimes it'll be a cloud of 2 or 3 things .. ( that can still DRASTICALLY narrow down their location ... )
wlesavo: lol 40% vs pb4 and 50 vs illedan is enough to push me on their subs
rwilson: nice
pb4: push you up ?
wlesavo: yes
pb4: :D
wlesavo: from like 10-12 to 7
YannT: pushing small improvements :)
YannT: not doing well :D
icecream17: oooh 2nd place
icecream17: YAS. Cars collide with each other. huh
icecream17: What? 20998th? That doesn't even fit on the square...
Zenoscave: Are any legend players detirministic?
elderlybeginner: how to find which easy puzzles gives bronze achievements
elderlybeginner: ?
elderlybeginner: maybe achievement showcase shows them all?
struct: It shows all yes
elderlybeginner: :(
then it's time for medium ones
struct: Its a good ting
struct: thing*
Marcose: anyone?
