Chat:World/2020-08-16

From CG community
Revision as of 11:50, 15 June 2021 by Chat Log (talk | contribs) (Created page with "<img src=/a/43135502422075> eulerscheZahl: interesting read @JBM especially the parts prior to my own activity on CG, how the system was back then <img src=/a/45775905514269>...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

eulerscheZahl: interesting read @JBM especially the parts prior to my own activity on CG, how the system was back then

Allis: How long have you been the best?

eulerscheZahl: https://dbdr.github.io/cg/podium ask dbdr to update this then we know

eulerscheZahl: at some point CG changed their ranking formulas. with the old one i would still be far away

Allis: Interesting.

Allis: Are some categories weighted more heavily than others?

eulerscheZahl: and contest points decay over time now. back then you still profited from a good rank on a contest with many players 2 years prior

Gorbit99: cp calculations are mentioned somewhere

eulerscheZahl: questionmark at your rank on the profile page/start page

Gorbit99: yeah

Gorbit99: and cp under that

aCat: morning ;-)

eulerscheZahl: hi

aCat: hmm results are stable I see ;-)

eulerscheZahl: results for what?

aCat: Illedan contest ;]_

aCat: I've been watching you talking whole vening

aCat: *evening

eulerscheZahl: i'm around 147500, just haven't submitted

aCat: Yeah, some of next contributions have to be multi/opti for sure

aCat: watching you guys having fun is fun ;-)

eulerscheZahl: you are free to join

aCat: I know

aCat: I will

aCat: probably when I will provide enough content for the course that I won't be in fear of deadlines

aCat: I will start making the content myself ;-)

eulerscheZahl: i think we need a sokoban, maybe as a puzzle. too classic not to have it

aCat: definitely start with marslander :smile_cat:

aCat: yes

aCat: sokoban is in my plans for sure

eulerscheZahl: was about to call dips on it :(

aCat: I can send you a picture of the games / problems ai noted to add on CG ;p

eulerscheZahl: cool, go ahead and send it

aCat: actually I have about 3 such lists, overlapping for sure

aCat: not to mention projects aimed to be foll contest-ready multis

eulerscheZahl: sounds like you have more ideas than time to implement them

aCat: my usual issue in nearly every domain :crying_cat_face:

aCat: OK, I've sent them :joy_cat:

eulerscheZahl: thanks, will have a look

eulerscheZahl: your handwriting though. for some i can't even tell if it's English or Polish :D

eulerscheZahl: but some are readable, especially on your last picture

aCat: I have the same issue :P

aCat: first is the newet list - sokoban there missle right column :D

eulerscheZahl: yes, I saw that

eulerscheZahl: did you plan it as a puzzle or optim?

aCat: probably optim

aCat: but thet's problematic

aCat: as usually the problem with hardcore sokobaning is not the quality of solution

aCat: but to solve the level at all

aCat: so these means < 100% passed

eulerscheZahl: i agree. wrote a solver once. either it failed of found the shortest solution

aCat: but that's the issue to resolve in a future

eulerscheZahl: gets tricky with more boxes involved. then i might be stronger playing it by hand

Default avatar.png Momynul_Islam: http://chat.codingame.com/pastebin/38802455-0f9d-4d8e-9a50-4ad094aea5f4

eulerscheZahl: Momynul_Islam remove the GAME OVER line

jrke: and "bingo" also not needed

eulerscheZahl: for a moment it made me think it was a code for this puzzle: https://www.codingame.com/training/medium/bingo

jrke: its power of thor right?

eulerscheZahl: yes

Westicles: A good sokoban puzzle would be designing the longest level giving a set of walls. Like this

Westicles: http://sokoban-gild.com/

aCat: I have to write it down,. might be nice idea

aCat: ahh let us allow multiauthored contributions - I have feeling that some of my nonstandard ideas euler will like :P

eulerscheZahl: interesting idea. but needs a highly optimized solver to do the scoring

eulerscheZahl: then referee can timeout too, not just the user code

dbdr: I don't think it's necessarily bad if levels are hard to solve rather than hard to optimize

eulerscheZahl: i suggested having multiple authors long ago. when the SDK was still not public but about to be released

eulerscheZahl: hard to solve => makes a good puzzle I think

eulerscheZahl: not an optim

eulerscheZahl: same as for Space Maze

dbdr: NS is a great optim

dbdr: generated a lot of activity and research

jrke: dbdr is your bot hardcoded in samegame?

eulerscheZahl: so you suggest doing the same for sokoban? that might be an idea

eulerscheZahl: solve the most levels

dbdr: jrke: offline search, of course

aCat: dbdr NS?

dbdr: possibly, with harder and harder sokoban levels

eulerscheZahl: https://www.codingame.com/multiplayer/optimization/number-shifting

dbdr: but I was also responding to your conversation

eulerscheZahl: that could work, valuable contribution

aCat: ah yeah, Ive seen it, really cool contrib ;]

dbdr: actually what could be an issue if it's "easy" to get the optimal, several tied #1

aCat: ok, you get my hyped to check off the list next puzzle I'm working on ;]

aCat: breakfast, then cooking, then code ^^'

eulerscheZahl: i'll play soccer

dbdr: in this heat??

eulerscheZahl: every sunday if possible

eulerscheZahl: lockdown was a hard time for us

jrke: yeah!! lockdown was too much...

eulerscheZahl: we play at almost all weather. snow, heat, rain

dbdr: in the ice storm I got yesterday, I guarantee you would have stopped :D

eulerscheZahl: ice storm? are you kidding?

eulerscheZahl: it's august

dbdr: Hail - Hagel

dbdr: it happens more when it's hot

jrke: dunno what yesterday was independence day for india :) a kite flying day

eulerscheZahl: saw a heavy one as a child, maybe 20 years ago

dbdr: ice balls the size of peas falling hard, covering the ground

Default avatar.png Momynul_Islam: http://chat.codingame.com/pastebin/0d752bb1-e8a0-4149-bab2-d8eae80e223b

eulerscheZahl: even shattered the roof gutter

dbdr: was very impressive to watch

eulerscheZahl: bigger than peas, more like chessnuts

dbdr: wow, that would seriously hurt

eulerscheZahl: i was glad i was inside

dbdr: big kinetic energy

eulerscheZahl: Momynul_Islam i already told you what's wrong with your code

Default avatar.png Momynul_Islam: i don't know

Default avatar.png Momynul_Islam: tell me

Default avatar.png Momynul_Islam: can u fix it

eulerscheZahl: eulerscheZahl 08:46AM Momynul_Islam remove the GAME OVER line

Default avatar.png Momynul_Islam: give me your code

aCat: noone will give you his code

Default avatar.png Momynul_Islam: http://chat.codingame.com/pastebin/479645f0-2033-4eba-a5f9-a1fdefa7a7ac

aCat: when you pass the puzzle

aCat: you can watch other people's solutions under the solutions tab

Default avatar.png Momynul_Islam: i remove the game over line

Default avatar.png Momynul_Islam: but it still not working

Default avatar.png Momynul_Islam: ??

Default avatar.png Momynul_Islam: @eulerscheZahl

tutubalin: hello

Marchete: I've read past chats. euler, no, please, no more sokobans, I hate this game. I think it's from an old CGA sokoban game in PC, I just can't stand it. There are more games to implement, we already have one (Bender)

tutubalin: game viewer uses PixiJS, but docs say we need to use Java for that. How does it convert to JS? Can we use all the features of PixiJS, like filters for example?

Marchete: java for backend, PixiJS as frontend

tutubalin: but how they are bound?

Marchete: the real game is the referee in java

DaNinja: Momynul_Islam your N and S are wrong

tutubalin: Marchete

https://www.codingame.com/playgrounds/25775/codingame-sdk-documentation/entities-shapes-sprites

dbdr: sokoban on CG with the 4-color CGA pink palette, that's an idea!

Marchete: this

dbdr: but unironically

Thyl: Hello World

dbdr: hi

tutubalin: so we create a sprite on Java side. does it generate PixiJS code as result? Or how it works?

Marchete: well, that's a predefined way to do it

Marchete: like a CG framework

Thyl: hi dbdr

Marchete: you can tweak it to show different stuff

Marchete: check contributions, there are two of them that shows different stuff

Marchete: a 3D teapot https://www.codingame.com/contribute/view/52965be9af1e1e9f5b35487a3cc449f24bf2

Marchete: you can download the game source to see how it's done

Marchete: and https://www.codingame.com/contribute/view/5326e0ad9aaa1c84010096f697c2dc81eee2

Marchete: a different one :D

Default avatar.png Momynul_Islam: DaNinja thanks bro

Marchete: or example games https://www.codingame.com/playgrounds/25775/codingame-sdk-documentation/example-games

eulerscheZahl: was afk. I agree Marchete: sokoban has similarities with existing games on CG (e.g. Bender and Space Maze)

eulerscheZahl: tutubalin the CG SDK works the following: you write java code to create objects (rectangles, images, ...) that gets serialized to a json string. the SDK also provides JS code to deserialize it and create pixi objects from it

dbdr: (08:25:51 AM) eulerscheZahl: i think we need a sokoban, maybe as a puzzle. too classic not to have it

)

eulerscheZahl: let's make a sokoban poll then

dbdr: you will vote on both sides? ;)

Gorbit99: so when is pool getting made?

eulerscheZahl: it's possible to get around the SDK and create your own pixi objects or doing entirely without pixi

eulerscheZahl: you can also do parts on your own while using the SDK for others

dbdr: where's the pool? I wanna swimm

eulerscheZahl: i'm not sure about sokoban. cool game but will it be popular here?

Gorbit99: probably depends a lot on how much people have to wait for the fall contest

eulerscheZahl: space maze was a lot of work for 3 users to solve it and some more to at least look into it

eulerscheZahl: not counting myself in for the 3 solvers

dbdr: there are so many puzzles

dbdr: less visibility than optims and multis

eulerscheZahl: my own game passed the thibaud review to become a contest

eulerscheZahl: next step: convince jupoulton

dbdr: well done!

aCat: the one I've seen eulerscheZahl ?

eulerscheZahl: y

eulerscheZahl: i'll need testers if it gets approved. i'll get back to you (plural) then if it happens

aCat: ok, so I have to sit to this one for real so I can provide you good feedback

aCat: and schedule some talk

eulerscheZahl: you don't have to. but would be nice :)

aCat: as I have some comments on this one :smirk_cat:

eulerscheZahl: i won't rebuild the whole engine :P

aCat: I was hoping so ;-)

aCat: noo, but definitely have to work on ending condition

eulerscheZahl: we won't discuss this here

aCat: match eas resolved much earlier than the game really ends

aCat: I'm talking in generals

aCat: but ok, true :D

dbdr: let's extract information about this contest, bit by bit, validator style ;)

aCat: I would just say

aCat: that I LOVE THE TITLE

aCat: naming is just first-class

aCat: and no more spoilers from this cat

eulerscheZahl: :)

dbdr: thanks, all the information I needed ;)

DaNinja: I enjoyed Space Jam even though I still havent got 100%

DaNinja: but still hoping Space Jam episode 2 gets published some day

jrke: recolored are random or can be hardcoded

Gorbit99: afaik they are actually just one of the previous cases, but recolored

jrke: ya but are colored are changed every time or they are hardcoded maps

dbdr: hardcoded

MSmits: good morning guys. Managed to get transpositions working last nice

MSmits: night

dbdr: nice

MSmits: solves 14 moves before the end now as opposed to 4. Quite nice. I also wasn't able to solve any of the testcases in the IDE before

dbdr: how much effect?

MSmits: now I can solve all except the codingame one and the official testcases

MSmits: dont think it had much effect on my scores though. But it saves time if anything

MSmits: still didn't use zobrist btw. I went with a uint64[15] bitboard for the map and xorshfited each line and the gamescore up to that point into a hash

jrke: why is this im getting 50% in real ide and 100% in contibution ide with same code

MSmits: check which testcases fail

MSmits: its probably recoloring

MSmits: each official testcase is a validator but acat also recolored them and added them

MSmits: kinda redundant, but you have to take it into account

jrke: if recoloring then it should get 50% in contribution ide also

MSmits: did you hard code solutions jrke?

jrke: yes

dbdr: contriibution might have different testcases

MSmits: did you take into account recoloring jrke?

dbdr: yay, I'm improving my case 7 score. was one of my weak points

jrke: i have just hardcoded all test cases in contribution ide

MSmits: then it's recoloring

MSmits: wait did you hardcode 40 cases?

jrke: ok thanks will work on it

jrke: not 40 just 20 and made it for 2 cases

MSmits: ah

jrke: will add something for recoloring

MSmits: btw, what I did for recoloring is use my own colors. The first time a color appears it is assigned a color

jrke: ok

MSmits: that way it doesnt matter what the game tells me the color is, only the order matters

jrke: nice

Default avatar.png AKHILASREE.M: :innocent:

Default avatar.png AKHILASREE.M: hi everyone i m new to this codingame.

Allis: Is there some way of knowing what caused the notification badge on the Contribute tab?

Gorbit99: hey there

Allis: Or does it just signify "something got updated" and it's up to the user to find out what?

Gorbit99: it has a new over the thing that updated

Gorbit99: it just bugs sometimes imo

jacek: i feel followed

Marchete: :eyes:

jrke: MSmits thanks now 100%

YurkovAS: 60k in samegame with MC (online). 80k with small improved MC

MSmits: gj jrke

MSmits: YurkovAS I got 112k online with MCTS

MSmits: probably could have gotten somewhat higher with submit spam

YurkovAS: MSmits thanks. i'm try mcts later

YurkovAS: MSmits what algo you use in hypersonic? Can't improve beam search higher 150 place

MSmits: it's beam search

MSmits: but I keep all states up to ply 20-25

MSmits: I prune very well so that I dont get so many states

YurkovAS: What width (best states count)?

MSmits: that's really hard to say, I do states differently from everyone else

MSmits: a state for me is a map state

MSmits: excluding the player

MSmits: instead of keeping a position, i keep all possiible positions with this map state

YurkovAS: oh, you keep all states...

MSmits: so I have a "possible position" map

MSmits: 1 state is actually maybe 20-30 states

MSmits: as long as you do nothing to change the map (use a bomb, get an item etc.) it doesnt matter what route you took to your next objective

MSmits: so I floodfill from the last change of the map

MSmits: and only make a new state when an action is taken that changes the map

Default avatar.png PerpetualCocktail_6633: hello

Default avatar.png Parthparthu: hello

MSmits: I have several maps, bomb maps, item maps etc and they are all 128 bit

MSmits: I xor these together for a single state into a hash and i have an unordered set or map (forgot which)

MSmits: for transpositions

YurkovAS: MSmits thanks. not easy

MSmits: no it was pretty grueling

Default avatar.png Darshu010: hi

MSmits: in the end i only needed like 5 ms to make an optimum choice ignoring opponents

MSmits: with more search time or more beam width I got no improvement

MSmits: I was going to use the rest of the time to minimax survival/killing players, but I never got good code for that. That's probably the difference between me and the players above me

YurkovAS: i'm try beam search in samegame with 5k-10k best states, best score = 30k. But with MC 60k. Thinking about check MC in hypersonic :thinking:

Marchete: in hypersonic just beat kymiyuki bots, it's an auto top10

Saptarshi: can anyone tell me how do I use loops in python in a more simpler way...like the one line if else statement

YurkovAS: don't understand kymiyuki code

dbdr: step 1: learn japanese

Marchete: nobody understand :D

Andriamanitra: @Saptarshi can you elaborate on what you mean by "simpler"?

Andriamanitra: if we don't know what you're currently doing it's hard to know is there's a simpler way :D

Marchete: just change simple numbers

Marchete: like the evaluation, beam width and depth

Marchete: it has bugs also

dbdr: Marchete, how much CPU power are you using? :)

Marchete: like incorrect simulation in some cases

Marchete: 5 threads

Marchete: but diminishing returns

dbdr: desktop CPU?

Marchete: yes

dbdr: yeah, that's always dimishing

Marchete: like 3 min -> 150k

Marchete: 1 hr -> 160k

Marchete: infinite -> 165k

dbdr: yeah

dbdr: wait, you can get 150k in 3 minutes?

Marchete: or maybe 5

Marchete: I can't recall

Marchete: it may be just luck

dbdr: that's 9 seconds per case

dbdr: less than online

Marchete: no

Marchete: 5 min * 5 cores

dbdr: oh, threads

dbdr: still, 45 seconds

Marchete: 25 min / 20 tests

dbdr: well, Neum*n was saying his CPU is 4 times faster than mine

dbdr: so I think those numbers are reasonable

Marchete: I have an Intel core i7 8700K

Default avatar.png JBM: thanks eulerscheZahl, glad you liked it!

dbdr: yep, 4.5 times faster (adding cores) :D

dbdr: https://www.cpubenchmark.net/compare/Intel-i7-5600U-vs-Intel-i7-8700K/2456vs3098

dbdr: only 1.5 faster per thread

Marchete: yeah, must be certainly faster

Marchete: and it doesn't feel at full speed

Marchete: when you use a laptop for CPU intensive use

Marchete: most of them sounds like a jet

Marchete: sound9*

Marchete: whatever

dbdr: it's fine. I hate noise

dbdr: you might think of those power laptops

dbdr: not sure what the usual term is

dbdr: mine is rather optimized for low power/heat

dbdr: 15W

Marchete: mobile workstation or gaming laptops

dbdr: right

Marchete: but I mean normal laptops that start melting on high CPU use

dbdr: depends what you consider normal ;)

Marchete: I had one that cried a lot while the NumberShifting rush hour

Marchete: annoying fan sound that is noticeable and doesn't let you concentrate

Marchete: poor soul, it took like 8 hrs for a single level

dbdr: yes. if my fan starts up, I know something is wrong, like a CG tab wasting CPU

SPDene: if you have some superglue, you can eliminate that fan sound :P :P

Marchete: I just removed the fan, problem solved!

dbdr: using efficient software also works

Andriamanitra: this fucking piece of shit clash can go fuck itself with a sharp object

Andriamanitra: why would the constraints say n can be huge if none of the test cases require it

Andriamanitra: struggled for 15 minutes to find out everyone else did it the easy way that only works for small n

jacek: o:

Marchete: :O

KyawMyotun: :joy :

KyawMyotun: https://www.codingame.com/clashofcode/clash/13395356aa451f42abc5926f3f5095b5ea738e4

eulerscheZahl: dbdr 12:38PM using efficient software also works

did you turn off the fan in the system settings?

dbdr: what system settings?

dbdr: no, fan starts when it's needed. I rather make sure cpu is not used without good reason

eulerscheZahl: just a bad joke...

eulerscheZahl: i removed 1 fan from my computer physically. wasn't spinning well

eulerscheZahl: always touching some part and making a terrible sound

eulerscheZahl: and at a position with little relevance

dbdr: wasn't that part out of place?

dbdr: happened to me once with a cable, but I moved the cable

eulerscheZahl: i never figured out what exactly was wrong

eulerscheZahl: i guess the fan wasn't perfectly centered anymore, touching its own frame

dbdr: oh ok

eulerscheZahl: so i unplugged it

dbdr: desktop?

eulerscheZahl: yes

dbdr: eulerscheZahl, still running?

dbdr: SG

dbdr: Westicles passed 173K

eulerscheZahl: nah

MSmits: I'm still working on meta mcts.

Marchete: do you fixed your bug?

MSmits: ah yeah

MSmits: it was stupid (as usual)

MSmits: I added transpositions. What I do in mcts, is store the childIndex of the transposed Node so it can share children with other transpositions. So if the childindex was still 0 it would check if a transposed node existed and get a new index from there. But after I set this, I didn't change the childindex variable in the function. So it was still 0. I would get all sorts of weird inconsistent maps

MSmits: it was literally just changing childIndex to node->childindex to get the right number and all was solved

wlesavo: how fast is checking for existence of transposed node?

MSmits: whats a more important question is how fast is it compared to expanding a node because if the transposed node isnt found it needs to do its own expansion

MSmits: it's begligible basically. Because expanding a node involves doing a flood fill for every possible move from the current map

MSmits: negligible that is

wlesavo: oh, i see

MSmits: my map is a uint64_t[15], to get a haskey I xor all lines together

MSmits: then i look it up in an unordered set

wlesavo: didn't think of that, rollouts with floodfills definetly are quite slow

MSmits: yeah

MSmits: the biggest benefit from transposition is the solving I think

MSmits: I can solve the game 14 moves before the end now, that's online

MSmits: so only using 50 ms I guess

wlesavo: nice

MSmits: dont know how it is offline, havent checked

MSmits: if you do the horizontal lines thing and you dont have transpositions, you cant even solve on turn 1

MSmits: I solved at turn 4 i think

MSmits: now it's easily managed

MSmits: ohhh I just figured out what the bar under my rank means

MSmits: thats progress toward next rank.

MSmits: does anyone see a semi filled bar?

MSmits: I see it for xp

MSmits: I wonder if they interpolate cp or rank for this bar

wlesavo: what bar

MSmits: above quest bar

MSmits: map

MSmits: you're not guru, you should see something

MSmits: 67 ranks to go till guru for you, but in cp terms it's probably a larger fraction of the distance between the start of grandmaster and rank 100

wlesavo: ah i see, basically its cp on a range from previous rank to the next one, for guru it could be to top#1

MSmits: so i wonder if they use rank or cp

MSmits: hm yes, but it doesnt have to be cp

MSmits: it can also be rank

wlesavo: cp for me

MSmits: oh ok

MSmits: says nothing for me. just "23rd"

wlesavo: hm, maybe im wrong actually

MSmits: it's based on rank then?

MSmits: that would be a bit misleading. You could get the sense that you're almost there, when in cp terms it could still take thousands

dbdr: https://i.snipboard.io/Cq24fH.jpg

MSmits: if the junp from 101 to 100 is very big

MSmits: stop hacking dbdr :P

dbdr: why?

MSmits: you're not 0th with level 255 :P

dbdr: I meant, why stop?

MSmits: oh, I actually don't have a good answer for that

dbdr: :nerd:

MSmits: :P

dbdr: hacking has a noble meaning

MSmits: in expert circles, not in laymans terms

dbdr: and who are we? ;)

MSmits: dunno what you are, but I am a hacking noob

dbdr: "A computer <https://en.wikipedia.org/wiki/Computer> hacker is a computer expert who uses their technical knowledge to overcome a problem."

wlesavo: MSmits rank confirmed, 83.75% for me which is consistent with progress bar

dbdr: first sentence on WP :)

MSmits: though, I did ace my internet security class :)

MSmits: ow if thats a hacker, then we all are

MSmits: ah ok wlesavo thanks, mystery solved

dbdr: :+1:

MSmits: I'm doing a game design class in november and a functional programming class in february

MSmits: as a student I mean

dbdr: though I don't think that sentence is idea

MSmits: might be a bit general

dbdr: it's missing something about creativity, doing things in unexpected ways

dbdr: wow, interesting classes!

dbdr: are you full time working + classes, or...?

MSmits: yeah, looking forward to the game design one. Not sure about the functional programming one, but i'l lprobabyl like that too

wlesavo: in terms of cp its close to 77%, so a little bit missleading but does not really matter much

MSmits: I work monday, wednesday and thursday this year, got an extra day off to study

MSmits: paid i mean, it's a 1 day a week scholarship basically

dbdr: I think FP is really good to know, even if you don't end up using a purely functional language

MSmits: yeah I think so too

dbdr: nice setup

eulerscheZahl: the C# LINQ stuff is functional btw

MSmits: yeah I figured that. I always wanted to become better at that

eulerscheZahl: but there are more extreme examples like Haskell or Clojure

dbdr: yes, most recent languages incorporate some FP ideas

dbdr: more and more over time

MSmits: python has a lot of it too doesnt it?

eulerscheZahl: yes

MSmits: the course will use elm

eulerscheZahl: i only know Elmo

MSmits: but the ideas are what matters I guess

MSmits: Elmo is an earlier version

dbdr: and the teacher!

MSmits: hence the Elm0

eulerscheZahl: no, this one https://www.youtube.com/watch?v=vSYadh2xmcI

MSmits: lol I got that :p

eulerscheZahl: oh

MSmits: elmo appears a lot in jokes on late night shows somehow

MSmits: mostly because of misbehaving on times square

dbdr: John Olliver did one with Elmo I think

MSmits: more than one :P

MSmits: pretty sure he did more than one on his months at the daily show alone

dbdr: https://www.vox.com/2016/4/18/11449756/john-oliver-sesame-street-lead-last-week-tonight

MSmits: ah yeah remember that

MSmits: lol that tucker calrlson getting upset over sesame street. That was a new level of sad...

kovi_not_hardcode: ah, better this way

dbdr: :D

wlesavo: lol

dbdr: pseudo updated on pidgin but not webchat

dbdr: I guess reload would do it

wlesavo: no_hardkovi

dbdr: will someone grab the kovi username now?

wlesavo: dbdr you are free to go :smiley:

dbdr: i'm fine :)

wlesavo: and ill grab dbdr :slight_smile:

dbdr: already got a FLU

MSmits: lol kovi really dislikes the hardcoding :)

dbdr: oh, he got scared

wlesavo: its like euler and coc

kovi: was thinking about that dbdr :)

dbdr: I was joking, but who knows

kovi: and i dont expect it from you, but who knows

kovi: yes

AntiSquid: just make an alt, use VPN wlesavo, nobody needs to know your dirty kovi nickname stealing secret

dbdr: I thought you created an alt, to have both a online and an offline submission

wlesavo: AntiSquid ill steal kovi_not_hardcode and will make a shitty submit to optims :smiley:

kovi: not a bad idea. but i just wanted to show that may 12x point is not the worst of the hardcoders, but the best of the nonhardcoders :)

MSmits: what I don't like about online submission is the greater luckfactor. You need to spam submit to get a better score

dbdr: wlesavo lol

AntiSquid: pb4's nickname was once up for grabs, took it for a few hours, nobody noticed @_@

MSmits: evil action snoone notices are tight

dbdr: MSrnits is free...

MSmits: free?

wlesavo: lol

dbdr: even you get fooled :)

dbdr: it is

dbdr: I should have done it

AntiSquid: MicroSoft right now its

MSmits: ohh

MSmits: the name

MSmits: how is it free?

MSmits: oh r n its

MSmits: sneaky

wlesavo: double lol

dbdr: ok, I'll do it

MSmits: no no

Marchete: yes yes

AntiSquid: you could get eulerscheZahI just replace L at the end with uppercase i

wlesavo: Marсhеtе is free

dbdr: cyrillic?

wlesavo: yes dbdr :slight_smile:

dbdr: they might check that. or not

wlesavo: nah, its for sure invalid characters

wlesavo: invalid = free

AntiSquid: what does dbdr stand for anyway?

dbdr: Dbdr dBdr dbDr dbdR

MSmits: database doctor?

MSmits: he fixes all your SQL issues

dbdr: :scream:

AntiSquid: using mongoDB though

wlesavo: mystery solved

dbdr: too much work

dbdr: wlesavo: did you check bulls and cows scores recently?

RJMontalvo: my peanut butter jelly bot died of over spreading :(

MSmits: dbdr anything weird happen in B&C?

dbdr: the magic of randomized validators

MSmits: randomization got cracked?

dbdr: nono

MSmits: spam submit?>

dbdr: yes

dbdr: we don't even have the referee I think

MSmits: thats what kovi wants for samegame :p

MSmits: if you're not hardcoding then there is no difference between the validators being random or the bots doing random things

MSmits: in fact the more parameters you put in your bot that you fit using submission feedback, the more your bot looks like a hardcoded bot

Default avatar.png scofield2: :grinning:

dbdr: rtue

MSmits: this is why all those NN's on boardgames always play the same moves. There really is not that much difference between a oware NN and a oware bot with a massive opening book

MSmits: jacek's oware bot is deterministic 50 plies deep with only 1 variation. If I play the same moves each time

dbdr: yeah, NN is just the learning mechanism

wlesavo: oh wow, nice random dbdr :slight_smile: but i believe your lead comes from some features

dbdr: thanks for believing in me :D

MSmits: mmh been running 4 instances of samegame solver all day

wlesavo: well others tried spam submit as well, there deffinetly should be some magic twist :slight_smile:

MSmits: only 1 improvement so far

kovi: 291? :o

dbdr: wow, that's stagnation MSmits

kovi: something is fishy there in b&c with (reasonaly) latest submit the first 10 had 2x304. since than i rarely got below 310

dbdr: randomness is like that

MSmits: yes randomness is fishy

**dbdr slaps MSmits around a bit with a large fishbot

MSmits: :four_leaf_clover: :fish:

wlesavo: generation probably could be improoved in b&c not much though

dbdr: what could be improved?

wlesavo: i think im not handling some edge cases and it is possible to generate a number with additional bull

wlesavo: which could give an advantage, and also as we disscused number can be generated online, up to the last possible number

jacek: b&c?

wlesavo: bulls and cows 2

kovi: wel, 304 seems like a hard limit for my current algo (over lots of attempts), so i might be very lucky on the first few

eulerscheZahl: i spam submitted the cows :(

MSmits: shoulda gone with the bulls for a better rank

eulerscheZahl: oh, didn't even notice the latest movement at the top

kovi: and considering the spread/range (315 +/-10), dbdr is definitely much better

eulerscheZahl: yes

dbdr: I have one 291, one 294 and one 299

eulerscheZahl: https://prnt.sc/u0gztn i'm nowhere near this

dbdr: so I might have something better since you never got any, but also a lot of luck involved

kovi: http://chat.codingame.com/pastebin/85000c79-e02b-4f1b-a418-2495b983099f

kovi: and since than just a bit better than euler's

dbdr: eulerscheZahl I think you knew about BrainSolver at 298

kovi: anyway, i agree that online submit also has its problems

dbdr: teasing me about how badly I had been beaten :D

wlesavo: lol, yeah, remember that

dbdr: "thanks for the motivation" ;)

kovi: especially as testcase-scores are not normalized (for zombies 1 testcase can dominate)

MSmits: yeah hardcoding solves this somewhat, if everyone has the max score, then it no longer dominates

dbdr: yes. but it's tricky to normalize since by definition you don't know the max possible score

MSmits: so maybe you should make all new optims then dbdr

MSmits: win them first, then normalize

dbdr: :D

eulerscheZahl: maybe i saw it and forgot again

kovi: yes, tc has relative scoring

dbdr: ah, old age...

wlesavo: in the case of SG offline is fine because of this standartised testset, would be fun to have more testsets like this

dbdr: you mean other games?

wlesavo: i mean more games yeah

kovi: well, best would be have two scoreboard :)

dbdr: not ... enough ... games ...

eulerscheZahl: create one

eulerscheZahl: or many

dbdr: one with kovi?

wlesavo: actually before i didn't even think someone would bother to make such a testset for a random game to benchmark different algos

dbdr: maybe it happened organically

kovi: not sure how many make strong attempt with hardcode. probably a dozen

wlesavo: maybe SG is not just a random game ofc, but still

dbdr: they just made those 20 cases for that casual js game site, no?

dbdr: then researchers took hold of it and spent years on those 20

dbdr: it's actually funny to see those research papers referring js-games.de :D

kovi: oh, i was not care earlier, are the actual judges known cases?

wlesavo: hm, maybe you right, but the point stands, its fun to compare results to the others even outside the sg, not that i have much to compare :smiley:

dbdr: yes, I agree

dbdr: but in other cases, we can say CG is the standard

dbdr: I'm sure research papers will eventually refer to CG

dbdr: kovi, the 20 cases you mean?

kovi: westicles improved? wasnt it 172k earlier?

wlesavo: cg constrains make perfect sence for many games, yeah

dbdr: he did

Marchete: research papers are like "I used distributed MCTS on a supercomputer and got 78k points"

MSmits: i wonder if he fixed his case 9 :)

MSmits: he was convinced some scientist frauded the case score

MSmits: i can't believe some people can do this for a living

Marchete: and some guys with a crappy laptop are better in js-games.de

MSmits: haha yeah

MSmits: westicles and his crappy laptop :)

Marchete: the supercomputer part is true

MSmits: no i know, these guys get to use their faculty's supercomputer

eulerscheZahl: 78k is at least 156k on CG leaderboard. but still a waste of a supercomputer

MSmits: they get assigned x hours of use

eulerscheZahl: does anyone check the program before approving the time?

MSmits: doubt it

wlesavo: how the score is calculated on jsgames? by ten lowest score sets?

dbdr: 10 GOTO 10

eulerscheZahl: sum of all wlesavo

eulerscheZahl: and for CG you have to do *2

MSmits: because of recolored cases...

eulerscheZahl: as the cat doubled the validators for whatever reason

MSmits: which i still think is silly

wlesavo: oh, i see, we have 40 cases

Marchete: I think it's the same score

wlesavo: forgot about that

Marchete: at least I submited Test 1 to check

MSmits: eulerscheZahl if you consider it a regular puzzle, the recolored part would make for an exatra test, but no need to recolor all 20

eulerscheZahl: well, it's not a puzzle :P

MSmits: that too

MSmits: Marchete what search type do you use btw?

MSmits: forgot to ask

Marchete: I ended matching the validator and returning the solution[i%20]

Marchete: similar to Numbershifting

Marchete: variations of SA

MSmits: oh ok

Marchete: maybe I'll try some MCTS algo

MSmits: mine is really simple and does well

MSmits: if I added more features it might be better

Marchete: so add it

eulerscheZahl: any fancy scoring?

MSmits: i tried a few from the papers

MSmits: didnt help

MSmits: no just high score

MSmits: i use each nodes highscore divided by parent high score in place of node value

kovi: besides hardcoding i still havent read any papers

MSmits: so if a child has a the max score, the node value = 1

eulerscheZahl: i didn't read any papers either

eulerscheZahl: too lazy

dbdr: you hardcode kovi?

MSmits: i read one on mcts

kovi: bad english

MSmits: had lot of fancy stuff that didnt work for me :P

dbdr: :D

MSmits: the papers or you kovi_not_hardcode?

MSmits: cuz... some papers :P

kovi: i had lots of failed heuristics. and some working well

MSmits: i only use one heuristic

MSmits: tobucolor i think it is

MSmits: in the random rollout

MSmits: other than that its just all fairly optimized code, nothing special

eulerscheZahl: i tried rewarding few stones without any group but meh

MSmits: i tried node value as average score instead of high score

MSmits: but that peformed worse as opposed to what the paper said

wlesavo: i have a fixed order for ccolors for a single rollouts, the only one that really worked, but i guess it can be worse than tabu color

MSmits: what do you mean fixed order?

MSmits: do you just set an order and then do the rollout?

eulerscheZahl: death to green

Marchete: there are a lot of papers that use acronyms, and there is nothing more in the internet about them

Marchete: Heuristically Guided Swarm Tree Search

wlesavo: if possible i try to choose first color in shuffled list of colors

Marchete: I've seen like zero info apart from 1 damn paper

MSmits: maybe try different search words

dbdr: Marchete: it's a troll paper ;)

eulerscheZahl: and searching for "same game" isn't that simple either

MSmits: like "Run away, this tree has bees"

dbdr: refer to amazing results from non-existing technique / paper

MSmits: eulerscheZahl it's like searching for breakthrough

eulerscheZahl: yeah :D

dbdr: always wondered about the name

dbdr: same as what?

eulerscheZahl: same colors

MSmits: the cells are the same color

MSmits: they disappear

dbdr: ah

dbdr: :D

MSmits: is this guy nr 1 in our optims?

MSmits: jeez

dbdr: rather sounds like "same game as that other game", which does not make sense

wlesavo: MSmits yeah, basically fixed order, in case of simple MC with not a lot of rollouts it works quite ok, but generally it should be worse then tabucolor

dbdr: yeah, dumb

MSmits: :P

Marchete: what algo was using W?

Marchete: because I tend to pick the worst possible

wlesavo: dbdr that was what i though, same old game, like the very simple game with intuitive rules, lol

MSmits: wlesavo i am not sure, tabucolor in general gives higher score, but you also keep using the same branches

MSmits: varying branches also has value

MSmits: thats why i am hoping meta mcts will help a bit

kovi: fixed order help with limited time

MSmits: yeah

MSmits: thats what i was thinking, with hardcoding it may end up hurting you

MSmits: less chances for lucky high score

Marchete: rule of thumb, as you can work offline, always give a chance to worse options

MSmits: good rule

Marchete: like 90% your super algo, but 10% random

MSmits: yeah

MSmits: the mcts paper used epsilon greedy 0.0002

MSmits: does that mean 0.02% chance to pick random?

wlesavo: simmilar to policy convergence theorem

wlesavo: MSmits yes

MSmits: seems extremely low chance

MSmits: 2 in 10k

wlesavo: well if you have unlimited time its ok

MSmits: they didnt though

wlesavo: oh

MSmits: they used a minute o rless

MSmits: i would at least use 1% I think

dbdr: lol

wlesavo: or decreasing by time value

kovi: i think fixed order hurt online as well. first day i tried dual-taboo, it sometimes helped

dbdr: a ridiculously low random is a nice trick to be able to prove "given enough time, it will find the optimal solution"

wlesavo: i think it depends on the range of your score :smiley:

MSmits: ahh i remembered the wrong value

MSmits: it's 0.3%

MSmits: http://chat.codingame.com/pastebin/d96a86a6-656d-4521-8b9e-995d0cfb96f0

Marchete: I'm not academic, but so many papers seem just fillers to me

Marchete: like filling CV and/or goals

MSmits: yes seems to me as well

wlesavo: especially in computure science yeah, but in other field as well

MSmits: btw, about the quote from the paper just now. Do they mean chance epsilon to have the first move of the playout be different, or every move? Or have a chance for every move?

dbdr: I think there has been pressure on academia that pushed in that direction as a perverse effect

kovi: well...how is it different from our competition. inching for the last 0.0x% improvement to beat others :)

dbdr: MSmits: I think every move

MSmits: so every time you do a move in the playout, you reroll the dice?

dbdr: the goal of academia should be to increase knowledge, and just can't quantify that in a useful way

dbdr: if you evaluate people based on how many papers they publish, they will publish more

MSmits: if knowledge becomes hard to find in a mountain of filler papers, it actually reduces knowledge

dbdr: I agree

dbdr: the evaluation policy is the problem

MSmits: which papers are approved you mean?

wlesavo: dbdr in russia last couple of years it went in extreme mode

dbdr: a lot of major breakthroughs have been from people who were not able to show results for a decade or more

Westicles: Ha, I found an old writeup of my approach. It sounds so stupid

MSmits: I would not mind reading your stupid stuff

wlesavo: publish it, a decade went over :slight_smile:

Marchete: I want to hear stupid things, Westicles :ear_tone1:

Westicles: http://chat.codingame.com/pastebin/ab240ea0-1cb7-4862-9584-216b943a6b38

MSmits: that doesn't sound stupid at all, it sounds like something I would need to think about for a day before I figure out why it works

Westicles: I think it is just the same thing as NS

MSmits: I never understood what you did there either, aside from it being SA

wlesavo: i think its very close to what i tried in NS as well

dbdr: sounds much more complicated than what I had in NS

wlesavo: but yeah, some details are quite interesting

MSmits: started two instances of solver for testcase 2 with epsilon 1%... see if it helps

kovi: my solution has similar to that westicles

MSmits: it's my worst score compared to the maximum I've found

kovi: except the repeat million times of course as online :)

MSmits: submit spam does that part

wlesavo: kovi and a one minute part :slight_smile:

kovi: i dont spam submit, just a few to get some average

MSmits: I'm just kidding, it wont let you

MSmits: I hate what happens on submit spam.

MSmits: first it lets click robot thingies which is fine

MSmits: but then it says "try again in a few minutes" FOR AN HOUR

dbdr: imagine getting "try again in a few years"

MSmits: lol

dbdr: on the whole account

MSmits: I deserve that for running so much cg benchmark :P

dbdr: you've reached your 2020 quota, try again next year

wlesavo: right before the contest

dbdr: :D

wlesavo: or after submiting a broken bot

MSmits: lol

MSmits: that actually happened to me in locam speed

dbdr: spam?

dbdr: or just the meltdown?

MSmits: no i submitted a broken bot

MSmits: but i didnt know it was broken because the feature i coded was in the next league

MSmits: I missed one space

MSmits: couldnt wait for submit to finish so i was just hoping it would work

eulerscheZahl: CG will go out of business just because of smits spamming games

MSmits: heh I dont do it as much now :)

MSmits: but I did some crazy amounts during my PCR testing. Thousands a week.

dbdr: thousands of benches?

MSmits: games

MSmits: at most I think 8000 games in a week

MSmits: it's mostly because PCR needs so many to get a signal out of the randomness

dbdr: ~1 game per minute non stop for a week :)

MSmits: sounds right yeah

MSmits: thanks to me, codingame is ready for server load during contests :P

kovi: i approx. did ~1 per minute latest games with pure submits

kovi: (hundrends of submits for few weeks, 50-100 games per submit)

MSmits: for a week non stop?

MSmits: wow

MSmits: allright, you are worse than me

MSmits: didn't know it was possible

dbdr: no, that should trigger rate limiting

eulerscheZahl: i also think there's a quota per game, isn't it?

eulerscheZahl: so you can benchmark 10 different arenas at the same time

MSmits: yeah thats true

dbdr: harder to analyze in your brain

MSmits: I never misused that though

MSmits: but could have

kovi: but i meant submit, not bench

MSmits: I could benchmark several boardgames at the same time if I wanted

dbdr: oh right

eulerscheZahl: yeah, we saw you the last contests kovi

eulerscheZahl: also: CG has 2 queues. 1 for submit, 1 for IDE games. the IDE has higher priority

MSmits: which is dangerous

MSmits: with people using CG

eulerscheZahl: that's why they added the delays when CG Spunk became popular

eulerscheZahl: as testing users slowed down submits

kovi: bench is better within top5, submit spam is better when not there yet

dbdr: I think when you are far from top 5, you can usually stop something to improve without spamming

dbdr: *spot

kovi: my spamming is never to have better score, but to assess latest changes

kovi: vs. other (more realistically than single submit)

dbdr: I was not speaking about better score

kovi: oh spot, i see

kovi: well, yes but spotting/getting an idea is one part, checking how much that helps is another

dbdr: yes, but knowing that precisely is not very relevant

dbdr: it might even lead to a local maximum

kovi: not precisely, but sometimes the result from a single submit is irrealistically good/bad

dbdr: yes

dbdr: I guess my point is, improving the logic of a bot is more important than optimizing against a field of bots that will changes anyways

dbdr: except near the end

kovi: valid point

kovi: but with limited time one has to assess current improvements to have better prioritization of further improvements

kovi: wow, euler back to 4th

kovi: and i still cant reach illedan

dbdr: eulerscheZahl: CEST is UTC+2, not UTC-2 ;)

eulerscheZahl: ?

eulerscheZahl: what are you talking about?

dbdr: joking

dbdr: bc you submitted just above MSmits, close to 6pm CEST

eulerscheZahl: context?

eulerscheZahl: ah, the illedan "contest"

eulerscheZahl: already half forgot about that again :D

dbdr: and end is at 8pm UTC

eulerscheZahl: chat is laggy

dbdr: hider ;)

dbdr: very laggy

eulerscheZahl: and i can beat smits whenever i want to

eulerscheZahl: considering that i have the skills which i don't always have

dbdr: that trash talk :D

dbdr: ?

eulerscheZahl: considering is the wrong word

kovi: yay, another odd score

dbdr: :D

dbdr: easy to fake ;)

eulerscheZahl: let me rephrase: it's my own decision when i want to submit a solution to beat smits

kovi: ;P

dbdr: provided you have a higher local score

eulerscheZahl: which i knew i had

eulerscheZahl: but probably not higher than his local

dbdr: that means you are improving slowly too

dbdr: if you submitted your best

eulerscheZahl: stuck now

eulerscheZahl: i mean i can go up like +20

dbdr: I thought you said you stopped

eulerscheZahl: started the solver again

eulerscheZahl: when i wrote it, it was true

eulerscheZahl: but at that time i didn't have my strongest in the arena

dbdr: dbdr asks euler stop "I stopped" euler restarts

dbdr: ;)

eulerscheZahl: you swapped the first 2 lines

dbdr: less funny

eulerscheZahl: i would just lie to you rather than stopping in the middle of a search that only saves at the end

dbdr: what's your score on case 7?

eulerscheZahl: http://chat.codingame.com/pastebin/61912f94-d120-4e9d-89e6-a7ac61ffdced

dbdr: I have 2771, definitely a weak point

dbdr: thx

kovi: case20 is the same as standard test20?

eulerscheZahl: yes

eulerscheZahl: http://www.js-games.de/eng/highscores/samegame/lx

dbdr: you beat me in 4 cases

eulerscheZahl: you beat me in 16

eulerscheZahl: or at least tie

kovi: :)

eulerscheZahl: http://www.js-games.de/eng/faq these FAQ are nostalgic

eulerscheZahl: When I told LX about my idea he called me crazy. "How do you want to do all the clicking stuff?" I knew it could be possible somehow but I had no concrete concept at the time.

dbdr: what does LX mean?

dbdr: oh, someone

eulerscheZahl: this one http://www.js-games.de/eng/team/lx

Marchete: that page feels so 2002

eulerscheZahl: looks a little bit like the D troll

eulerscheZahl: you can claim a score under your username. it will send you password via HTTP without encryption

eulerscheZahl: we should test for SQL injection and XSS

dbdr: based on a FAQ, it might just work

eulerscheZahl: but the guestbook has a google recaptcha :o

MSmits: seems excessive

MSmits: who tried to DDOS the guestbook?

eulerscheZahl: ddos is lame

MSmits: sure

dbdr: r (n 0 s 0 v 2 l 0 oa 0 ob 0 oc 2 od 0 oe 0 of 0 og 0 oh 0 c 1)

dbdr: looks like CG replay

eulerscheZahl: where did you get that?

dbdr: <meta http-equiv="pics-Label" content='(pics-1.1 "http://www.icra.org/pics/vocabularyv03/" l gen true for "http://js-games.de" r (n 0 s 0 v 2 l 0 oa 0 ob 0 oc 2 od 0 oe 0 of 0 og 0 oh 0 c 1) gen true for "http://www.js-games.de" r (n 0 s 0 v 2 l 0 oa 0 ob 0 oc 2 od 0 oe 0 of 0 og 0 oh 0 c 1))'>

dbdr: html of homepage

MSmits: eulerscheZahl thats just how dbdr thinks, usually he translates it before typing, but thats raw dbdr neuron communication

dbdr: syntax looks lispy

MSmits: mmh first results on 1% epsilon came back, nothing spectacular yet

MSmits: (this is an hour per run)

dbdr: new optims should be only allowed during winter

MSmits: agreed

dbdr: and ban the southern hemisphere to avoid that issue

MSmits: on the northern hemisphere ofc

eulerscheZahl: getting warm in here too

MSmits: lol

MSmits: I've been battling mosquitos to keep the windows open here

dbdr: watch out mosquitos sucked in by the fan and grilled on the cpu

MSmits: this happens?

dbdr: I guess that works as antimosquito

dbdr: might need those fancy case lights

MSmits: yeah lol

Westicles: I'm very impressed my new 18c desktop isn't getting hot

dbdr: you know the origin of the word "bug"?

MSmits: if it's not getting hot, you still have a crappy computer Westicles :P

eulerscheZahl: you no longer have that old laptop Westicles? :o

MSmits: dbdr maybe it comes from bugs?

MSmits: just riffing here

Westicles: Nah, I couldn't resist when dazzled by all those cores.

dbdr: frying on early computers and thus causing "bugs", yes

MSmits: wow really

eulerscheZahl: now dbdr officially has the worse hardware

eulerscheZahl: together with JBM

dbdr: yay \o/

Default avatar.png JBM: unexpected hl

MSmits: you dont need good hardware to rickroll people

eulerscheZahl: hi

Default avatar.png JBM: can't wait for the dbdr streams

dbdr: because of the hardware?

eulerscheZahl: i saw a video of him at least

eulerscheZahl: speaking at a conference

MSmits: dbdr?

eulerscheZahl: y

MSmits: cool

MSmits: I assume

Default avatar.png JBM: oooh now we're curious

dbdr: dbdrconf

Default avatar.png JBM: doctoring satabases again?

Default avatar.png JBM: satabase, that's a nice typo

dbdr: nice one

MSmits: starbases?

eulerscheZahl: something about a software for molucules, i don't remember

Default avatar.png JBM: the successor to scsibase

eulerscheZahl: he had a molecule as a profile pic initially

Default avatar.png JBM: i don't remember that far

MSmits: which one

Default avatar.png JBM: ethanol, most likely

MSmits: mmh ok

dbdr: https://i.pinimg.com/originals/37/df/da/37dfda943a0bff11fd79421f52a84e9f.png

MSmits: dbdrunk

eulerscheZahl: no idea what that is

dbdr: good memory about the profile, eulerscheZahl

MSmits: meth obviously

dbdr: I was a nobody at the time

eulerscheZahl: you were the java UTTT legend

dbdr: still with that pic?

eulerscheZahl: i think so

dbdr: possible

MSmits: dbdr inspired me to stick with C# until legend

eulerscheZahl: google image search suggests: Eminem Darkness

MSmits: so, meth

dbdr: Jul 7 2018 prince.png

eulerscheZahl: long ago

eulerscheZahl: i'm still not 100% used to it. new profile pics always confuse me

MSmits: thats when i was working on uttt the first time

dbdr: common, I registered end of 2017

MSmits: so matches up

eulerscheZahl: that's half a year as a molecule

eulerscheZahl: long time

dbdr: changed a few months later. now 2 years later you're not used to the new one? :D

eulerscheZahl: nope

eulerscheZahl: smits is more consistent

dbdr: https://i.snipboard.io/5znuJt.jpg

eulerscheZahl: and that toad was my first avatar too (i admit i had a few others temporarily)

dbdr: then https://i.snipboard.io/io1uXr.jpg

eulerscheZahl: we had a golf cooking contest?

dbdr: cooking was calm

dbdr: then I focussed on golfing

dbdr: but I had to stop that trend, was getting crowded

MSmits: coulda added some floating numbers for NS

eulerscheZahl: and now you are back to running through long corridors

MSmits: and jumping, dont forget the jumping

eulerscheZahl: and stabbing

eulerscheZahl: that reminds me of the stabbing game we once played

dbdr: https://i.snipboard.io/tm1Izi.jpg

eulerscheZahl: http://stabby.io/

dbdr: how does it work?

kovi: yay, got you illedan

MSmits: kovi_not_hardcode stabbed Illedan?

dbdr: yay!

MSmits: :evil:

eulerscheZahl: damn, had no stamina left

dbdr: killed eulerscheZahl :D

MSmits: glorious

eulerscheZahl: (╯°□°)╯︵ ┻━┻

Illedan: :skull:

Illedan: Try again kovi )

Illedan: ;;)

kovi: nooo

Illedan: :P

dbdr: it's a bit silly

eulerscheZahl: i got bored too :D

dbdr: the only way to win is not to stab first :D

eulerscheZahl: exactly

dbdr: very moral

kovi: westicles, have you tried your code online?

dbdr: Illedan is creepy and sneaky, we already knew that ;)

Illedan: Was hiding :D

Default avatar.png JBM: exactly how the face-off with mirror prince should be played

dbdr: ^

eulerscheZahl: are you still hiding Illedan?

dbdr: I think he has 180k+

Default avatar.png JBM: if i don't stream fp tonight, i'll speedrun pop

Illedan: Going to be glorious

dbdr: porque no los dos?

Default avatar.png JBM: because pop doesn't qualify as fp?

dbdr: doesn't prevent you from doing both

Westicles: kovi, no it would be too slow.

Default avatar.png JBM: true

dbdr: bonus points for doing them at the same time

Default avatar.png JBM: lol

Default avatar.png JBM: well a pop-bot oughtn't be that hard

Default avatar.png JBM: in the grand scheme of gameplaying bots

dbdr: hard coded or learning?

kovi: what is the best for test20?

eulerscheZahl: did you see the link i shared after you asked about test 20?

dbdr: ‏‎5265

eulerscheZahl: http://www.js-games.de/eng/highscores/samegame/lx


dbdr: it's not always the best. but in this case it is

kovi: oh, i c and yours?

eulerscheZahl: there seems to be an unpublished paper having higher scores

eulerscheZahl: btw sampson is first here on CG

eulerscheZahl: mine:

eulerscheZahl: http://chat.codingame.com/pastebin/0e827e4e-9447-4d9d-9a40-030b43f826ce

eulerscheZahl: + means i cleared the board

dbdr: it's published, MSmits linked to it

eulerscheZahl: oh

MSmits: hmm not exactly

MSmits: the paper i linked is published

Westicles: oh really, the new one from Tristan? Does it give scores?

MSmits: but it has references to other research that is unverified

MSmits: not sure if it is published

dbdr: so there is a published paper with scores from possibly unpublished papers :)

kovi: interesting. i have tried locally and got to 4700 on 1st run (3 minutes)

kovi: must be lucky

dbdr: or diminishing returns

MSmits: which testcase kovi_not_hardcode?

kovi: case20

MSmits: thats higher than mine

MSmits: 4503

Illedan: 3887 :D

MSmits: max from sources is 5265

dbdr: js.de has that

MSmits: then that might be the source

Westicles: Oh, that one. Pah

kovi: nvm. was a lucky run

wlesavo: easy 2107 :smiley:

kovi: wow, 4900 now

eulerscheZahl: you should start hardcoding

kovi: what is yours westicles?

Westicles: yeah, 16-20 are the best to make you feel good. 65 of one color

Westicles: 5239 for 20

kovi: true. others are more "pure luck"

Westicles: 1-10 are random, 11-15 45 each color

kovi: well i cant tell the range, just 1,5,10,15,20 :)

Westicles: 4900 is good, you should play it out on jsgames

Westicles: Oh, I see... you don't know your move list I guess

dbdr: do you have a scrpt for that Westicles?

kovi: you did with 5239?

kovi: you have the point, i didnt record :)

kovi: the movelist

Westicles: No script... remember, I was the guy who couldn't figure out the python script on NS

dbdr: :D

kovi: balanced games are not that fun i guess i could battle for 2nd place on hardcoding

dbdr: why not #1?

kovi: i can only compare a few score with eulers, and from that i interpolate

eulerscheZahl: you can try the other too kovi https://www.codingame.com/demo/847363c996b5db478bdbf6bf192c2535825e84

kovi: i mean, those few are enough to assess that im consistently 10% above

kovi: (or can be)

kovi: but yeah, that is just with repeated longer runs. with more time additional code/heuristics can be added

Marchete: no more improvement for hours, I think I'm done with it

eulerscheZahl: i did a massive improvement on testcase 1

eulerscheZahl: took 2h to find it

aCat: so close 4, 5, 6

eulerscheZahl: http://js-games.de/eng/highscores/samegame/lx got my new solver working, i'll beat you now :P

aCat: what time is the contest end?

aCat: or you don't care and will improve your algos untill the world's end :joy_cat:

**eulerscheZahl doesn't care

aCat: :heart_eyes_cat::kissing_cat:

Westicles: ooooh, euler that is impressive

aCat: congratz Westicles :-)

Westicles: thanks, but I think euler is coming after me

aCat: BTW will you guys post some postmortems?

eulerscheZahl: maybe later

aCat: ok

aCat: or at least if you don;t want to get fully public send me a private note about your approach

aCat: for the science reasons :smirk_cat:

Westicles: is there a chat history somewhere? I sent it earlier but it scrolled off

eulerscheZahl: https://cg.spdns.eu/wiki/Main_Page

eulerscheZahl: dbdr's wiki

dbdr: wow, eulerscheZahl, amazing score, gz! and the top ones are so close to each other

dbdr: seems that this might be close to optimal

kovi: gratz

eulerscheZahl: curious to see how it runs on the other testcases

dbdr: :scream:

aCat: :scream_cat:

wlesavo: :scream: gratz

MSmits: gj eulerscheZahl

Default avatar.png Tuntraek: Hi, I'm having trouble with the native editor, ctrl+c, ctrl+v, etc are not working and Idk why, anyone got an idea why?

eulerscheZahl: VIM mode?

eulerscheZahl: left side: settings what's your edit mode?

Default avatar.png Tuntraek: ok, lemme see that..

Default avatar.png Tuntraek: yeah, it was emacs

eulerscheZahl: then i was close at least :D

Default avatar.png Tuntraek: yeap, that fix it... lol thanks!

Default avatar.png Tuntraek: btw, what are the copy, cut, paste shortcuts in emacs or vim?

eulerscheZahl: that's a google question if no one else steps in. no idea

dbdr: vim: v to select, y to copy, p to paste

eulerscheZahl: select what? all?

dbdr: shift-v to select lines, ctrl-v to select rectangles

dbdr: v to movement

dbdr: v + movement

Default avatar.png Tuntraek: thanks!

dbdr: vw selects a word, for instance :)

eulerscheZahl: gg vG is select all then?

Counterbalance: until EOF

dbdr: :+1:

dbdr: ^ future vim pro there

Counterbalance: mah, i've been vimming vor 24 years and still know nothing

eulerscheZahl: i'm too old for this now :(

eulerscheZahl: almost 29

dbdr: vi" select inside " quotes

dbdr: Counterbalance LD

Counterbalance: delete line in the bottom of the screen? :)

dbdr: zb puts the current line at the bottom, then it's easy ;)

dbdr: does that count? ;)

Default avatar.png LIONSxLEADER: lol is bashar one of the bots?

Default avatar.png scorpio3.0: gg

Default avatar.png scorpio3.0: gg

aCat: work in progress :-)

aCat: https://imgur.com/ucQrEHD

aCat: ah - there is mistake in showing mistake !

aCat: :joy_cat:

IfIHadATail: when does the Puzzle of the Week renew?

Scarfield: once a month

Scarfield: usually wednesday i think

IfIHadATail: seems legit

IfIHadATail: lol okay thanks : )

Scarfield: np :)

jacek: a month, weekfield?

Andriamanitra: why must reverse mode exist

Andriamanitra: half of these are just "guess what operation i decided to do with ascii values of characters this time"

Scarfield: yescek, was quality joke

kovi: yay

wlesavo: nice kovi

kovi: and pretty stably above 125k now

Scarfield: :muscle:

Andriamanitra: is there any enforcement against cheating in clash?

Andriamanitra: some guy submitted 100% in 25 seconds and this was a clash that i think most of us couldn't complete even in the full 15 minutes with all the edge cases

Andriamanitra: https://www.codingame.com/clashofcode/clash/report/13401038fea405615af59b5ab49062aead56148

Andriamanitra: incredibly frustrating problem too this oracle thing

reCurse: No rule against copy/pasting your previous attempt

Default avatar.png JaggBow: Can you get the same problem twice in CoC?

reCurse: Yes

kovi: lovely speedups

Default avatar.png Sintex: not proud of that one, but it passed the tests lol

MSmits: Andriamanitra it's not really cheating

MSmits: ah chat scroll

Andriamanitra: if that's not cheating i have no idea what is

Andriamanitra: i guess hacking the whole system might count, or maybe there's also technically no rule against that :grinning:

darkhorse64: TBali is a human bot

Andriamanitra: even if you've solved it before there's no way you dig up the answer manually in under 25 seconds so you can drop the human part

darkhorse64: I am a noob in CoC but I have already solved one in less of 25 seconds

Andriamanitra: i have too

Andriamanitra: but that's not one you solve in under a minute

darkhorse64: *less than*

TBali: Hi, yes that was a copy paste of my previous solution. First attempt was of course much longer for this one, but well within the 15 min limit.

TBali: I did not know if it is against the rules.

TBali: I am currently go for the 500 CoC achievement so Coc-ing a lot.

darkhorse64: A good memory is a very powerful weapon

TBali: I hate to solve same things twice, so if it repeats I copy my old solution

Andriamanitra: goes a bit against the spirit of the game in my opinion

Andriamanitra: when i get repeats i might solve it in a different language or something

TBali: After each CoC I save my solution to a big fille together with the first input/output pair

darkhorse64: :popcorn:

TBali: Okay I understand your point, and I don't wnat to ruin anybody's experience.

TBali: However if I meet a CoC I have the solution what shall I do? Wait artificially before submit?

Scarfield: new rule: Must wait 10min before copy pasting :p

Scarfield: what happens if you leave a CoC? just to avoid the grey area of cheating

darkhorse64: I played one CoC against you TBali: you output before i finished reading the statement

darkhorse64: :sob:

MSmits: well at least it wasnt your slow typing

MSmits: :p

darkhorse64: slow mind

TBali: Sometimes I met with some guys from the top of leaderboard. They are indeed fast...

TBali: plus they are also short (mostly in Ruby) . In PHP I am getting better and better in golf, but ruby still beats me most of the time.

Default avatar.png peppece: how do some ppl type so fast

Andriamanitra: i've played against the top3 multiple times and while they're fast it never feels unfair because they still take the time to write the code (even though i'm sure they've seen most clashes before)

darkhorse64: They press the right keys

Scarfield: with more than 2 fingers :p'

Default avatar.png peppece: do clashes repeat often?

TBali: it depends how much you play :-)

TBali: There is a limited number CoC puzzles in the pool.

Default avatar.png peppece: oh k

TBali: I am at 350 and now it repeats a LOT

Andriamanitra: i have a bit over 500 clashes and i've not had too many repeats yet, maybe a few dozen

TBali: that's strange

BenjaminUrquhart: I'm at 1238, I get a fair number of repeats

Default avatar.png peppece: one of u must be wrong lol

BenjaminUrquhart: not necessarily

Default avatar.png peppece: how so?

Kuzey: mods someone copied code from the internet on clash

BenjaminUrquhart: the clashes you get depend on your opponents' ranking in CoC and a bit of randomness

Default avatar.png peppece: ah i get it

BenjaminUrquhart: if you keep getting newer players in your lobbies you'll see a lot of the easier ones

Default avatar.png peppece: right so if u get stronger peopl everytime u keep seeing new ones

BenjaminUrquhart: yes, the "harder" ones

Default avatar.png peppece: kk

Andriamanitra: i guess tbali is getting only the hard ones since he's high level

Scarfield: hmm, i get some failed tests when submitting Same Game, but cant seem to fail a testcase?

BenjaminUrquhart: oh if we're talking about level, I'm at 238 for clashes

Default avatar.png peppece: just out of curiosity

Default avatar.png peppece: do people really get a job out of this site

TBali: Not sure about hard ones, I get one-liners as well

BenjaminUrquhart: ah yes the ones you solve in under 10 seconds

TBali: the one you quoted was one of the longest solutions I had in CoC.

TBali: pretty annoying edges cases, i remember

BenjaminUrquhart: which one was quoted?

TBali: scroll up to a link

BenjaminUrquhart: it's out of my history

BenjaminUrquhart: I opened an IDE window

Scarfield: https://www.codingame.com/clashofcode/clash/report/13401038fea405615af59b5ab49062aead56148 incredibly frustrating problem too this oracle thing

Scarfield: the link ^

BenjaminUrquhart: that doesn't tell me what the prompt is

BenjaminUrquhart: and I can't see solutions since I wasn't a participant

Scarfield: well, i scrolled up to a link

BenjaminUrquhart: this? https://www.codingame.com/contribute/view/4982f173dce367ca6a5cbc810907bb085189

BenjaminUrquhart: only result for oracle in clash

Allis: That's the one.

Allis: Hilarious that TBali submitted in 21 seconds, but I'm glad he's honest about the fact that he pastes previous solutions.

BenjaminUrquhart: I also do that for annoying ones

BenjaminUrquhart: :upside_down:

Allis: I think it's poor sportsmanship, but obviously I understand not wanting to solve it again.

BenjaminUrquhart: this one seems fairly simple to me

BenjaminUrquhart: except I can't test it for some reason

Allis: Pretty much all of the ones that don't require some physics knowledge are "fairly simple". Speed is usually the name of the game.

Andriamanitra: it doesn't seem too bad until you look at the test cases

Allis: And if you're pasting a previous solution, well, you're kinda starting the race at the finish line?

BenjaminUrquhart: the point is to have fun, are you really having fun by recoding something you didn't enjoy the first time around?

BenjaminUrquhart: but I see what you're saying

BenjaminUrquhart: "Couldn't find the question related to this contribution".

BenjaminUrquhart: no u

BenjaminUrquhart: it's just a bunch of if-statements looking at the validators

Allis: There's a way to be clever about it that doesn't involve much special-casing.

BenjaminUrquhart: the sample solution is simple at least

TBali: I think the shortest I got was: echo (($n * $n - $n) / 2) . "\n"

Allis: On that note, it seems like it'd be a nice feature to allow multiple solution languages.

TBali: or something like that

BenjaminUrquhart: :thinking:

Andriamanitra: it's almost like the author knew of all the traps he deliberately planted

BenjaminUrquhart: ok give me an example

Andriamanitra: like spaces before "?", abbreviated words that end in ".", question words that have 's after them

Andriamanitra: .split is rendered useless

Allis: Thankfully, Ruby copied a whole bunch of Perl's regex craziness.

BenjaminUrquhart: .split("('s)? ")

BenjaminUrquhart: (I use a lang with regex splitting)

Allis: As do I, but I'm not sure I see how it helps?

Allis: Splitting at all is definitely a trap for that one.

BenjaminUrquhart: removes the 's from words that end with it since it's a conditional part of the split regex

Andriamanitra: it helps for the question mark thing but then you need to be careful not to use the same thing for finding out if the last letter is a vowel

Andriamanitra: question word thing *

BenjaminUrquhart: the question mark is easy, not sure why the spaces are a trap

BenjaminUrquhart: last char of string == '?'

MSmits: It's a trap!

BenjaminUrquhart: hi are you still doing yavalath

MSmits: not currently, why do you ask?

BenjaminUrquhart: curious

MSmits: not much to do there atm

Andriamanitra: because you can't count the number of words ending in consonant by doing words.size - vocal_word_count

MSmits: If i get opposition, I will go back

BenjaminUrquhart: 4 points ahead I see

BenjaminUrquhart: understandable have a nice day

MSmits: you too , going to bed soon btw, not much day left. Also, work tomorrow :(

MSmits: been 6 weeks off, always sucks the first few days

BenjaminUrquhart: Andriamanitra that's not really a hurdle though? It's one of those things you try, see it doesn't work, and move on to counting consonants the long way

BenjaminUrquhart: I haven't done anything on the site since pacman challenge

Andriamanitra: well it's a hurdle you can manage but hurdle nonetheless

BenjaminUrquhart: fair enough

BenjaminUrquhart: gonna play a few rounds to see if I'm completely wrong and the puzzle quality has changed a lot since I last came

BenjaminUrquhart: oh hi

Andriamanitra: i feel like i've had terrible puzzles the last couple days

Andriamanitra: but maybe i'm just imagining it

BenjaminUrquhart: well there's that CoC footer now

BenjaminUrquhart: annoying smh

reCurse: I don't think anyone takes clashes seriously enough to care about enforcing rules especially when it's something you solved before