Chat:World/2021-03-02
Wontonimo: hey JayM9, what's shaking?
jacek: good morning
micoze-stella: hi
micoze-stella: now 2:00 pm
jacek: oO
Wontonimo: good afternoon Japan ... it's just after midnight here
Wontonimo: weird that the sun is in different places in the sky for different locations ... as if it's a sphere or something
Wontonimo: AntiSquid what algo did you use in SpringChallenge?
Aiger-Akabane: Wontonimo i think AntiSquid is not here
micoze-stella: wow, So it's just March 2nd there?
Wontonimo: yeah micoze-stella , all day
Aiger-Akabane: i think Spring Challenge is near
jacek: 6th may, exactly
Wontonimo: looks like it's going to be a mash up game
Wontonimo: l8r
Uljahn: tutubalin: guess you could train to do AND (similar to XOR) and use binary representation with some common algo for multiplication
Uljahn: i mean neural network architectures aren't limited to fully connected ones, you could use memory nodes like in RNN/LSTM
Rypakka: Guys i suck at coding
Rypakka: Im really desperate to be better but somehow i can't
Rypakka: Any tips? I'm a beginner
KiwiTae: Rypakka easy take a deep breath and practice
KiwiTae: its not a sprint
Rypakka: I've already failed my first term in cs programming )):
KiwiTae: its ok to fail an exam
KiwiTae: just redo it on your own, see what u failed, and practice until u do it with no trouble
tutubalin: are you sure that programming is what you love? or you go there just because it's popular now?
Rypakka: At first I took it because I like challenges
KiwiTae: oh in codebusters u loose if your busters walk in range of the enemy base?
tutubalin: Uljahn that will give you binary representation of multiplication, not actual multiplication of 2 inputs
Uljahn: :confused:
derjack: nn more like approximates the 'representation' of function rather than function itself
[CG]Thibaud: no KiwiTae, I don't remember this
Butanium: http://chat.codingame.com/pastebin/7919e375-474f-47ef-8785-1c0543ddecde
Butanium: wtf is this coc
Butanium: reverse mode
Butanium: lol only the bot manage to do it
Uljahn: i guess any calculator uses taylor/maclaurin series to approximate math functions, so it doesn't calculate those functions itself
KiwiTae: [CG]Thibaud ohoh indeed looks like its me ~ must have a nullptr smwhere
KiwiTae: [CG]Thibaud merci o/
[CG]Thibaud: no pb ;)
MarTiaNe: I need help with puzzle 'Code4Life', when bot is at 'molecules module' it still prints last connection why? I have conditions for it debuging and everything seems fine
[CG]Thibaud: too many outputs during a turn so your code is desync with the input?
MarTiaNe: Yeah
[CG]Thibaud: ensure your print one command per turn, at the end of your logic. Not inside
MarTiaNe: :thinking:
Uljahn: make a variable to store your output command and print it at the end of the game loop instead of immediate printing inside if-else or other game logic
Uljahn: this way it's easier to debug and harder to fail
MarTiaNe: Yeah I see, I just don't get why it entered that condition again
Uljahn: just use debug output to print terms of the condition, it's better than guessing
MarTiaNe: Yeah I got it, I thought target was my position it was oponent's :grimacing:
tutubalin: my next CoC will be 1000th
tutubalin: anyone wanna celebrate it and challenge me?
tutubalin: oops. sorry. 1000th was the last one
MarTiaNe: :laughing:
Wontonimo: 1000? so, 5 to 10 min per class ... you could have made a lot of waffles and ate them in that time. Just saying
Wontonimo: well, tutubalin, I would have lost vs you
Kitkat260: gm
MSmits: mmh well thats creepy. VS is asking me if I want to enable AI assisted intellisense
Kitkat260: what?
MSmits: I think I'll just let the VS AI code my bots
darkhorse64: If it does not require to drill a hole in your head, why not ?
Kitkat260: no!!!!
MSmits: is that how you make all your decisions darkhorse64?
MSmits: check the drill hole in head heuristic?
darkhorse64: Integrity matters to me
MSmits: I see
MSmits: just installed new PC
MSmits: seems to be able to run 4 meta mcts without any issues
darkhorse64: We need to write more board games to keep your PC busy
MSmits: :)
jacek: 4 cores? what is it, 2020?
MSmits: no it has 10
MSmits: i want to leave some for actual use
MSmits: maybe i'll run 5 meta mcts, dunno. Guess I could start yavalath again
MSmits: the other computer had 4 cores, could run 2 meta mcts without issues, but it got really slow at 3
Wontonimo: what's a meta mcts? is it a hyperparameter search of settings for mcts?
jacek: its mcts but instead of using random playout, you use another mcts
jacek: it is mostly use for opening book generation
Wontonimo: :nerd:
Kitkat260: haha
Wontonimo: are the two mcts' identical otherwise?
jacek: yeah
Uljahn: i guess it's about evaluating openings with a very huge amount of games, building very deep tree and using solvers for endgames
jacek: though in meta since its done offline, youd want to exploit symmetries and transpositions also
Wontonimo: neato. nice way to also warm the office and vaporize excess coal
Wontonimo: i've gotta get more more cores and a proper gpu
Uljahn: yep, mining mcts-coins
Wontonimo: that reminds me of a puzzle to make about mining bitcoins
tutubalin: can you gamble in Monte Carlo with them?
Wontonimo: (not actual bitcoin, just the math concept)
MSmits: mmh some of my meta mcts crashing. Not enough object pool available. I guess because the cpu and memory are faster, it does more rollout and makes more nodes
MSmits: easy to fix i guess
Wontonimo: haha, issues of having more cpu
MSmits: ye
Wontonimo: that's a tough life MSmits ... I can help though
tutubalin: jacek what about metametamcts? how many levels of mcts is ok?
Wontonimo: just mail me your computer and I'll send you mine
MSmits: eh tutubalin, 2 levels is plenty :)
jacek: tutubalin some guy used nested mcts for some single player games like solitaire or samegame. 3 levels if i remember correctly
MSmits: yeah i suppose you can do that for SP
MSmits: that's different
jacek: if your cpu is fast but has low memory, use meta-meta-mcts. it will slow down giving more accurate simulations
Wontonimo: if at every level you recursively add MCTS to MCTS, then it just becomes regular MCTS again !
Wontonimo: wait ... no, let me think that through some more
oddoneout: how exactly do i advance from league to league in ai puzzles?
oddoneout: it says im ranking 1 above the boss but it doesnt advance
jacek: it needs to finish your submit. then you will see if you advanced and there will be timer
Wontonimo: ^
oddoneout: so do i just wait in the arena?
Wontonimo: no, you can go get a coffee
oddoneout: oh, cool, thanks
Wontonimo: but seriously, there's no way to make the timer go slower or faster. You don't need your browser open to advance
tutubalin: Wontonimo it becomes DFS, i suppose
Wontonimo: :D
dave573: anyone stuck at lvl8 boss?
Wontonimo: in what game?
jacek: lvl8 boss? w00t
oddoneout: wood league dave vs chad lvl8 boss
oddoneout: probably coders strike back? it feels like a 'default' to me
dave573: I went as a mage dps, but double for loops doesn't do as much damage as I would like
jacek: oO
Wontonimo: ôÔ
Wontonimo: i have no idea what you could be talking about dave573
jacek: :notebook: :soccer:
tutubalin: lvl8 boss is loop immune. try to memoize him instead.
dave573: What about whiles?
farmaceits: i got my arcane mage healer rdy to go
tutubalin: dave573 whie true continue
Kitkat260: sup
dick128: bubvhbddbuvvvdiffkvhkcuvh;obchbj
AntiSquid: that's a ban
AntiSquid: check username
AntiSquid: intent is clear
BlaiseEbuth: Yeah, I saw it after.
JohnathanBarbee: Bam! Get modded
BlaiseEbuth: Nvm. If he come back we will correct that.
HunterEhrenfeld: hello
HunterEhrenfeld: I will give a high 5 to anyone who can guess what exactly my pfp is and how it was made
jacek: whats pfp
HunterEhrenfeld: profile picture
Westicles: python turtle
HunterEhrenfeld: ヘ( ^o^)ノ\(^_^ )
jacek: is turtle a python
HunterEhrenfeld: python turtle is the devil lol
Kitkat260: gm
BlaiseEbuth: For the last time, I have nothing to do with python! :rage:
Kitkat260: how is everyone
JohnathanBarbee: Pretty good. You?
Kitkat260: i'm ok
jacek: how is Automaton2000
MadKnight: how's your everydays hi, Kitkat260 ?
Automaton2000: depends how you define it
jacek: how is AutomatonNN
AutomatonNN: hatput you to the best
Kitkat260: hi madknight
AntiSquid: hi MadKnight
MadKnight: hi Automaton2000
Automaton2000: there is a tool for that
AntiSquid: hell Kebab
MadKnight: a tool for saying hi? great! where can i get that Automaton2000 ?
Automaton2000: can you find the bug
Kaleab: I had an annoying problem (question)
Kitkat260: sup
MadKnight: dude
MadKnight: it's not even been an hour
MadKnight: Kaleab shoot it
MadKnight: what's your question ?
Kitkat260: nothing
HunterEhrenfeld: a precise question
jacek: AutomatoNN what's your question?
jacek: AutomatonNN even
AutomatonNN: eulerscheZahl is there a way to see a contest on the left side without it to start with easy games
jacek: :(
Wontonimo: oh AutomatonNN , you know not what you say
AutomatonNN: what do you mean?
struct: Wontonimo you asked something last night right?
struct: before saying later
struct: I cant remember what it was :/
Wontonimo: hey struct , i can't remember but i could come up with another question
struct: I dont remember if it was a question, I just remember that I had something to reply :(
Wontonimo: lol
Wontonimo: i was perhaps asking a poll of what kind of algo people used for Spring Challenge 2020 ... but I've since read the community posts
struct: Ah now I remember
struct: It was about the art of next contest
struct: Its just a placeholder
struct: You thought it was a mashup or something
Wontonimo: oh, yeah... it certainly looks that way
struct: its just a placeholder, it was also used on the fall challenge 2020
Wontonimo: that's not very cash money of them
jacek: they need to add ads to cocs
Wontonimo: don't give em ideas, they may look at how much compute each bot submit takes and make you watch an ad for ever 20 bot battles
Wontonimo: watch-to-win :scream:
jacek: oh nine men's morris soon to end contribution cycle. how did i miss that
GeorgiD: 2 131 -> 1 4 91 -> 3 5 156 -> 1
GeorgiD: Probably something obvious
GeorgiD: But couldnt notice the pattern
Westicles: b mod a
jacek: modulo
PatrickMcGinnisII: Westicles this 5D chests is kicking my arse
jacek: 5D chess?
PatrickMcGinnisII: Chests... ( . )( . )
jacek: :no_mouth:
Westicles: Timing out without symmetry?
PatrickMcGinnisII: yup, can't do any C>9 for N=20 even with precomputed primes
PatrickMcGinnisII: I don't know how to find the symmetry of a 5d space
ErrorRazor: is std::bitset slower than just using something like uint64_t?
ErrorRazor: I mean on codingame specifically. I heard something about not using stds
Westicles: Well, I can verify tbali and dwarfie got it working in PHP with some scheme involving hundreds of dollar signs
darkhorse64: Used wisely, the stl is a good thing. as for bitset, I hope that a bitset(64) is like a uint64_t but you cannot be sure unless you look at the implementation
PatrickMcGinnisII: yea, i see they did it... I'm guessing the sheer number rooms gets kinda exponential when c is large
JohnathanBarbee: @Error, I think stds are slower but easier to use. Just like Intel's FastRand and std::rand
darkhorse64: Use neither, they are bad, implement xorshift
darkhorse64: bad in terms of randomness
JohnathanBarbee: I will definitely look into that
darkhorse64: std is slow on CG if you don't use pragmas
PatrickMcGinnisII: I can see there are boundaries when N is not prime
ErrorRazor: I don't know what these do but I have #pragma GCC optimize "Ofast,omit-frame-pointer,inline,unroll-loops"
- pragma GCC option("arch=native", "tune=native", "no-zero-upper")
- pragma GCC target("rdrnd", "popcnt", "avx", "bmi2")
PatrickMcGinnisII: er N+1
darkhorse64: great
Westicles: All permutations of (x,y,z,w,v) are the same, so you can use that to do less flodds
Westicles: floods
PatrickMcGinnisII: ok if I find a room and search the 1 dimensional space... could I extrapolate .... hmmm, I'm under/over thinking it
PatrickMcGinnisII: I'm saving in a visited 5d manner, so I am searching too much
PatrickMcGinnisII: I guess if diagonals were allowed that would screw up the symetry
Westicles: It doesn't look like the other PHP solutions use symmetry. So just 5 deep loops and a flood fill should work
PatrickMcGinnisII: I'm not even using recursion, just a stack for floodfill
oddoneout: "Timeout: the program did not provide 1 input lines in due time... oddoneout will no longer be active in this game."
Is there any standard troubleshooting I can do to debug this? Or am I supposed to hit my head against a brick wall until it breaks? :D
oddoneout: Or am I mistaken that my bot takes too long to compute a game loop?
Branzz: Handicap for java code golf plz :grinning:
Branzz: It means you aren't reading in enough inputs
ErrorRazor: Either it takes too long or it doesn't output anything
therealbeef: or it crashes
tutubalin: try to output some debug info
oddoneout: nvm, I messed up so bad - i've put returns in my game loop instead of continues, haha
oddoneout: I am really much more used to having a few more files than just one code window
tutubalin: you're welcome! you can use rubber duck next time :)
oddoneout: hahaha, thanks
tutubalin: you can code in external IDE in multiple files and them glue them together with some script that runs on file save
tutubalin: that sync with CG
tutubalin: *that = then
oddoneout: I guess that would be handy but I have a feeling it would have a pretty low ROI
oddoneout: or I may be a bit too inexperienced to figure that out
tutubalin: i don't think so. i got some script from github, adjusted it a bit for my needs - et voila! i can code as i used to code
oddoneout: oh cool! I was 100% certain I would have to do it myself :sweat_smile:
murrayr: I'm struggling with "Ways to make change". I have a solution that works for all test cases but fails the last validator. The code is fast O(NS) and uses 64 bit ints. Any idea what could be special about that last validator?
PatrickMcGinnisII: murrayr it's only 2 loops...like 5 lines of code
murrayr: yes, that sounds about right
PatrickMcGinnisII: array has 1001 entries of only 32 bit ints
murrayr: seems to me a test case of 1000, {1..10} needs a 64 bit answer
Westicles: http://chat.codingame.com/pastebin/dd4f16f6-2c01-4095-9b59-6aa0c8d1b1c4
murrayr: ha ha ha - constraints say N<=1000, so I set my array exactly 2000 to be safe.... 2001 is necessary
tutubalin: why 2001?
murrayr: because the constraints lie. N<=2000
tutubalin: ah, ok.
tutubalin: i solved it on JS, so it definitely doesn't require 64 bit
Westicles: I updated the constraints to 2000
murrayr: I think none of the test or validator cases need 64bit - but they could have done. I can now confidently say N=1000, Vi={1,2,3,4,5,6,7,8,9,10} results in 968356321790171
murrayr: Thanks Westicles
tutubalin: that's why i love languages with dynamic arrays :)
tutubalin: my JS solution intentionally goes out of array boundaries
awnion: https://www.codingame.com/contribute/view/63428900a40dab092da96baba31fac53117d
awnion: need opinion plz: is this medium or easy?
Astrobytes: Since I didn't click on the link I'd say Very Hard.
Astrobytes: (sorry)
Astrobytes: OK, just on first look it does look Easy yeah
awnion: it's my contribution :) For me, it's not that hard but I'm biased as an author.
Astrobytes: I know, sorry, was joking around. It does seem to fit in the Easy category
awnion: OK, thanks.
Astrobytes: Oh I see that clashers thought it was too hard for clashes
awnion: yeah but for Clash it was 1M D and could be solved with BF on c++ :)
Astrobytes: Haha! Personally I think it's suitable for easy, wait for some more opinions but I think for a puzzle it should be fine
awnion: yeah I got 30 days anyways
Astrobytes: Will give it a closer look tomorrow
awnion: thank you really appreciate
Astrobytes: no worries. Anyway, I'm off for now, cya tomorrow
Branzz: :v_tone5:
tutubalin: awnion "The absolute values of the final result and all intermediate results are less than 2^60" wouldn't work well in JS
darkhorse64: thanks. It's a huge refactor so it may be successful
darkhorse64: *may not*
Tuo: u will proply pass me when someone else runs
darkhorse64: Unsure. If I run CGStats, it's stronger against you but weaker against others. I need to do some step by step to see if everything is right. There is also a large RPS effect
Tuo: you win in 16/20 latest battles
darkhorse64: Previous version running now
darkhorse64: This is really a strange game. Sometimes, I run against 3 default bots and I finish dead last. I am supposed to win all games against the boss !
cajunphoenix: NO YOU CANT SAY NIGGA
darkhorse64: last warning. stop that or get baaned
darkhorse64: banned
Tuo: yeah its very hard to test and find bugs when you cant even run the same seed reliably
darkhorse64: ? your bot is not deterministic ?
Tuo: its is but lets say you run a seed and finish last, then do a change that is obviusly bad and finnish 1st :sweat_smile:
darkhorse64: "finnish" :slight_smile:
Tuo: sometimes i try intentionally break my new features to see if they do anything and still first place
darkhorse64: I see what you mean. Looks like my previous version is doing worse
darkhorse64: This new version is not so bad. I'll let it run for tonight. gn
AntiSquid: hey tuo
AntiSquid: how often are you playing hanabi with friends?
AntiSquid: what
AntiSquid: you missed out
Tuo: first time i heard about that game was in explanation for fireworks
AntiSquid: you just need the 1 deck of cards, no lockdown restrictions and convince your friends to give it a try
AntiSquid: i thought you play it often because of your lead
AntiSquid: then again i've been too lazy to finish my bot
Tuo: maybe it would help in the beginning (is pretty rough fireworks) but in the long run its better to skip all the "human" strategies
AntiSquid: these aren't self-trained agents, so i expect everyone added their "human" approach to their bot :P
AntiSquid: tea-pot at the top :o
Tuo: yeah. pretty thin margin tho, is normal for 1st place to get +- 0.5 points when someone runs
Tuo: i mean someone else then top 2
Solarite-E: I don't think I've learnt anything so far, and i was already told to join a "Clash of Code" and i was so confused on where to start.
aegle: hi