CyberLemonade: wow someone had a nice time
LelouchVC2: Hey, Jeff Cavalier, AthleanX.com
rkhapov: gays, do u now flags of C++ compiler?
jacek: yes, we do. do you?
rkhapov: i dont
R8T3D: > when people don't submit their solutions :(
R8T3D: on clash
[4.0]poorgoomba: i submitted
R8T3D: THANK YOU!
R8T3D: Wow I don't understand that but thanks regardless
R8T3D: Much appreciated
jacek: #pragma GCC optimize "Ofast,unroll-loops,omit-frame-pointer,inline"
- pragma GCC option("arch=native", "tune=native", "no-zero-upper")
- pragma GCC target("rdrnd", "popcnt", "avx", "bmi2")
Qoter: AntiSquid why did you ban rkhapov?
jacek: try #include <map>
YurkovAS: jacek in uttt or other mcts multi do you use mcts-solver? Is solver has winrate 65% vs standart mcts?
jacek: i always use solver. for me this is the "standard" mcts
jacek: winrate depends on game, but i think i in uttt i got something like 55%
jacek: in yavalath is a must-have. even if mcts solver doesnt help much, it doesnt worsen mcts either
karliso: How do you go about implementing mcts solver?
karliso: Do you see a node that ends, and then check if previous ends as well?
jacek: i have bool terminal in a node
karliso: So you check if all children terminal?
jacek: if i encounter terminal state, if this is a win for player, the its a lose and terminal for a parent
jacek: if this is a lose, i check all siblings if they are lose. if yes, parent is a win, and grandparent is a lose again
YurkovAS: jacek thanks
jacek: whats with the sudden interest in uttt in #ru :thinking:
darkhorse64: Don't tell me that all these people get so high in UTTT without a solver :scream:
YurkovAS: +5 new users in uttt legend from #ru in last one week. they found all tricks
darkhorse64: Except for teccles heuristic, there are not so many tricks in UTTT
CyberLemonade: darkhorse64 what's that?
darkhorse64: teccles heuristic ? When you play for the first time in an empty miniboard, send your opponent to the same board. I haven't tried it, so I don't know if it applies on rollouts and/or move selection
CyberLemonade: I see, thanks
YurkovAS: darkhorse64 solver is hard to understand. do you have good example? not pseudo code from "Monte-Carlo Tree Search Solver.pdf"
tutubalin: jacek when there's no challenge, we help new guys in #ru with CSB or puzzles. If there are no new guys, we discuss UTTT.
Astrobytes: lol, we only discuss UTTT when re Curse isn't here :D
jacek: pseudo code from that paper is kind of mess, like theyre mixing negamax propagation with normal
Astrobytes: It's horrible
darkhorse64: Score Bounded Monte-Carlo Tree Search is an improved version. I implemented from this paper.
darkhorse64: Yeah, it's horrible so you have to read carefully between the lines to realize it's plain simple
YurkovAS: darkhorse64 thanks
Astrobytes: Yep. I read the paper, then I saw the pseudocode and started questioning whether I had understood it. Had a conversation with Smits to make sure I wasn't going crazy
darkhorse64: Several pages paper for something that can be explained in two sentences
jacek: hopefully the results there were legit
darkhorse64: Just tried google translate on #ru. It does a pretty good job. There's going to be lot of ne comers in UTTT top
darkhorse64: Once they stop banging their heads on the solver, I fear that
tutubalin: too late, we've already annexed legendary league
tutubalin: next major goal is to do that on Python
darkhorse64: It's one thing to reach legend, it's another to reach the top
JohnCM: @darkhorse64 that paper sounds interesting
darkhorse64: The boss is an MCTS which performs at 25K rollouts. Good luck !
JohnCM: can you post a link to the score bounded monte-carlo tree search paper
jacek: 25k in second turn?
darkhorse64: Yes. 20-25k second turn. That's what I heard on the chat
jacek: so what im doing wrong if i get 10k in 2nd turn?
Uljahn: using solver and 1ply check? :smiley:
jacek: well yes
darkhorse64: You have smart rollouts ?
Uljahn: i guess vanilla MCTS is just plain random rollouts without heuristics and tree reusing
jacek: theyre not smart. 1ply check is just about choosing winning move and avoiding losing one
jacek: i have teccles, solved, 1ply check and tree reuse. in rollouts some small bias towards moves that would win small board
tutubalin: jacek do you use bitmaps?
jacek: yet having several times slower bot i am at the top
darkhorse64: Yes. Tree reuse and solver do not reuire game knowledge. Moreover, they are "free". Playing winning moves and avoiding losing moves is also a nice improvement
jacek: i use bitboards
tutubalin: how do you check if position is terminal?
Uljahn: "i have bool terminal in a node"
tutubalin: i mea in random playout
jacek: if my main board is terminal. i have table of 512 possible states and just do lookup
jacek: well any board, not main board
jacek: for (move : availableMoves) -> makeMove() check() undoMove()
tomatoes: i tried to check if possible to win then do it, but no much difference
WINWINWIN: still cant get the minimax to work for Othello :sob:
WINWINWIN: worst part is that it looks so simple
darkhorse64: How do you use teccles: biasing the rollouts, biaising the selection, book move ?
jacek: just in the beginning. if both players use teccles, it will end up in 18th move
darkhorse64: As a book move then.
jacek: WINWINWIN you cant get the minimax part, or the game mechanics?
WINWINWIN: The minimax keeps returning None as the best move
WINWINWIN: But printing to stderr shows that the actions are being generated
darkhorse64: tomatoes: wow #2. Did you add an opening book ?
WINWINWIN: And the actions are accurate with the initial input
tomatoes: no, i have no idea how to store it
darkhorse64: So you climb this high without any hardcoding (except the first move, I guess) ? Any special trick ? Just say yes if you don't want to share. I'm ok with that
tomatoes: im selecting nodes by heuristic
tomatoes: no random and ucb with a bit patched formula
darkhorse64: You found one. It's a common and accepted knowledge that there is no heuristic for UTTT. On other board games, combining evaluation and MCTS yields a powerful bot
Astrobytes: Smits is gonna love that leaderboard when he sees it :D
darkhorse64: He claims he can get to #1. It may be harder than he thinks
Astrobytes: Indeed. He thought he was only up against karlis o
WINWINWIN: But its only 0.15 points back to #2?
WINWINWIN: Wouldnt a lucky resubmit put him back at #2?
Astrobytes: Oh right, gotcha lol
darkhorse64: Yes. It means nothing except that there is a new contender
tutubalin: jacek when do you create new node?
Astrobytes: Yep. And when he comes back with his new bot (I believe he's working on that now) then it'll be tuned for karlis o and not tomatoe s
WINWINWIN: In the top areas of legend league, do people add modifications to cater to the opponents>
jacek: smits use special opening books for them
Astrobytes: Guess he's gonna have another few months work :P
jacek: tutubalin when i expand visited once node
jacek: he also planned othello
Astrobytes: Oh nice
tutubalin: jacek so you don't create it on every random move, right?
tutubalin: one of my failed attempts was when i created a node for every played state
tutubalin: for example, during random playout i created a node for every played move
Astrobytes: Isn't that what struc t was doing? Creating nodes for all his playouts
tutubalin: that was slow and took too much memory
tutubalin: it would work fine if we had a lot of time and memory
jacek: i dont create any nodes during rollouts
tutubalin: but with limitations we have it's waste of resources: most of these nodes will never be visited again
tutubalin: hmmm.... ok...
tutubalin: do you turn optimizations on?
jacek: of course
tutubalin: do you use standard C++ random or custom one?
JohnCM: yea rollouts can be huge, especially if the state space is stochastic (affected by random factors)
tutubalin: hm... than i have no idea why you've got 10k rollouts on 2nd turn
tutubalin: probably your rollouts are much better than mine
Astrobytes: Are they heavy rollouts jacek?
jacek: not much heavy, no
jacek: or i alwats get lucky resubmits
tutubalin: have you tried to profile?
tutubalin: what's the most time consuming part?
jacek: long ago, most of the time in my games the rollouts are time consuming
tutubalin: my last idea: probably there's a bug in rollouts counter, and you have 10x times more of them )
jacek: shouldnt have used int16_t for that
tutubalin: ahahaha )
Astrobytes: Seriously? :D
dbdr: Rust #2 on uTTT :)
dbdr: gg tomatoes!
tomatoes: thanks :sweat_smile:
dbdr: have you been using rust for long?
darkhorse64: I'll wreck him in the IDE then
tomatoes: no, half year maybe
dbdr: what did you use before? hard transition?
tomatoes: python mostly, not hard, but i also dont use most rust features on cg
tomatoes: threads, async, lifetimes etc
dbdr: makes sense. multithreading is rarely useful on CG, regardless of language
dbdr: lifetimes can be, but most of the times there is a simpler solution without them
dbdr: if you don't use them coming from a GC language, it seems you get it :)
dbdr: technically you always use lifetimes in rust, it's just that the compiler infers them and checks them automatically for you :)
tomatoes: thanks rustc
dbdr: indeed :)
struct: yeah, that was what im doing on uttt Astrobytes
struct: But I had no idea that mcts did not work like that
Astrobytes: Easy enough mistake to make
struct: Also the expand takes up 40% of the performance
Astrobytes: True Expansion.
jacek: darn had 1 timeout against java coffee cup
struct: fix it
eulerscheZahl: please stop spamming the chat
eulerscheZahl: #your_room feel free to click that and chat there
eulerscheZahl: reCurse your turn
AntiSquid: hey guys what are you doing? @_@
AntiSquid: just one minute away from chat and this is what happens :/
Hjax: this is nothing compared to last night
Hjax: last night someone spent about half an hour flooding the chat
AntiSquid: i saw that when i entered chat
AntiSquid: it was form 2 AM my time so no way i'd be up :p
eulerscheZahl: i think we could expect more dedication from you. disappointing
AntiSquid: i don't get paid enough to do this
eulerscheZahl: dedication doesn't need payment
struct: Yeah, my chat experience was ruined yesterday :(
eulerscheZahl: one star because there is no way to give 0
dbdr: give 2^32 stars and count on overflow
eulerscheZahl: 2^31 and hope for am overflow to a negative rating
eulerscheZahl: yes. but if you don't trust whoever told you this, why should you trust us?
eulerscheZahl: https://en.wikipedia.org/wiki/NumPy "Written in Python, C" almighty wikipedia, the most trustworthy source of all
eulerscheZahl: python is turing complete. if you can do it with C/C++, you can do it with python too. but speed is a good motivation to still use C/C++
Uljahn: you can build modules even with Fortran
eulerscheZahl: you can even compile Fortran with some C compilers
eulerscheZahl: fortran is the first high level language
eulerscheZahl: interesting history. they even delayed it to add compiler optimizations
eulerscheZahl: not to "burn" the language. when coders decide that it's too slow, they have a negative attitude to an improved version already
eulerscheZahl: disclaimer: i never tried anything with C+Python
in general you limit yourself to primitive types such as int, double when you code some cross-language code
tomatoes: more like you need to write data to c
Uljahn: isn't it what happens when you call numpy functions
tutubalin: AntiSquid there was collateral damage. you banned rkhapov for asking about compiler flags
Hjax: its a combination of the performance of C and the trend that more experienced developers prefer C
Hjax: oh i wasnt scrolled
Hjax: good old chat
tutubalin: AggYz you cannot just inline C code inside Python. but you can create a C-module which then can be imported from Python code
tutubalin: Of course C code should be aware that it is used with Python and follow some conventions
AntiSquid: "boomer language" lol, is older than some boomer's parents
AntiSquid: fortran is
AntiSquid: automatonnn lives matter
AutomatonNN: i feel like i can't say anything about the contest
reCurse: Ohh is that a penis joke
AntiSquid: i am only D user so i am minority!!!
AntiSquid: im going to burn and loot your house if you discriminate me with your joke AggYz
tutubalin: this D minor joke was really good one
dbdr: I had missed it, thanks tutubalin, and gg reCurse
reCurse: Alright that's enough shitchatting
Hjax: cg chat is weird today
Snef: Our savior \o/
reCurse: Any more takers?
AntiSquid: no manners at all
reCurse: Dunno why chat to teenagers is like catnip to cats
jrke: What a chat !
jrke: Recurse i tried hard but didn't even completed 1 lap in CSB against you :neutral_face:
Hjax: watching the good csb bots play is so interesting
Hjax: i watched a game where re curse's pod bounced off an enemy pod, creating a perfect gap for his second pod to get through
reCurse: It's definitely not about racing anymore past a certain point
Hjax: its about being a massive bully
Hjax: and just blocking forever
reCurse: Yeah and cooperation
reCurse: Tooltip says moderator
AntiSquid: is it just NN or are there any conditions you added to your CSB bot?
reCurse: It basically just means I keep the chat clean as best I can
reCurse: AntiSquid: The only "just NN" bot is Fenrir afaik
AntiSquid: sqrt() FelixvanBeusekom, not sure what you mean
Hjax: #include <math.h> sqrt() ?
jrke: ya for c++ use #include <cmath>
jacek: and for D <dmath>
eulerscheZahl: and for java it's JMath?
jacek: oh boy
AntiSquid: Tamizh no clash link invites here
tutubalin: only UTTT discussions are allowed here
jrke: whats UTTT means ?
jacek: maybe reBless will explain
tutubalin: Ultimate Topic To Talk
darkhorse64: You are not a real CodinGamer if you have no UTTT bot
reCurse: Is that CGatekeeping?
struct: gg jacek
jacek: boring game, make new one
darkhorse64: This game needs a boss
dbdr: someone spoke too soon
darkhorse64: trictrac is a tough nut
jacek: hmm maybe i should skip eval at last moves, i think i can solve it at 12-13 plies before end
OiYouYeahYou: @HexaHonut Hello stranger!
struct: No time to make a boss :(
jacek: in the beginning i had 5 strike win with him but later i was losing
darkhorse64: I can make one for you but it will rank around 7-9
struct: Sure, I would also make contribution private so they cant download the code
darkhorse64: There is a link to the referee, that's enough
jacek: you want to add boss after its been approved?
jacek: what would happen to the bots now?
struct: bot will submit
struct: all above get promoted
struct: after it gets 100%
dbdr: will ppl get demoted from woo1 to wood2 first?
struct: everyone starts at wood 2
struct: so yeah
dbdr: but people are already in wood1
struct: wood 1 becomes wood 2
dbdr: ok, cool
jacek: especially at morning
eulerscheZahl: that leads to a follow-up poll: who has a linkedIn account?
PatrickMcGinnisII: Next poll: who has a LN acct just to weed out fakes?
PatrickMcGinnisII: I have 0 contacts, but I am on LN... what's your point eulerscheZahl?
eulerscheZahl: my point is: i don't have "the CTO lurking" on my LinkedIn. But I don't have a LinkedIn account, so it's not surprising
eulerscheZahl: my social media account is this one on CG
jrke: ya same here no whatsapp,FB,etc just CG
dbf: "my social media account is this one on CG" - whats about russianaicup.ru?
eulerscheZahl: i don't use that one for chatting
eulerscheZahl: i know nothing about RAIC players who are not active on CG as well
eulerscheZahl: jeez, that storry PatrickMcGinnisII :o
eulerscheZahl: and the employer didn't even verify your accusation?
NASADeveloper102: Hi, I'm new to codingame. What's there to do here?
NASADeveloper102: Sorry if I'm interrupting a conversation
eulerscheZahl: if you are new to coding: https://www.codingame.com/training if you feel confident already: https://www.codingame.com/multiplayer/bot-programming
PatrickMcGinnisII: I don't know exactly what they did, but I'm sure they firedrilled his work computers and email, because he did lose the job.
if you have spended sometime in coding :smiley:
can anyone send me the solution of this code
jrke: no-one will send his solution make your own solution
NASADeveloper102: o _ o, wouldn't it be better to solve it yourself?
LelouchVC2: I'll send love
reCurse: Cheating won't get you anywhere
NASADeveloper102: The solution of the code is
< Patience >
PatrickMcGinnisII: Most americans are easy to find, LN is just 1 more way to gather info. Getting a name, addy,email,phone,DL # is easy... I'm wide open IDC, but don't mess w/ computer geeks. ;)
jrke: why everybody ask for solution :thinking:
NASADeveloper102: Maybe they just need more programming knowledge to solve the problems here.
NASADeveloper102: Which I guess would be the objective of this place, to push people to learn more programming?
NASADeveloper102: Or to get more cash :)
jrke: hey i wanna ask to recurse,Agade and all top rankers in CSB are they using smitsimax or anything else
reCurse: pb4 and Agade published a document saying what they do
PatrickMcGinnisII: Some of the more challenging puzzles really take a deep dive into math oriented algorithms, that i get...ascii-art ... c'mon ... learn to google or learn to code
eulerscheZahl: pb4git account
PatrickMcGinnisII: base85... interesting... I did a base88, it works for hardcoding big numbers
jrke: recurse do you have Phd(s) in computers
eulerscheZahl: agade has a PhD I think. But in Physics
Astrobytes: w lesavo too iirc, again in some branch of physics
Astrobytes: hi jrke
jrke: hey astrobytes
jrke: just going to sleep in 10 mins
jrke: no problem
NASADeveloper102: Should I really be doing coders-strike-back if I don't have a lot of experience with nueral networks? Well I guess the first question I would ask is, "does it require one"?
jrke: not till silver its easy till silver
NASADeveloper102: Easy seems like a very opinionated statement here.
eulerscheZahl: 100000 players and the arena. and top4 are neural networks
eulerscheZahl: still plenty of players without
NASADeveloper102: I see.
jrke: i just have simple simulation and in legend with 189 rank
eulerscheZahl: there are also minimax, monte carlo tree search, genetic algorithms, ... on the leaderboard
jrke: smitsimax also
NASADeveloper102: Simple still seems pretty complicated.
eulerscheZahl: personally i don't even think that CSB is a good game for getting started
NASADeveloper102: Perhaps it's simple to you. It may not necessarily be to me.
reCurse: Neural networks are no concern to you until you care about breaching top 4 legend
Hjax: pr2 turned out to be pretty straight forward and fun
jrke: but CSB have tutorial in starting i think so
Astrobytes: Just depends on your experience I guess whether CSB suits you or not
Astrobytes: Plenty other games to try out, that's for sure
eulerscheZahl: it took me a while to understand the turning angle and why i should slow down. just missed a proper description of physics
eulerscheZahl: "beginner friendly" :rolling_eyes:
jrke: but these things are in gold euler
Hjax: ghost in the cell is also pretty easy to get started with
NASADeveloper102: What is the incentive for getting legend?
Hjax: pride! xp!
Hjax: thats about it
Astrobytes: + bragging rights if you can get to the top
eulerscheZahl: occasionally there are contests on CodinGame. You can win tshirts and other goodies there. Multiplayer games are just for fun and learning
WINWINWIN: Astrobytes, have you ever encountered a continous None output in minimax?
NASADeveloper102: I see.
Astrobytes: No, because I don't use python WINWINWIN :P
jrke: hey winwinwin
Hjax: if minimax is returning none, you have a bug
eulerscheZahl: but we have to wait a few months till the next contest :(
Astrobytes: But sure, I've had no move returned when I had bugs
jrke: ya october or november
Hjax: i suppose theres no hope of another surprise contest like OOC?
jrke: next contest
Astrobytes: Wait for second wave of COVID Hjax
eulerscheZahl: a surprise contest would require an existing game
Hjax: i guess the unofficial contest was the surprise contest
eulerscheZahl: OOC was an old prototype, as no one had anything else
eulerscheZahl: went much better than expected
Hjax: i see
Astrobytes: Was a lockdown-special
eulerscheZahl: i think the next bot contest i'll play will be mini RAIC
eulerscheZahl: no date announced yet
dbdr: Hjax: step 1, start a new pandemic
eulerscheZahl: what soup do you want to try?
jacek: frog soup
Astrobytes: Toad-licking should do it
Hjax: maybe we can convince thibpat to do another unofficial contest in a month or two
Astrobytes: oh dammit
eulerscheZahl: simpsons is like xkcd
jrke: im going to sleep bye to all
Astrobytes: Definitely eulerscheZahl
Astrobytes: goodnight jrke, sleep well
jrke: gn astrobytes
USE-TO: We need python3.8 support
Astrobytes: hey jacek, good work in Othello, tric is struggling to get back on top (for now)
jacek: yeah he resubmitted and was 2nd
Astrobytes: I saw :)
trictrac: gg jacek yout bot is strong
jacek: so lesson today: keep your lucky resubmit
Astrobytes: It is very strong, what did you improve today? Or you don't wanna say
jacek: better move ordering, fixed killer heuristic so it wont overwrite best move from TT
jacek: as well adjusted weights for eval
Astrobytes: Nice. Was a very clear improvement, well done again
LelouchVC2: We need C++22 support
LelouchVC2: I want features that I don't yet know I want
Hjax: move gen in othello looks annoying
Hjax: because if a move is legal or not depends on the state of the rest of the board
Hjax: its not like chess, where you just look at one piece
Hjax: some chess engines dont even care about check making moves illegal, and just allow the king to be taken
Hjax: because forced loss is avoided anyway
Astrobytes: It's not too difficult, just some shifting and whatnot
jacek: you just need to copy paste clever bitboard tricks from internet and your good to go :v
Hjax: oh hey look, the chessprogrammingwiki has an othello page
Astrobytes: I think I need to look for some of those
Hjax: i found someone just precomputing a table of all legal moves for a given row
jacek: for testing eval i use random xot openings so i could see if eval is really better or just better against my old bot
Hjax: and then just rotating their bitboard
Hjax: well, keeping rotations of their bitboard
struct: If you think othello is annoying
struct: you should see yinsh
Hjax: i clicked on yinsh once
Hjax: i noped out of there real fast
Astrobytes: "i clicked on yinsh once" - meme-worthy
PatrickMcGinnisII: fair numbering puzzle took me a couple hours, but got it down to 18 lines
PatrickMcGinnisII: I'llc all today a success, now to mow the frickin' lawn
AntiSquid: mowing lawn is probably more enjoyable
Hjax: like, for continuous integration?
Hjax: if so, then its an automated process for building / running / testing the code you push to a repository
NASADeveloper102: I doubt this, but I'll ask anyway. Does using C++, or any compiled language decrease execution time in bot programming puzzles?
LelouchVC2: If you're asking because the puzzle says ur solution is taking too long
LelouchVC2: Then no I'd say. It should be referring more to time complexities
NASADeveloper102: Okay that makes sense.
NASADeveloper102: I haven't gotten that, but I figured that was a thing.
LelouchVC2: Try to make your program do less work - make it work smarter not ahrder
struct: NASADeveloper102 you mean on multiplayers?
Uljahn: do you mean time limits?
NASADeveloper102: As I said, I haven't gotten an error regarding the time it takes to execute code.
NASADeveloper102: And I don't mean time limits.
struct: Which multi?
NASADeveloper102: Any of them
NASADeveloper102: Will the answer change per each one?
NASADeveloper102: Okay then
struct: Just wanted to know if it was multies
struct: so compilation time should not matter
struct: what matters is execution time
struct: Which is the same for all languages
NASADeveloper102: I know the compilation time shouldn't matter.
NASADeveloper102: But thank you for that latter piece of information.
Hjax: you are given the same amount of time, but obviously a faster language will get more done in that time
struct: This doesnt apply to puzzles though
struct: at least for some
Hjax: theres a reason you dont see any python bots at the top of uttt
NASADeveloper102: I see.
struct: Rust has release mode
struct: C# does too
struct: C++ has pragmas
NASADeveloper102: Release mode?
Hjax: when you run a language like rust in the IDE, its compiled in "debug" mode, which is slower but gives you more helpful errors
Hjax: when you hit submit, its compiled in "release" mode, which is much faster
NASADeveloper102: I see.
NASADeveloper102: Do you recommend a bot programming competition I should try first?
MPSI_LIV_CronierDamiano: hey guys
MPSI_LIV_CronierDamiano: is there someone who know ocaml ?
struct: Which type of games do you like nasa?
NASADeveloper102: What type of games do I like?
NASADeveloper102: I like RPGs like pokemon.
Hjax: well a beginner should probably play a game that isnt too hard to wrap your head around, im currently looking at Ghost in the Cell, which looks pretty straight forward
NASADeveloper102: Thanks for the recommendation, I'll look into it.
Hjax: i also recently played Platinum Rift 2, that was pretty fun and straight forward as well
NASADeveloper102: Okay, I'll probably write these down.
NASADeveloper102: These all have leagues right?
Hjax: platinum rift 2 does not
Hjax: it just has a normal leaderboard
Hjax: still an enjoyable game
struct: Fantastic bits
struct: best game
NASADeveloper102: Okay, cool.
Uljahn: yinsh is better :smirk:
Hjax: i would not recommend yinsh to a beginner
struct: yinsh might be a bit hard for starting
NASADeveloper102: I see, okay.
NASADeveloper102: Do you mean beginner to programming, or beginner to puzzles?
NASADeveloper102: Or perhaps a mixture of both?
NASADeveloper102: I see.
struct: These modules are driving me crazy
struct: (╯°□°）╯︵ ┻━┻
NASADeveloper102: Also another question, why is coders strike back so, infamous here?
Hjax: because it has by far the most users on its leaderboard
Hjax: also the game is pretty difficult imo, but maybe im just bad at games like that
struct: Game is hard to get to top 50/100 maybe
struct: But I think it's a good to learn
NASADeveloper102: Okay, thanks!
Hjax: personally i find grid based games to be way more straight forward
struct: It has a guide on how to write a sim written by a cg user
NASADeveloper102: A "sim" meaning simulator?
Hjax: a lot of games on this site benefit from having a sim in your bot, to figure out the outcome of your actions
NASADeveloper102: I don't quite understand the benefits of a simulator in a bot.
Uljahn: predicting the future
NASADeveloper102: I see.
NASADeveloper102: So you take the game variables and you use those to predict the future.
NASADeveloper102: I assume?
Hjax: if i do this, and he does that? what will happen? a sim answers that question
struct: Then you evaluate
struct: The best state
Hjax: yeah then you ask yourself, i do this, and he does that. do i like the result? what if he does something else?
NASADeveloper102: Whenever creating "self-sustaining" programs I usually use a finite state machine, is that what you mean by "state"?
NASADeveloper102: I think that makes sense.
Hjax: you go through all of your actions, and all of your opponents actions
Hjax: and find the combinations that result in the best situation
Hjax: assuming your opponent plays optimally
NASADeveloper102: Wouldn't this result in a large execution time?
NASADeveloper102: I guess not if you write it well enough, or the game is simple enough.
LelouchVC2: HEADLINE: A bot that uses Random numbers to generate random moves destroys opponents
NASADeveloper102: 0 _ o
Hjax: if you are curious, look up the minimax algorithm
NASADeveloper102: Okay, sure.
Hjax: generally bots that use it run until they run out of time, and go with the best move they found
Hjax: but you have plenty of time to look quite a few moves into the future
Hjax: depending on the game
NASADeveloper102: Thanks, I never knew this.
NASADeveloper102: I really appreciate this.
struct: Also there is AVX :)
NASADeveloper102: Is that another algorithm?
NASADeveloper102: Is AVX a bot programming puzzle?
struct: Advanced vector extensions
NASADeveloper102: o _ 0
NASADeveloper102: What exactly does this mean?
struct: Imagine you have 8 floats
struct: and you want to add those 8 floats to another 8
struct: with avx those 8 floats take the same time as 1 float
NASADeveloper102: :thinking: Huh.
struct: This you can skip
struct: If you don't need it
NASADeveloper102: :thinking: Okay
Idleless: If you know what a float is, it's double the precision of a float (twice the size, same function). If you don't know what a float is, it's a real number (a number with a decimal place. example: 3.1234)
struct: more bits
Uljahn: Automaton2000: what are numbers?
Automaton2000: i need a new object
NASADeveloper102: This is a bot for codingame?
NASADeveloper102: Where does
Uljahn: ye, there are two of them
NASADeveloper102: Where does it get it's messages from
NASADeveloper102: also what is the second bot
NASADeveloper102: Huh neat.
AutomatonNN: eulerscheZahl is there a way to see the code and it works for me i was wrong about that in the code
Idleless: @AggYz https://hackr.io/blog/float-vs-double
NASADeveloper102: Also hasn't this been here for a long time? I mean the bot. So how does it choose the messages from such a large amount of them?
Hjax: it doesnt choose messages, it generates them, learning from chat
Uljahn: with Markov chains
NASADeveloper102: Wow I just suspected that it chose a random message based on one already spoken.
Automaton2000: some of the things i do in the next round
NASADeveloper102: What is the point of clash of code?
Hjax: depending on who you ask, there isnt one
NASADeveloper102: What circumstances would make you have to write code for an answer to a prompt quickly?
Hjax: job interviews
NASADeveloper102: And you would have to beat other people.
Uljahn: or bots
Hjax: not everything that is fun has to be useful
NASADeveloper102: I was thinking a coding competition
Hjax: if you think its fun, go play it
Hjax: ive played a bit of it, its alright
NASADeveloper102: Well I guess if it's useful in a job interview.
NASADeveloper102: I should try to spend at least an hour a day doing it.
struct: Clashes have downsides
NASADeveloper102: Because a social media life, is something I lack.
Hjax: honestly i like playing a multi has more benefits to a job interview than clash
Hjax: i feel like*
reCurse: Social media is the worst invention of the last decade, so no worries
NASADeveloper102: What benefits would outweigh the ones ones for clashes?
struct: in clash you will get bad coding habits
Hjax: code written quickly has the tendency to not be particularly good code
NASADeveloper102: But couldn't you say that you can learn to find a solution to a problem quickly and then write code for it?
NASADeveloper102: But I guess the same problem arrives.
NASADeveloper102: The quality of the code could potentially be lowered.
struct: I have 3695 clashes, and I don't recommend them
reCurse: It's not code you should be interested in but problem solving
reCurse: For clash there's definitely diminishing returns
struct: most clashes are math problems and parse strings anyways
NASADeveloper102: Is there a memory limit in multiplayer puzzles?
Hjax: yes, 768 mb
Uljahn: multis teach you staying focused on large tasks, thinking of architecture, performant data structures and optimizations i guess
NASADeveloper102: You probably can't include other libraries that aren't in the standards right?
Hjax: https://www.codingame.com/faq list of libraries by language here
Uljahn: check https://www.codingame.com/faq
NASADeveloper102: I mean like pandas?
Hjax: python does get pandas yes
NASADeveloper102: Do you think the PS5 is going to live up to the hype it's getting?
NASADeveloper102: Well that was surprising?
tomatoes: games > specs
NASADeveloper102: Huh, yeah you're right.
NASADeveloper102: I don't play console games, but I was thinking about getting an xbox or play station.
NASADeveloper102: So I can be introduced into the console gaming world with their accessories and money making schemes.
NASADeveloper102: And then I can feel the anger among gamers that have to spend money on seemingly meaningless things, that are required for some games.
NASADeveloper102: or features of them.
tomatoes: i have only psvr, used it ~2 times
NASADeveloper102: . . .
NASADeveloper102: How do you use something a negative amount of times?
NASADeveloper102: I assume, "psvr" is play station virtual reality?
tomatoes: around two times
NASADeveloper102: Wouldn't it make sense to put ~2 and not -2
tomatoes: as i did
NASADeveloper102: Wait a minute....
NASADeveloper102: That's tilda?
NASADeveloper102: What is this font?
tomatoes: tilda, yes
Reksio: Hi All. Does anyone know how Submitted application deal with Jit compilation? I know that code which first time runs for 1800 microseconds can run in just 15 microseconds if it runs 2nd time just after the first run. Is every submitted puzzle run twice at least?
Reksio: Or even every test case...
Reksio: Or is some kind of AoT (CoreRT) compilation used?
eulerscheZahl: your code is run once on every testcase. the speedup to account for JIT languages is to give 1s for the first turn. you have less time for later turns, whenever there is a loop of repeated input and output