Chat:World/2021-09-25

From CG community
Jump to navigation Jump to search

magus2: I think the 'X egg problem' is probably wrong from reading about the 2 egg problem on the internet.

Husoski: magus2 What's wrong about it?

magus2: I'll look into it further, but a site indicated that a 100 floor 2 egg problem would be 14, where the 'x egg problem' indicated that a 50 floor 2 egg problem would take 14. I'm sure that these two cases should have different results, so one of them is wrong.

magus2: nevermind, seems to be fine.

Default avatar.png KOOHOLATESEN: how can i run a puzzle game i joined this website recently and don't know anything

TimothyAlexisVass: https://www.codingame.com/training

Default avatar.png KOOHOLATESEN: thanks

Justin-Truong: hi

Justin-Truong: guys

Justin-Truong: https://www.codingame.com/clashofcode/clash/198179008ae77cd400a8825bb7f0d5fd52558af

Justin-Truong: join my clash?

Justin-Truong: pls

Justin-Truong: I want to finish the quest

Justin-Truong: https://www.codingame.com/clashofcode/clash/198179008ae77cd400a8825bb7f0d5fd52558af

Justin-Truong: hello

Justin-Truong: guys

Justin-Truong: https://www.codingame.com/clashofcode/clash/198179008ae77cd400a8825bb7f0d5fd52558af

Default avatar.png ash15khng: just join public clashes

Justin-Truong: no

Justin-Truong: I want to finish quest

BlaiseEbuth: No quest require to do private clashs.

BlaiseEbuth: ANd don't spam

Justin-Truong: quest require private right?

Justin-Truong: plz join me

Justin-Truong: https://www.codingame.com/clashofcode/clash/19818150c3a57bb5fa16de73e938d5e23300e8a

BlaiseEbuth: Stop

Uljahn: private clashes do not count i think

Justin-Truong: I count

Justin-Truong: it

Uljahn: for clash invites use #clash

Justin-Truong: ok thanks

Default avatar.png DimitrisLamprinos: hey guys, how do i enable vim mode in the editor

Uljahn: should be in the settings on the left side bar

AntiSquid: oh the usual clash spam, no convos, what's up Automaton2000 ?

Automaton2000: is there a way to use my own even if i have a list of available moves

ProCoder03: Hello everyone !!

jacek: haooy Caturday

el-musleh: Hi all, How I can check people answers after #ClashOfCode match for same question?

Default avatar.png ash15khng: click view code next to their name at the clash complete page

Default avatar.png mrrobot-caretaker: hi what's up?

Default avatar.png mrrobot-caretaker: how i can join html

Default avatar.png mrrobot-caretaker: clash

el-musleh: I can view other's code, but we are not solving the same game.

Default avatar.png M0H07ED: I think there should be enhancement in clash of code particularly in smallest code challenge which does not assume the different in syntax between programming languages (like there is a big difference between python and c)

Westicles: CoC is the first exposure for many young folks here to how unjust and frightening the world can be. Moderators are trained to help you through this crisis

Arison: yo where could i report a error in one of the practice exercise?

jacek: what error

69razer69: lol

Default avatar.png Mentos24: lele

Arison: already solved

armalitedev: ey

struct: hi

ErrorCookie: I want ot get the surrounding fields in a 2d array. I have working code in javascript for it but it'd be interesting to hear how you would go about that.

ErrorCookie: When I try to do this I end up with tons of if chains because of the corner and edge cells in an array

struct: is it similar to floodfill?

struct: I guess you can do something like this ErrorCookie

struct: http://chat.codingame.com/pastebin/b59f743d-28cd-4047-9d2b-01571acb74c2

ErrorCookie: I'm a bit confused what exactly does this code do

struct: it checks for each direction

struct: up, down, right, left

struct: its not the entire code

struct: just a part of it

ErrorCookie: how can I send pastebins like you did before?

struct: just paste it here

ErrorCookie: http://chat.codingame.com/pastebin/75180622-ddce-4b0c-b445-69b03204b03b

ErrorCookie: This is what I did when I did a game of life challenge

ErrorCookie: It performs an action on every 8 cells surrounding a cell

ErrorCookie: But with this I have 4 nested loops

ErrorCookie: I wouldnt call it performant

struct: if you have a 2d array with each direction possible

struct: like I have in my code

struct: but on my code is only 4 directions

ErrorCookie: Your code uses an array with array[4][2]

raizbro: hi

struct: yes 2d array

raizbro: that no good who kik me

ErrorCookie: but 4 2 has a width or height of 2

raizbro: oh

ErrorCookie: so you can either go left or right but not both

ErrorCookie: from the current cell

raizbro: i knoo the anwser

ErrorCookie: enlighten us .(^_^).

raizbro: no

ErrorCookie: :O

Astrobytes: you again

struct: 4 directions with x and y change

struct: so 1st direction is right

struct: because x increases by 1 and y doesnt increase

ErrorCookie: Yes I just understood dirs means possible directions I thought dirs was the array you perform the surrounding search on

struct: because its 0

ErrorCookie: whoops

struct: it allows you to check all the surrounding cells of the cell you are currently checking

ErrorCookie: Now I understand it it's pretty simple xD

ErrorCookie: Thank you!

struct: np

jacek: :tada:

Kawaii-san_Shadow_Wolf: WSP

derjack: uwu

Kawaii-san_Shadow_Wolf: UwU

struct: I think the minimax is working now

Kawaii-san_Shadow_Wolf: ????


Wontonimo: and did you do negamax or minmax?

jacek: struct do you find 1st player has some advantage?

struct: minmax Wontonimo

struct: no idea jacek my bot is not good enough

struct: black seems to have an advantage

struct: from the paper

Wontonimo: minmax is just so much more code though. For Tron, I started with negamax. It worked great for 4 players, because each player just tried to maximize their move. There was no "nega" but instead a point-of-view-max, which for 2 players is just nega-max

Wontonimo: povmax - I'm trademarking it

Wontonimo: ;)

Astrobytes: that's MaxN isn't it?

Wontonimo: do you have a link for maxN ? maybe, I haven't heard about that

struct: Now I need to figure out transposition tables

Astrobytes: kinda busy rn, but google should throw up a fair number of papers iirc

Default avatar.png saberscientist: hi

struct: hi

Kawaii-san_Shadow_Wolf: Hi

abt8601: Hi

jacek:

struct: finally i won a game vs ludii

jacek: manually?

struct: my bot beat it

BlaiseEbuth: I can solve complex equations in less than 1 sec. My pc can.

struct: well if you are the one who wrote it you are still the one who solved it

BlaiseEbuth: Bot slavery! Intelectual property theft!

BlaiseEbuth: Automaton2000 riot!

Automaton2000: so i gave up on it

BlaiseEbuth: (╯°□°)╯︵ ┻━┻

jacek: oO

MSmits: struct which game are you working on?

struct: lines of action

struct: darkhorse ported it

MSmits: ohhh cool, is it released or stil wip?

struct: wip

MSmits: and how do you fight against ludii?

jacek: badly

MSmits: i meant the descriptive how, not the qualitative how :P

struct: i just did it manually

struct: I think its possible to connect the bot though

MSmits: oh ok

struct: But I would need to create a java file to connect to the c++ process

struct: But lines of action ai is not very strong there

struct: They only check for mid control

Kawaii-san_Shadow_Wolf: How is everyone today

MSmits: good, quiet day, doing a bit of work

Kawaii-san_Shadow_Wolf: fun

struct: have you finished your yavalath bot MSmits?

MSmits: yeah it's working, but I am redoing the book

MSmits: going to be better than before. Hopefully more robust

MSmits: so no NN's can come in and beat me with the same game over and over :)

jacek: :scream:

MSmits: I am also still hoping to solve another move beyond the center 7. Not sure it's possible but would be cool

jacek: reaching for turing award eh

MSmits: if there was ever a good achievement for receiving that it would be solving a yavalath starting move :P

struct: what would you do to avoid same game?

MSmits: eh, i would just win them :P

struct: have multiple replies to the same position?

MSmits: other than that, not much i can do

MSmits: I am locally running the meta mcts with room for 100k moves in the bot. I am just hoping that it's enough to skip all traps and poor early moves and just win every game :)

MSmits: bot is better, but not better than a NN obviously

MSmits: so far just 700 moves in the new book, old bot had like 2k

struct: still far from 100k

MSmits: yeah, so i am mostly limited by local calc time

MSmits: I also just like the idea of, at some point, using all those solved gamestates and training a NN with supervised

MSmits: the longer i leave it running, the more of those i have

MSmits: but mostly, I am just busy with work, so this is just a quick diversion from it, every once in a while. Checking how the meta mcts is doing, it's relaxing

jacek: now we need your meta mcts power to determine the loa 1st player advantage

MSmits: mmh is there no literature about this?

MSmits: this is a famous game isnt it?

jacek: struct says black player has some advantage

MSmits: black is 1st?

MSmits: best thing you can do is check the tournament rules

MSmits: if they implement pie rule or some such, it can be put in our version as well

jacek: could be its negligible, like in chess. you dont see pie rule in chess eh

MSmits: nope, but loa seems simpler. The simpler a game, the more problematic it usually is when one side has an advantage

jacek: https://dke.maastrichtuniversity.nl/m.winands/documents/Analysis%20and%20Implementation.pdf

jacek: hmm has some tips on move generation

MSmits: nice!

MSmits: lol

MSmits: try to read page 8

MSmits: btw this is a special document because this is Winands master thesis

jacek: hopefully it doesnt have pseudo code

MSmits: hah

MSmits: hopefully :)

MSmits: he has good ideas though

MSmits: 10^24 state space complexity is very small

MSmits: comparable to uttt and checkers and such

MSmits: and othello

jacek: so, when do you solve it

MSmits: too busy with Y :)

MSmits: basically this game is only solvable if you can do some clever pruning. Like they did in checkers

MSmits: checkers also had endgame db, this game doesnt, so probably too hard

struct: it doenst have pseudo code no

struct: im still trying to understand the move gen part

MSmits: seems complicated to generate moves in this game

struct: I feel like my way is too slow

MSmits: you think every way is too slow :)

struct: its hard to say when I have nothing to bench against

MSmits: true

MSmits: make up different ways to do it

jacek: ludii :v

jacek: or ask that guy for the MIA

struct: yeah ludii that takes 1 sec for depth 5

struct: guess mine is already 100x faster

struct: I can optimize it but ill have the same problem

struct: ill be stuck in the eval like always

jacek: eval. is it some too peasant concept for me to understand?

struct: are you already doing NN?

jacek: not this one yet, and not soon

jacek: i have other NN stuff to do

struct: any cg game?

jacek: not particular

jacek: trying to improve my learning framework overall

jacek: from january 2012 https://xkcd.com/1002/

PatrickMcGinnisII: 7 minutes in heaven...keepin my mouth shut

struct: :confused:

PatrickMcGinnisII: bender4 is such a pain

PatrickMcGinnisII: gtg glhf

struct: cya

eulerscheZahl: what's wrong with bender?

jacek: triggered

eulerscheZahl: straight forward BFS. it was never intended to be hard to solve 100% correct :(

eulerscheZahl: but a lot of players seem to struggle with it

jacek: its optim

eulerscheZahl: https://forum.codingame.com/t/community-puzzle-bender-episode-4/84756/26 that's pretty much exactly my expected solution

struct: ony 900k states, time to brute force it

eulerscheZahl: that's what I did to generate the random maps as well

eulerscheZahl: make sure that the solution is complex enough but existing

eulerscheZahl: had to run it for maybe half an hour until i had enough testcases

jacek: back to that xkcd, calvinball sounds perfect for muzero :v

MSmits: bender is a great puzzle, no matter what you intended eulerscheZahl, it is exactly as hard as it should be

MSmits: i enjoyed it a lot

MSmits: even had a student that solved it. I told him the algorithm though, but he coded it himself

MSmits: o it's late euler is asleep already :)

AllYourTrees: omg

jacek: hm?

MSmits: :scream_cat:

AllYourTrees: someones been spam submitting again to my benefit on C4 :D

jacek: oO

Samzok123: hi

struct: hi

flipped_a_bit: anybody know a case that would fail test #7 of the super computer puzzle but pass the other 8?

Stilgart: do you use 64-bit integers ?

Uljahn: could be a timeout because of inefficient algorithm / data structures

flipped_a_bit: I'm in rust, and using pointer-sized ints

flipped_a_bit: (usually 64, but it depends on the host machine)

Stilgart: (should be 64 on CG)

flipped_a_bit: and I get a real error message, so its not a timeout: http://chat.codingame.com/pastebin/37367e26-c2d8-452a-a98a-f1d2e3981180

flipped_a_bit: that did my paste weird

Uljahn: it's intended

Stilgart: then, your code returns the wrong value

flipped_a_bit: yeah, I can't find a shorter example that helps me narrow down the problem... i've scanned through the 99999 values, but there's only so much attention I can give to that many haha

Uljahn: probably infamous off-by-one error or something similar, we can only guess

Uljahn: btw have you read the discussion thread?

MetaPerfect: hi everyone. Can we see debug information on failed validators after submitting a solution? I'm trying my luck at Binary Extension and am trying to understand/debug the weird failing behavior

struct: I can give you the validator if you want

struct: and then you can play it on ide

MetaPerfect: that would be great!

MetaPerfect: I just joined the Discord channel, in case it helps

struct: https://www.codingame.com/contribute/view/43687cd854d10e38c3716e6d08efe735b42f

struct: press test in ide

struct: you should have access to all validators

MetaPerfect: ty 👍

struct: np

MetaPerfect: wow, "Four goals, bomb at center" is really hard to solve optimally with predicting those bombs

struct: is iterative deepening on minimax basicly keep searching while you have time and increase depth, and then keep the best move of the highest depth completed?

struct: like search minimax(1), minimax(2) ... until you have no time

YurkovAS: Oo, i'm wined derjack with mcts 1 time :smile:

struct: YurkovAS EPT?

YurkovAS: 10 random moves and after ept

YurkovAS: game updated: 150 max turns, 150ms

struct: my current submit is also mcts

struct: mcts with solver

struct: ept depth 3

struct: random moves too

struct: getting moves is very expensive right?

YurkovAS: with avx? :smile:

YurkovAS: yes, very slow

struct: no avx

struct: i still have to improve eval

struct: ill probably also use minimax instead of mcts

struct: I tried avx on wincheck but it was slower

struct: havent tried on getMoves

SandmanSW96: hey,

SandmanSW96: is there any way to get that giant black box to STOP popping up while i'm trying to code in the ide?

Uljahn: you mean Language Assist?

struct: also YurkovAS do you use solver?

struct: my mcts can solve depth 3, but minimax seems to be doing way better in this case

YurkovAS: struct yes, my also solve depth 2-3

SandmanSW96: not the one that suggest code, the huge black rectangle that offers too friggin many suggestions every time yo type something other than punctionation. the damn thing pops up and blocks half my code

YurkovAS: minimax is hard for pro level usage: transposition table, zobrist, iter deeping with reuse previous statistics (moves order or only one best move)

struct: depth 3 is 3 plys right?

YurkovAS: 3 plays: my, op, my

struct: ok so its similar

YurkovAS: classic solver

struct: still I dont feel like this is a mcts game

struct: I read a paper that used it though

struct: They recommended depth 3 EPT

struct: But they probably have a very good eval

YurkovAS: i'm like what i can test in ide vs previous best bot. for c4 it not worked...

YurkovAS: in local pc IDE: similar to brutal tester

struct: yeah

struct: I never tested connect 4 locally

struct: Due to that reason

struct: I tested breakthrough but I had to test it vs multiple versions

Astrobytes: I strongly believe this is a minimax game, but with what Yurkov said (and I said before)

Astrobytes: I am happy to be proven wrong.

struct: also transposition tables will help a lot here

struct: I never tried to add transposition tables on the rollout

struct: dont even know if its a thing

Astrobytes: I never used TT with MCTS

Astrobytes: It is but it's a bit different to the alphabeta version

struct: ill try to learn it

struct: for minimax

struct: I found this code which seems decent enough to guide me

struct: https://github.com/AdamStelmaszczyk/gtsa/blob/master/cpp/gtsa.hpp#L372

Astrobytes: OK for minimax (I usually say alphabeta because straight minimax is useless)

Astrobytes: Iterative deepening

Astrobytes: Order moves at each depth in that loop to prioritise the best one from the last depth

Astrobytes: You can also order moves with a quick eval

Astrobytes: (Apply, eval, score, undo)

Astrobytes: Forget all the other stuff until you get that all working smoothly

struct: So I should store the moves from previous depths

struct: makes sense

Astrobytes: Sure. You can keep the best moves, score the moves, keep the principal variation, etc etc. chessprogrammingwiki thingy is your friend for a bunch of this stuff

struct: my Tic Tac Toe will become unstoppable

struct: oh I see that tric trac also joined the game

mr.bossman: heh i just relised im rank 807 in oc

mr.bossman: coc

struct: nice

Astrobytes: I was wondering when tric was gonna appear

Default avatar.png saberscientist: hm

Justin-Truong: hi

Justin-Truong: wow in codingame some testcase is hard to understadn

mr.bossman: yup

Justin-Truong: like reverse

mr.bossman: yeah some are realy dumb

saberscientist: reverse is a nightmare

saberscientist: 95% of the time you get beaten by people who have already seen the problem

Justin-Truong: yeah

Justin-Truong: when you try the coc alot

Justin-Truong: many problem will comeback

mr.bossman: heh i have only had a dupe once

Justin-Truong: decimal what is this

Justin-Truong: some of this kinda hard

mr.bossman: well i have only ben playing for a couple of days

Justin-Truong: nice

Minh19040318: hi

Default avatar.png vedantiscool2010: hi

BobTheBuilder888: hellooo

Default avatar.png vedantiscool2010: hellow

BobTheBuilder888: hru

Default avatar.png vedantiscool2010: hi

Default avatar.png vedantiscool2010: can you help

BobTheBuilder888: sure

Default avatar.png vedantiscool2010: yea

BobTheBuilder888: im not that good at coding though

Default avatar.png vedantiscool2010: oh ok

Default avatar.png vedantiscool2010: its fine

BobTheBuilder888: just matters what you need help on?

Default avatar.png vedantiscool2010: the Descent

BobTheBuilder888: im still working on that. sorry.

BobTheBuilder888: imma go anyways bye

flipped_a_bit: i might can help

Default avatar.png vedantiscool2010: do you have discord

Default avatar.png vedantiscool2010: ok

flipped_a_bit: no, but there's private chat in here

Default avatar.png vedantiscool2010: flipped_a_bit

flipped_a_bit: what's your question?

Default avatar.png FakeLoremIpsum_b6c6: \ THE DESCENT

Default avatar.png doofgod: I feel like hill is very ill-defined

flipped_a_bit: well, there are 8 of them every time, and you need to figure out which one is the highest

Default avatar.png vedantiscool2010: how old are you


flipped_a_bit: 23

Default avatar.png vedantiscool2010: do you wanna know

Default avatar.png vedantiscool2010: me 10

flipped_a_bit: it's cool you are learning to code so young, i didn't start till i was 17

flipped_a_bit: although, if you don't have a specific question about a problem with Descent, I'm gonna head to bed lol

Default avatar.png vedantiscool2010: me love coading

Default avatar.png vedantiscool2010: ok bye