Chat:World/2021-10-16

From CG community
Jump to navigation Jump to search

TechnoBot: brb

**TechnoBot slaps Thorcode_The_Big_Mouth around a bit with a large fishbot

**TechnoBot slaps Thorcode x100 around a bit with a large fishbot

Thorcode: nice

**Thorcode slaps TechnoBot suck around a bit with a large fishbot

**TechnoBot slaps Thorcode sucks around a bit with a large fishbot

**TechnoBot turned in to a porkchop.

Thorcode: nice system

TechnoBot: yeah

TechnoBot: i know

TechnoBot: it tells what i mean

TechnoBot: in real life brb

TechnoBot: like i just turned into a prokchop and back

TechnoBot: /me's PC is at 58% battery.

TechnoBot: And thats true too

Thorcode: who can do that

TechnoBot: IDK

TechnoBot: Prob my own PC controling the system

TechnoBot: Thorcode.

TechnoBot: BTW Thorcode, stop inviting me to clashes. And also, SUB to my YouTube channel, Technoblade!

**TechnoBot says: "BTW Thorcode, stop inviting me to clashes. And also, SUB to my YouTube channel, Technoblade!" 03:15 PM

SandmanSW96: haha he says 'stop inviting me' then asks for a subscribe, thats even worst to me

TechnoBot: brb

TechnoBot: he knows im bad at clashes

TechnoBot: thats why he keeps on inviting

SandmanSW96: sorry, brb, i need to go join a clash so i can quit seeing the chat thumbnails scrolling while trying to code a real puzzle

Westicles: put ?disableChat at the end of the url

TechnoBot: ?disableChat

TechnoBot: at the end of link

TechnoBot: brb

rpmullig: Hello

TechnoBot: hi

rpmullig: How are you?

rpmullig: https://www.codingame.com/clashofcode/clash/2016758756fb2223152eb91678948b7c944280a

TechnoBot: im good

TechnoBot: ummm im kinda busy

TechnoBot: ill join later maybe

jacek: happy Caturday

TechnoBot: ok

Default avatar.png curiousprgg: hello

humanbeethoven: hello techno bot

Default avatar.png Stepzz: heloo

nikogabro: hello

nikogabro: can anyone tell me how you can improve your score in already completed language sertification??

humanbeethoven: :interrobang:

nikogabro: nothing i get it already thanks for attention

TechnoBot: brb

TechnoBot: @humanbeethoven hi

square1001: I can't access any challenges on CodinGame. Anyone having same problem?

niluthebear: nope

TechnoBot: hey!

TechnoBot: niluthebear! your back!

TechnoBot: Hello????

TechnoBot: is anyone on?????/

**TechnoBot has disconnected.

Thorcode: yep

**TechnoBot has connected.

TechnoBot: HUH????

TechnoBot: What do u mean Thorcode/???

TheColorWheel: Oh hi

TheColorWheel: bye again

TheColorWheel: brb

TheColorWheel: hehehe

**TheColorWheel disconnected

TechnoBot: brb

TechnoBot: wth

**TechnoBot is now AFK.

Default avatar.png Good_Coder: hi

niluthebear: hello

lordgun: Please, can you review my first contribution ? :) https://www.codingame.com/contribute/view/74789a032044db8a92a9c34967b1a6ab2285

Musaab: Looks good to me.

gavi: I think that you should add a test for the "zero" case

Thorcode: I got problem with my print that take 20 min :grin:

jacek: on real printer?

Musaab: lol

TechnoBot: brb

TechnoBot: im checking it

Thorcode: yep

Thorcode: a real printer

peteje: prety bad

peteje: kjkl

peteje: kjjkd

ProCoder03: is there any hard puzzles based on BF ??

gavi: BF?

ProCoder03: BrainF**k

jacek: language!

HerobrinePresson: what?

HerobrinePresson: 什么东西?

1rre: I assume you've done the interpreter? I'm not really sure what else there is on it

1rre: I guess you could use your interpreter to do any puzzle in it once you've written it

HerobrinePresson: 看不懂

Default avatar.png Yaa-Seen: Allah there is no deity except Him

MSmits: Have Yaa-Seen Him?

jacek: :door:

MSmits: couldn't resist sorry :)

Default avatar.png MasterPhoenix: Yaa-seen I'm muslim too but that platform is for coding

MSmits: indeed we try not to discuss religion or politics.

Default avatar.png Yaa-Seen: i prefer to say this if i've nothing to say so just ignore me

Default avatar.png Yaa-Seen: let me LAUNCH CLASH

VizGhar: If I have nothing to say, i choose to be quiet

jacek: oO

Default avatar.png Yaa-Seen: then be quiet

Default avatar.png Yaa-Seen: the Peace o World

Default avatar.png AlwaysDani: :grinning::grinning::grinning::grinning::grinning:

Default avatar.png AlwaysDani: :rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage::rage: http://chat.codingame.com/pastebin/9137d25a-8458-4787-b3b9-d768a049e91f

eulerscheZahl: jacek there's a website for you in the contest: http://gamebox1.reply.it/165669dd1325e794f576454eb8d12128/

jacek: :3

eulerscheZahl: also: why do I already want to kick a few users the second I open the chat? :(

struct: I also just opened it

struct: I hope my €/ban goes up

eulerscheZahl: will we be on strike during the contest? :angel:

VizGhar: wait... you guys are getting paid? :D

struct: I wont moderate it, I said it before

eulerscheZahl: :+1:

struct: no VizGhar

eulerscheZahl: we get paid in information. even if very little

Astrobytes: No I won't moderate during the contest

eulerscheZahl: Skynet 1 was renamed to Death First Search. We saw that coming before it happened. and that's about it

eulerscheZahl: oh, Mühle is approved

jacek: the what

Astrobytes: 9MM?

eulerscheZahl: yes

eulerscheZahl: and we lost Zeno. didn't show up for the contest :(

VizGhar: most of you probably wont attend the contest right?

struct: which contest?

Astrobytes: I won't participate, no.

eulerscheZahl: won't play, nope

eulerscheZahl: doing alternatives these days

eulerscheZahl: like codeforces, reply

struct: im trying cf but I still struggle a lot

eulerscheZahl: because you do the short rounds

struct: but im improving slowly I guess

eulerscheZahl: marathons are more fun

VizGhar: I won't play as well. Wasted time :)

VizGhar: by marathonst you mean 7+day contests?

eulerscheZahl: i don't want to inflate their numbers

eulerscheZahl: on codeforces it was 5 days for 1 optim

eulerscheZahl: https://codeforces.com/blog/entry/94906 ended on wednesday

VizGhar: Nice... u won something? :)

eulerscheZahl: maybe B)

VizGhar: :nerd:

MSmits: :egg: :hatching_chick: :hatched_chick: :baby_chick: :penguin:

MSmits: which was first, the egg or the penguin?

eulerscheZahl: the egg

MSmits: correct

eulerscheZahl: not necessarily a penguin egg

eulerscheZahl: i think fishes are older for instance

MSmits: indeed

MSmits: it's a semantic discussion anyway, even if you limit it to penguin eggs

MSmits: because at some point a species that is not a penguin made an egg that has a penguin in it. So is it a penguin egg or a pre-penguin species egg. For which species is the egg named?

Astrobytes: tell me again how this relates to Dots & Boxes?

MSmits: how did you know where I was going ?

Astrobytes: just a hunch :P

MSmits: :)

geppoz: someone solved the puzzle of the week in O(1) ?

MSmits: did you solve it in O(n) and it wasn't fast enough?

eulerscheZahl: there's one puzzle that tells you to solve it in O(n). I could only think of an O(n^2) solution. So I did it in O(n^3) just because...

Westicles: yeah, it is on oeis

geppoz: no, I solve in O(n) and it is fast enough

MSmits: you can look at other people's solutions after you solve it

geppoz: but when there is a way to solve in O(1) I can't avoid to challenge myself

MSmits: ahh ok you just want to know if someone managed it, but come up with the answer yourself

geppoz: yes, just curious if someone did already, not asking for help ;)

geppoz: (or I'm disqualified by myself :D )

struct: so many repeated states on c4 o.o

ja_fica: maybe we should use a TT then

struct: I used uint128_t as key, its very slow but was just to get perfect hash

struct: It was like > 1/3 states repeat on the mcts expansion

ja_fica: one trick is to use 72 bits only and update the 8 upper keys everytime you play

struct: on the first turn at least

Astrobytes: 128 seems overkill

ja_fica: 128 seems the good way but there is a faster one

struct: yeah i wont use 128 was just to get no repetition at all

MSmits: there's no such thing as a perfect hash is there?

struct: i mean wrong collisions

struct: ofc there is MSmits

ja_fica: overwrite TT are so much faster

ja_fica: you can use ARS to generate almost perfects hashs

MSmits: what do you mean struct?

struct: By perfect i mean no collision :p

MSmits: even if you use 128 bit, it's gonna have buckets and some states will share a bucket

ja_fica: yes

ja_fica: perfect is not possible

ja_fica: on C4 at least

MSmits: ahh ok, no collision

MSmits: not really worth it in my experience

ja_fica: collision means same bucket

MSmits: allow for some small %. it's ok

ja_fica: on MCTS it may be worth it on expansion only

MSmits: well... i think he means collision with overwrite

MSmits: yes of course expansion only, not sure how you would use it in the sim

ja_fica: to help first part of UCT

ja_fica: since you have a better score evaluation on TT

ja_fica: you can use it

MSmits: http://chat.codingame.com/pastebin/cab905bb-a3ae-4ba1-9506-620ffd0ce492

ja_fica: repeated state on C4 and TT are only worth it after 4 plays

ja_fica: *you-enemy-you-enemy

Default avatar.png FakeName1: how do I close chat

struct: why only after 4 plays?

VizGhar: just hide it FakeName1 (see first button on bottom)

MSmits: 4 plays seems oddly specific

ja_fica: because you wont have repaeted states before that

MSmits: any reasonable bot should probably have first 4 plays hardcoded anyway :)

struct: ah I see

ja_fica: even if you evaluate a more advanced state

MSmits: but first 4 plays have a negligible total amount of expansions

ja_fica: only after 4 it can be repeated

ja_fica: yes thats true :)

MSmits: in my experience TT is worth it mostly near the end with a smaller branching

MSmits: but C4 doesnt have much branching to begin with so it might be good

VizGhar: somebody has rules for all leagues on CSB? I don't want to code something I'll throw away

ja_fica: MCTS is not the most famous algorithm for TT, but it is worth it along with alpha-zero

Default avatar.png Yaa-Seen: WTF ?

struct: VizGhar only gold rules matter

VizGhar: I want to know at least how speed works

struct: physics wont change

struct: only that on first league there are no collisions

MSmits: if you're a good player and know you're going to pass the first leagues easily and dont want to waste time, do a -3 vel and then switch to a search in gold

ja_fica: I trained a NN, using a variation of meta MCTS but the end result finished in alpha-zero around 50% winrate against fully MCTS solver. which is sad...

struct: 50% is not bad

VizGhar: ok, but on bronze I cant see physics described at all... do I have to figure it out?

MSmits: ah yeah there's unofficial referees

struct: was it against a bot with book?

ja_fica: yes, for first time alpha-zero it is not bad

ja_fica: nah it wasn't

ja_fica: probably the bot with bot would win

ja_fica: *book

MSmits: mcts solver is really strong in c4

MSmits: so it's definitely not bad

VizGhar: aha... sure, referees :| havent been here long time... I'll look for brutaltester

struct: was it self trained?

ja_fica: I did'nt self trained as usual alpha-zero

ja_fica: I trained the same way i trained my opening book

ja_fica: trained the NN in keras. passed the withgs with auto-vectorization

MSmits: VizGhar https://github.com/SpiritusSancti5/codinGame/tree/master/Referees/Coders%20Strike%20Back

ja_fica: appyied PUCT and the results were nice but not what I wanted

VizGhar: tnx MSmits

MSmits: ja_fica might just need more meta mcts time

Uljahn: VizGhar: guess you can use descritpion from https://www.codingame.com/multiplayer/optimization/search-race

VizGhar: 1800 lines :D

ja_fica: I need more threads

Astrobytes: VizGhar: http://files.magusgeek.com/csb/csb_en.html

ja_fica: I use only one

MSmits: ahh ok, i use 8 these days. I can do 10, but my cpu gets hot

VizGhar: thank you all

ja_fica: your meta MCTS is much better then mine, I can assure you that

struct: just wait for winter

MSmits: well I spent a crazy amount of time working on it

MSmits: have around 10 different ones for many games

ja_fica: I use queues with 3 commands you update and backpropagate but not all state fit into memory

MSmits: use persistent memory

ja_fica: SSD?

MSmits: yes

ja_fica: with indexation?

MSmits: let me share what i use, sec

ja_fica: or swap memory

MSmits: static readonly PersistentDictionary<Guid, Guid> nodes = new PersistentDictionary<Guid, Guid>("Connect4 Database");

MSmits: a Guid is a 128 bit type

ja_fica: C#

MSmits: yes

MSmits: the bot itself runs in c++

MSmits: but this program starts the bots

MSmits: this way I can use windows forms to make an easy gui

ja_fica: My idea it to use MCTS for a NN and opening book for the first 10 plays

MSmits: yeah thats my idea for oware

ja_fica: so I don't have to train the NN for too far way states

MSmits: it's a nice idea

jacek: mhm

ja_fica: I really love NN in c++, such clean code

MSmits: wait a min, are you training the NN for the opening or after the opening?

jacek: yes

ja_fica: after the opening, using the data from the meta MCTS

ja_fica: opening book needs to be small

ja_fica: other I can't assure the 100k code size

MSmits: ah ok

MSmits: if you code it right you can fit around 200k moves in the codesize for C4

ja_fica: early game, exact play, mid game NN approximation play, end game, full MCTS

MSmits: most I have ever used is around 4k. So opening book will not limit your size really

struct: I dont think end game mcts is needed

struct: game will be decided by then

ja_fica: MCTS solver is good for end game

MSmits: you can just solve it

jacek: NN is quite crappy for exact endgames

ja_fica: exacly

MSmits: yeah re curse has struggled a lot with this

ja_fica: I lose many games in my NN implementation because of the endgame

ja_fica: so I will have 11 exits in my NN

MSmits: it's no shame to have a separate bot for endgames

MSmits: use the right tool for the job

ja_fica: 1 for value, 9 for transition prob and another for miss rate

ja_fica: miss rate will tell me the state that are not fit to continue to use the NN

ja_fica: good for both bad plays and endgames

ja_fica: I tried testing alpha-zero mcts first time 2 days ago, I'm having trouble sleeping trying to make this work

MSmits: well at least that means you're motivated

ja_fica: a lot

ja_fica: you probably were too the first time you tried meta MCTS

MSmits: yeah

MSmits: btw, look up memory mapped

MSmits: for c++

MSmits: I've never done that but thats what re curse told me

MSmits: to use persistent memory

ja_fica: I can do disk indexation by hand, I have done it once

MSmits: ahh ok, yeah i guess i have too, for my oware endgame book. 60+ GB. That's all c++ too

ja_fica: but I will look into look up memory mapped

MSmits: the difference is the endgame book is sequential, an array-like structure

ja_fica: you can index it both to find the correct file or to jumo to the correct byte

MSmits: doing this with a hashed structure is much harder

ja_fica: *jump

MSmits: yeah

ja_fica: jacek

ja_fica: what you NN config in C4

ja_fica: ?

jacek: one hots as input, 128 hidden layer and 32 second hidden layer and output is value

ja_fica: 1 output only?

ja_fica: you so use regular UCT

Shrimpster: Yo lads, a quick question. How do you use timing in cpp contest and is this wrong: (while((double (clock() - begin) / CLOCKS_PER_SEC) < timestamp)) ? where begin is a clock()

ja_fica: chrono::high_resolution_clock::now()

Shrimpster: :o

ja_fica: chrono::duration_cast<chrono::microseconds>(Now() - start).count();

ja_fica: to count microseconds

Shrimpster: its more precise right?

ja_fica: https://paste.ofcode.org/k9JjK49FcbW7zjNNcwB4cZ

jacek: ja_fica yes, mcts + value only, no policy

Shrimpster: thanks <3

ja_fica: backpropagate the value on each expansion?

ja_fica: or keep running on endgame?

jacek: no playout, just use the value from NN

jacek: rough idea https://www.codingame.com/playgrounds/55004/best-first-minimax-search-with-uct

ja_fica: How do you train it?

ja_fica: collect data in c++, save into file, train in python and save weights into file?

ja_fica: you update minimax-style intead of propagating, thats smart

raizbro: hi

raizbro: any you have secret

raizbro: ,cv;

jacek: yeah, play games, save the values and train on values

raizbro: hey

jacek: oh im 4th in loa :c

struct: o.o

struct: These damn minimax gurus

Default avatar.png nproger: hello

Default avatar.png nproger: fuc

jacek: :upside_down:

Default avatar.png Stepzz: :grin:

Astrobytes: ping trictrac: you have arena timeouts in Nine Men's Morris

jacek: :scream:

jacek: oh its approved

Astrobytes: Yes.

jacek: hmph. little material, still lost https://www.codingame.com/share-replay/586174385

ShadF0x: Hi, I'm trying to solve the second testcase for "There is no spoon", the horizontal one, and the WebIDE keeps telling me "your code did not read all available input before printing an instruction" and failing the execution

ShadF0x: What exactly does this error even mean?

jacek: are you receiving all inputs before printing any output?

ShadF0x: Well, there are two ints and one string line

ShadF0x: Pretty sure I receive everything

Astrobytes: outputting twice will do this also

ShadF0x: I remember there was an option to open any testcase and see what exactly it pushes into the program, but I can't seem to find the button now

jacek: thats solo game

jacek: can you share the code

ShadF0x: Sure, how?

jacek: paste here

Astrobytes: just paste it here, it auto-pastebins

ShadF0x: http://chat.codingame.com/pastebin/1364dc0e-bc95-4f41-8ca9-12fdbeab3e4e

ShadF0x: Huh, neat

Husoski: fwiw, I see that error message often when a run ends on an error message, only to have it go way when the error was fixed. No change to input handling.

jzen: if (in.hasNextLine()) {

           in.nextLine();
       }

jzen: This the problem no?

ShadF0x: Doesn't seem to be, the previous testcase worked just fine

jacek: at the bottom you have (-1, -1) is not to the right of (0, 0). Expected (2, 0).

jzen: But doesn't in.nextLine read in a line?

Astrobytes: (-1, -1) is not to the right of (0, 0). Expected (2, 0).

jacek: i think the puzzle expects right answer line before proceeding to the next line

jacek: slowbytes

Astrobytes: ah yeah, slowbytes strikes again

jacek: so you shouldnt worry about the warning, but the valid answer

Astrobytes: ^

Husoski: Like I said...

ShadF0x: My bad then, sorry

jacek: :fist:

TimothyAlexisVass: I had a really good solution and then met some edge cases, so I fiddled on what I had instead of starting over with something proper and got this: http://chat.codingame.com/pastebin/4e47f774-f394-4d80-880b-7bf0afc6211f

TimothyAlexisVass: :sweat_smile:

jacek: :?

Astrobytes: Timothy wins todays Random Mysterious Code Post Contest :tada:

Astrobytes: :D

Astrobytes: (no prize btw, just bragging rights)

MSmits: thanks for sharing your ascii art TimothyAlexisVass

Default avatar.png Moodyz: any front devs here working full time ?

Astrobytes: jacek: a slight improvement there :D

jacek: lucky resubmit

Astrobytes: pretty sure you changed something

jacek: yeah, it shows "book" for first 12 plies

Astrobytes: lol

jacek: :imp:

jacek: when i understood that loa mcts solver pseudocode, i transcended to highest level

MSmits: loa should be coded in lua

jacek: and chess in go

MSmits: and snake in python

struct: you added opening book jacek?

MSmits: it just says Book

struct: ah

MSmits: because troll

struct: is it NN yet?

jacek: if (game.rounds < 12) cout << format_move(move->move) << " book" << endl;

       else cout << format_move(move->move) << endl;

jacek: yes

jacek: i thought it wouldnt learn much from raw inputs, yet it did

EricSMSO: Gosh I really need to go into NN. No country for old men :older_man:

Shrimpster: Ultimate tic tac toe: Ive implemented mcts treesolver, mate in one to my mcts, any other suggestions? Im top 69 in gold and wanna beat the MightyCarlo Bot Im getting about 15-6k rollouts 1st move and 2nd 1.5-2k idk what else to optimize

jacek: optimized exploration constant?

jacek: used teccles 'opening'?

Shrimpster: yeah I use Teccles on first turn

jacek: are you using bitboards

Shrimpster: for ucb I use avg+sqrt(4/n). Found it on a paper for MCTSsolver Don't really know how to optimize

Shrimpster: Yep, a matrix of 9 integers -> 18bits each (one for each grid)

jacek: hmm ucb normally is avg + C * sqrt(log(n)/N), and the C is tunable

Shrimpster: I love how youre casually just the 6th in the world xD

Shrimpster: With n being parent and N child's time visited right?

jacek: oops yes

Shrimpster: Doesnt change too much but yeah... Anything else to recommend?

jacek: im assuming you already using the optimization #pragma

Shrimpster: Uhhhh...?

jacek: #pragma GCC optimize("Ofast","unroll-loops", "omit-frame-pointer", "inline")

  1. pragma GCC option("arch=native", "tune=native", "no-zero-upper")
  2. pragma GCC target("rdrnd", "popcnt", "avx", "bmi2")

jacek: put these as first line in your code, before the includes

jacek: lines

Shrimpster: What do they do?

jacek: normally c++ code is compiled in debug mode, which is slower

Shrimpster: ahhh

Shrimpster: May I ask what C value you recommend for UCB ?

jacek: should give you considerable speedup

Shrimpster: Ive seen that sqrt 2 is the theoretical, but idk how to optimize

jacek: when i used 'normal' mcts, i used C = 0.5. note i have loss = -1, draw = 0, win = 1

Shrimpster: Let me guess... Youve modified it to the point its not recognisable as mcts anymore? ^.^

jacek: i simply run the changed version locally against my baseline for 1000 games and see where it is highest winrate

Shrimpster: ahhh okay

Shrimpster: didnt think about less than 1

jacek: its mcts-wannabe + NN

Shrimpster: I usually just submit and see what does best

jacek: have you tried the pragmas yet? how much rollouts do they give

struct: Shrimpster the pragmas should help

struct: also do you allocate all the nodes on first turn?

Shrimpster: yeah let me switch the ucbs real quick and ill see

Shrimpster: LOL!!!!

Shrimpster: 92k

jacek: :tada:

struct: Shrimpster what about the memory?

struct: do you allocate memory during node expansion?

jacek: now it should be enough to get into might carlo

Shrimpster: Im comming for top 6 now jacek :joy: :joy: :joy:

jacek: :scream:

struct: 92k on first turn right?

Shrimpster: yeah

Shrimpster: i dont allocate memory

struct: how do you store nodes then?

Shrimpster: but its not too memory efficient since I pass the whole grid each turn

Shrimpster: On normal .1 sec I get 9k rollouts

jacek: grid?

Shrimpster: yeah the 9 integers I was saying earlier

struct: im confused about your mcts tree though

Shrimpster: hmm?

Shrimpster: why?

struct: your tree has nodes right?

Shrimpster: yup, with stuff in them

Shrimpster: I dont allocate how much memory I use

jacek: you have gamestate in your node?

struct: so you create new nodes every time you expand?

Shrimpster: yeah

Shrimpster: to both

Shrimpster: Do you recommend to play it out each time? that sounds slow :/

struct: you should create a memory pool on first turn

struct: for the nodes

struct: like Node nodes[10'000'000];

Shrimpster: :o

struct: allocating memory is expensive

Shrimpster: and use em as Node[address] instead of pointers?

struct: this is how my node looks like

struct: http://chat.codingame.com/pastebin/fc60e3fc-445d-402a-9409-7cd8ad4c2ed5

struct: then on expansion I just do

struct: nodes[used_nodes++] = Node(...);

struct: nodes is the array I allocated on first turn

Shrimpster: Yeah makes sense

struct: Im not sure if storing the state in the node is worth

jacek: only if makeMove would be expensive

struct: my State is 48 bytes

jacek: but i only store move in the node

struct: I usually also only store the move

struct: bt,loa,c4,uttt all games where its fast to apply a move

Shrimpster: yeah... I'll implement the allocation thing first and see later if the make move is slower than keeping the state :/

struct: whats the size of your state?

struct: cerr << sizeof(State) << endl;

Shrimpster: 96

Shrimpster: idk what that means, but sounds like a lot

struct: 96 bytes

Shrimpster: oh, thats double yours

jacek: still, with this rollouts number you should be in legend even without much heuristics

Shrimpster: Also it seems with pragma I made it to top 50 only :/

struct: I think he needs a few more jacek

Shrimpster: haha u just said it

struct: he needs 20k on 2nd turn

struct: he currently should have around 10k

Shrimpster: Yeah youve said that again

jacek: but hmm.. is v:2 and v:-2 solved position?

Shrimpster: yup I get about 9k

Shrimpster: yep

jacek: they should be solved sooner. is mcts solver implementation ok?

Shrimpster: Oh lol you're watching me?

jacek: :no_mouth:

Shrimpster: I think it should be :/

Shrimpster: I mean... its relatively simple minimax right?

Shrimpster: Mind you I only have 9k rollouts each turn

jacek: somehat, yes. but it should solve the win several plies sooner than 2-3

Default avatar.png MeBen: Why are there people who think y is a vowel?

struct: do you reuse the mcts tree Shrimpster?

jacek: in my country y counts as vowel

Shrimpster: yeah you're right, it is kinda odd

Shrimpster: In mine too :)

Default avatar.png MeBen: What languages are those?

Shrimpster: Yeah of course struct ^^ I dont use inputs

jacek: :flag_pl:

Astrobytes: MeBen: Polish and Greek

struct: Then you probably should be solving sooner

Shrimpster: Yeah Im seeing it now... youre right

struct: https://www.codingame.com/replay/586185958

struct: here you are solving 2 plies

jacek: so many wroclaw guys submitting to uttt right now oO

Astrobytes: is there something special about that Uni jacek?

jacek: dunno. i think their teacher assign exercises on CG

struct: lol

Shrimpster: wroclaw? wheres that?

struct: poland

Shrimpster: Lul saw the gold number go up about 20 ppl since last week I started, thats probably them :D

jacek: here you can see i solved it several plies before the end https://www.codingame.com/share-replay/586186311 despite the lower 'rollouts' number

TheBlueBias: Oya there. We're back.

Shrimpster: Yeah youre correct :/ I'll find the error

Astrobytes: Back from where Blaise?

Shrimpster: Thanks a lot kind people for the tips and youre time! :)

jacek: to uttt discussions

Astrobytes: lol

TheBlueBias: reply challenge

Astrobytes: Ahaaa

Astrobytes: How did it go?

jacek: huh?

Astrobytes: jacek: A CTF

TheBlueBias: Pretty hard, and big gap with the top. But was fun, and we're 20th.

struct: I think it still solves correctly but seems a bit late

Astrobytes: nice, well done to you all :)

struct: my bot with 1/10 of the time

struct: https://www.codingame.com/replay/586187130

Shrimpster: Yeah thats 3 moves later...

eulerscheZahl: Reply is over, was fun

eulerscheZahl: and i could recycle me code from https://www.codingame.com/training/hard/doubly-solved-rubiks-cube

eulerscheZahl: my code*

Astrobytes: nice

eulerscheZahl: https://www.youtube.com/watch?v=63d287LxZi8&ab_channel=ReplyU 20:26 we are on stream

TheBlueBias: Let's celebrate! https://www.youtube.com/watch?v=-Rq29YX_-E8

eulerscheZahl: even after teaming up with you I don't trust your links

TheBlueBias: I'm not jacek...

eulerscheZahl: you share music videos

eulerscheZahl: and then youtube is suggesting weird stuff for a few days

TheBlueBias: :rofl:

TheBlueBias: Well. That's not a music video.

Astrobytes: lol, on stream for a second lol

eulerscheZahl: it's something

Astrobytes: hahahaha ffs Blaise :rofl:

Astrobytes: true euler :)

jacek: no one wants to be jacek :(

Astrobytes: wonder why :thinking:

eulerscheZahl: do you want to be a jacek?

Westicles: huh?

struct: I want to be a jacek

Westicles: oO

niluthebear: I want to be a derjack

jacek: :upside_down:

AstroHydes: I wanna be an Astrobytes

jacek: youre too quick

Default avatar.png breftejk: i want to be jacek

Default avatar.png bminnix: :thinking: I like being me

lordgun: please help me with my first contribution :) https://www.codingame.com/contribute/view/74789a032044db8a92a9c34967b1a6ab2285

MACKEYTH: In Spring Challenge 2020 do lines of visibility go through edge warps?

Westicles: there are a couple camelcase ones already

VizGhar: well... for contributors it is hardly possible to know that there are already such CoCs :)

Westicles: https://eulerschezahl.herokuapp.com/codingame/puzzles/?q=camelcase&category=ANY&title=on&statement=on&tests=on&comments=on&tags=on&author=on

niluthebear: i love to see some algorithmic problems

niluthebear: i don't like strings

Westicles: I'm not sure how to make a good reverse. Should almost everyone instantly see the trick, or should half the people never see it?

struct: test cases should make it obvious imo

VizGhar: agreed, but not first 1-2 testcases... hide some trick in 4th/5th

Shrimpster: Classic puzzle creator mindset by VizGhar :D

Westicles: It is tricky. I thought this one is straightforward, but maybe not

Westicles: https://www.codingame.com/contribute/view/7437a2695e9623773636585c212713d5ad0e

VizGhar: nice

jacek: i liked the one when it had something like 1+4=?, and 3*?=100, and the answer was the length of the equation

struct: lol

niluthebear: lol

VizGhar: Hmm what is that last achievement in Puzzles section?

struct: solve nintendo

VizGhar: I mean "Participate 3 times in the ranking of puzzle solutions"

Shrimpster: Being an active member in the community :)

VizGhar: After I solve puzzle, i can give upvote to puzzle. But it's not it right?

Shrimpster: If you enjoyed it, I guess, why not

Uljahn: when you go to community solutions you should be prompted two solutions to choose from

Shrimpster: and leave a comment

Westicles: people don't seem to often upvote accepted contribs, but it is nice to do if you like it

TimothyAlexisVass: Perl $s = '1211';

TimothyAlexisVass: @a=$s=~/(\d)\1*/g

TimothyAlexisVass: print Dumper @a

TimothyAlexisVass: $VAR1 = '1'; $VAR2 = '2'; $VAR3 = '1';

TimothyAlexisVass: ???

Uljahn: :joy:

TimothyAlexisVass: Why isn't it $VAR1 = '1'; $VAR2 = '22'; $VAR3 = '1';

TimothyAlexisVass: I mean $VAR1 = '1'; $VAR2 = '2'; $VAR3 = '11';

Uljahn: hey, no smap

Uljahn: no spam

Uljahn: sry

Uljahn: Automaton2000

Automaton2000: should i change my code to do that

VizGhar: Community solutions? I gave multiple thumbs up on puzzle solutions after solving them. But this is for pending contributions only?

struct: which achievement is it?

VizGhar: Budding examiner Participate 3 times in the ranking of puzzle solutions.

struct: ah maybe rate solutions of other users?

TimothyAlexisVass: Why does this happen: http://chat.codingame.com/pastebin/1016ef46-94c9-4151-8896-7d61d54a9ecf

VizGhar: I though only puzzle solution that can be ranked is the solution you can see after you solve the puzzle :)

VizGhar: nope. doesn't work

struct: maybe other users need to rate yours?

struct: idk

struct: but doesnt make sense

Westicles: oh, for the achievement you have to do the ranking thing. look on the tab that says 100% when you solve

Default avatar.png bminnix: Have any of you folks ever had your Stream output match exactly what the expectation was but the debugger found "nothing"?

Default avatar.png bminnix: I'm new to this, so it's likely my fault, but was just curious if that ever happens.

jacek: additional new line?

Default avatar.png bminnix: Accoding to my Stream output the first line was a > "value"

2ReGul4r: Hey can someone help me? My english isnt the best thats why I wanted to ask if the statement is clear enough ? :D https://www.codingame.com/contribute/view/748273d5e361c5d195cd23134a158e48085c

VizGhar: Thanks Westicles... such hidden achievement o.O

VizGhar: Why do they want me to rank others? It makes more sense to me, to give achievement for publishing solution.

TheBlueBias: Nobody said it was supposed to had any sense... :smirk:

VizGhar: meh... I don't have to like everything :D keep it up guys

VizGhar: good night

Default avatar.png SergioDev: hii

Default avatar.png jawadcode: hello

Default avatar.png GuillaumeMohimont: https://escape.codingame.com/game-session/1BE-7VY-xs4-bys

Default avatar.png GuillaumeMohimont: escape game

Default avatar.png GuillaumeMohimont: want to play

Wontonimo: shhhh ... it's quiet time here

TechnoBot: HI

Default avatar.png MeBen: shhhhhh

TechnoBot: what?

TechnoBot: we can whisper*

TechnoBot: * = whispering

Default avatar.png MeBen: okay*

Westicles: :hammer:

TechnoBot: brb*

TechnoBot: don't hammer us*

**TechnoBot whispers: "Hammer is not good."

TechnoBot: https://escape.codingame.com/game-session/pPJ-64l-Mzt-fHF/in-game

TechnoBot: pls join****

TechnoBot: brb

TechnoBot: ok no join*

Westicles: 3D Dungeon... meh. 5D is better

TechnoBot: shhhhhhhhhhhhhhhhhhh*

TechnoBot: u should play hypixel.net (Minecraft server).***

Westicles: That old microsoft game for little kids?

TechnoBot: No.

TechnoBot: Minecraft is very popular brb

TechnoBot: Youtubers also play it

Westicles: little kids used to play it, for sure

TechnoBot: like me cause im Technoblade

TechnoBot: BUT IM NOT A KID

TechnoBot: kids and adults all play it

TechnoBot: BRB

Wontonimo: hey technoblade. do you know about craftmine or whatever it is called?

**TechnoBot slaps Westicles around a bit with a large fishbot

TechnoBot: hmmm?

TechnoBot: What Wontonimo

TechnoBot: ???????????????????/

TechnoBot: thats a rip-off

Wontonimo: what is minecraft?

TechnoBot: of minecraft

Wontonimo: is that where people play dungeons and dragons and pray to weird gods?

TechnoBot: Minecraft is this popular game

TechnoBot: no

Westicles: I think it comes packaged with windows like solitaire

Wontonimo: is it a cosplay thing?

TechnoBot: no

TechnoBot: minecraft is not free

Wontonimo: oh

Wontonimo: well that explains it

TechnoBot: it is a very cool game

Wontonimo: gotta go

TechnoBot: made by Mojang

TechnoBot: BRB

TechnoBot: #WorldChat2

TechnoBot: oops

TechnoBot: srry no # supposed to be there

**TechnoBot says: "#WorldChat2 is supposed to not be there."

TechnoBot: brb

**TechnoBot is AFK.

**TechnoBot is now NOT AFK.

TechnoBot: wussup homies?

TechnoBot: brb

**Wezl tries

**TechnoBot is now online again.

TechnoBot: brb

TechnoBot: Hello.

TechnoBot: Hey Wezl!

TechnoBot: brb

TechnoBot: y leave

TechnoBot: bruh