Chat:World/2021-08-08

From CG community
Jump to navigation Jump to search

sugam: why did i get 75%

sugam: does that mean 75% of the test cases were correct?

hypers: yes

hypers: share the code

hypers: we can spot the problem

sugam: look

sugam: ohhh

sugam: I looped through all of it

sugam: while I assigned i as int, and the n was long. I got an error for long values

hypers: probably

jacek: oh my

sugam: guys do you have any tips for me to get remote internships? Companies at my countries are really bad and I think getting a remote internship Eu/US would be better for my growth.

jacek: youre magus reversed :o

sugam: ohh

MACKEYTH: Anyone familiar with the Fall 2020 challenge referee?

eulerscheZahl: the engine was simple enough to play the game without referee

eulerscheZahl: can't you just skip the introduction and come to the real question?

MACKEYTH: What's the behavior when casting a spell if you don't have inventory room for all the new ingredients?

dbdr: hey eulerscheZahl! where did you get so many CPs? :D

eulerscheZahl: hm.. CG Funge, Search Race, Dots & Boxes

jacek: i can troll again :tada:

eulerscheZahl: hello chat

eulerscheZahl: I had one more line for dbdr that didn't get through as the chat suddenly died

eulerscheZahl: and all of my submits at connect4 finish at rank 35 :/

eulerscheZahl: 25*

eulerscheZahl: convert from C# to C++: 25 => 25

eulerscheZahl: remember to use pragmas => 25

eulerscheZahl: add STEAL => 25

eulerscheZahl: don't reuse own tree on opponent STEAL (that I don't sim) => 25

eulerscheZahl: (╯°□°)╯︵ ┻━┻

martinpapa69: I spent 5 days to implement a tflite parser, and runner for cg

martinpapa69: to skyrocket my bot from rank5 to rank 25

eulerscheZahl: oh :D

eulerscheZahl: must be another game, my 25 is engraved in stone

martinpapa69: oware

Astrobytes: lol, time well spent eh :D

BlaiseEbuth: Howdy there!

martinpapa69: but...i still hope this gonna work

eulerscheZahl: hail our mod

BlaiseEbuth: Fear my badge.

Astrobytes: :D

eulerscheZahl: is MK back to mod as well?

BlaiseEbuth: Cg don't want me as mod, but the char disagree.

BlaiseEbuth: *chat

eulerscheZahl: did you know that chat and char differ by just one chat?

eulerscheZahl: *char

BlaiseEbuth: :3

jacek: so are you mod or not

eulerscheZahl: the answer is "yes"

BlaiseEbuth: I am \o/

eulerscheZahl: :D

eulerscheZahl: rip jacek :(

Astrobytes: :rofl:

jacek: nyoro~n

Default avatar.png IISadNightII: Good Morning

jacek: ohai

Default avatar.png BUCKFAE: Is Skynet 1 only broken for me? I pass all tests in the IDE but get 0% in the final evaluation

Default avatar.png Ace1315: nyanharo

darkhorse64: eulerscheZahl: for C4, use smart rollouts (avoid losing moves, play winning moves).

BlaiseEbuth: eulerscheZahl: If you want to win, avoid losing.

darkhorse64: stop plagiarism, please

martinpapa69: can you link a paper about this algo? sounds interesting

BlaiseEbuth: Don't mess with a moderator darkhorse64...

darkhorse64: On the Huge Benefit of Decisive Moves in Monte-Carlo Tree Search Algorithms

darkhorse64: I can kick moderators butt provided I stay polite

darkhorse64: and I act gently

BlaiseEbuth: blackponyx86

darkhorse64: 32 bits is so 20th century

BlaiseEbuth: Aren't you? :p

jacek: mhm

darkhorse64: don't mess with veterans

BlaiseEbuth: :older_man:

jacek: checking for win/lose moves... of course theres tradeoff. smarter rollouts, (a bit?) slower but better. must have in yavalath, meh in othello

jacek: i bet euler has it :?

darkhorse64: Yep, it's not a winning recipe but it works in C4. I based my recommandation on his C4 C# snippet

TobiasA: are any of you in legend in tron?

BlaiseEbuth: Yeah

darkhorse64: I do

darkhorse64: but far from the top

TobiasA: did you use minimax?

BlaiseEbuth: yup

TobiasA: did you make every other player have negative score when simulating their moves?

BlaiseEbuth: Yes.

BlaiseEbuth: 3 min, 1 max

BlaiseEbuth: when 4 players

TobiasA: and it worked fine?

BlaiseEbuth: Well, my bot is 62th

TobiasA: did it get you from gold to legend?

TobiasA: or you used something else?

BlaiseEbuth: I tunned it after the legend promotion, but that was already what I used in in gold.

TobiasA: hm

TobiasA: mine works but makes random dumb decisions sometimes

PatrickMcGinnisII: I survived Korn concert last night. I think I lvld up in RL. ;)

TobiasA: and i don't know what the problem is

BlaiseEbuth: This kind of algo depends a lot of your evaluation function...

jrke: flood fill eval works great in tron

TobiasA: that's what i used

TobiasA: the bot is stuck in its current rank no matter how much i try to make it better

Default avatar.png Xaniouks: So close but yet so far in solving the puzzle of the week :sweat_smile:

jrke: TobiasA your flood fill eval looks buggy i think

jrke: https://www.codingame.com/replay/574543436

jrke: turn 162

jrke: bad move at turn 162

BlaiseEbuth: ^

TobiasA: yea that's what i was talking about sometimes it works just fine then sometimes it just makes the most trash decision i have ever seen

TobiasA: i think it was trying to block the red guy

jrke: but see he is already blocked and blocking him more will lead you to block

jrke: and i also think that move at turn 19 is also bad

TobiasA: how do you think i should score the bots especially with more than two players because that's when it freak out the most?

jrke: wait i have one link lemme see if i can find

BlaiseEbuth: I have a flood fill, and it works fine. You should verify there's no error in your implementation.

jrke: i have flood fill also but for better explanantion understand this

jrke: https://vks.ai/2016-09-07-ai-challenge-in-78-lines

jrke: use this as your eval

TobiasA: thanks

jrke: thats flood fill though

jrke: see the `visual aid` part

TobiasA: yea that's exactly what i did

jrke: but it looks buggy

jrke: the replay i shared this - https://www.codingame.com/replay/574543436 you decreased your blocks you can visit on turn 19 as well as 162

WannaBeHum1n: guys

WannaBeHum1n: i ran out of time in a COC

WannaBeHum1n: but i wanna finsih it

WannaBeHum1n: finish it

WannaBeHum1n: how to do so ?

WannaBeHum1n: :sweat_smile:

WannaBeHum1n: notice me senpais

BlaiseEbuth: You can't.

WannaBeHum1n: cant i search for the pronlem

BlaiseEbuth: You can try to find it back here https://eulerschezahl.herokuapp.com/codingame/puzzles/

WannaBeHum1n: and do iit outside the coc

WannaBeHum1n: it was a reverse code

WannaBeHum1n: do i find them here

WannaBeHum1n: ?

BlaiseEbuth: If you know what to search

WannaBeHum1n: found it

WannaBeHum1n: thnx forr help

WannaBeHum1n: <3

BlaiseEbuth: :kissing_heart:

Default avatar.png BUCKFAE: What do i do if a puzzle is broken?

BlaiseEbuth: Fix it

Default avatar.png BUCKFAE: It's not my puzzle, how would i do that

Default avatar.png BUCKFAE: I get all tests correct in the ide yet my score is 0%

Default avatar.png BUCKFAE: when i submit

jrke: validators!!

BlaiseEbuth: Well, since the error is probably in your code, you totally can fix it.

TobiasA: maybe you hard coded:astonished:

Default avatar.png BUCKFAE: I did not hardcode anything yet the validators don't work

BlaiseEbuth: What puzzle ?

Default avatar.png BUCKFAE: Skynet revolution - episode 1

Default avatar.png BUCKFAE: https://paste.ofcode.org/ZNcZmr9NDpexm3v5suZZS3

Default avatar.png BUCKFAE: I know my bfs is bad, don't judge lol

Default avatar.png BUCKFAE: But this is not hard-coded

BlaiseEbuth: Thanks for reporting, I gonna send a message to the 37296 users who solved it, to tell them they can't have solved it since the validators don't work.

Default avatar.png BUCKFAE: There is a chance that the validators broke, or am i wrong?

BlaiseEbuth: No chance.

Default avatar.png BUCKFAE: The validators are even the same thing as in the ide and they work in the ide (at least from what i can tell)

jacek: welp it works for me

jrke: its showing timeout in validators

jacek: but yeah, using your solution it doesnt work in validations oO

BlaiseEbuth: Well. Somebody have a python solution that already got 100% ?

Default avatar.png BUCKFAE: Could it be the while True at the end? If the validators don't stop like the ide does i will loop forever there

TobiasA: timeouts are the worst i prefer long irritating c++ errors than timeouts

jacek: welp simple test in test and validator is the same

BlaiseEbuth: They aren't

BlaiseEbuth: Node ids are differents

jacek: anyway something's weird. they fail right away

BlaiseEbuth: Hard coding the first validator in python works for me.

BlaiseEbuth: So the problem really come from your code

Counterbalance: the s = sorted(..) looks problematic. It returns that, and then edges.remove() is called on that. If the elements are swapped, the edge won't be found. I think.

Default avatar.png BUCKFAE: oh crap, that's probably it

Default avatar.png BUCKFAE: i was assuming the edges are always sorted descending

Default avatar.png BUCKFAE: Thank you, that solved the issue. I added a test if the edge exists, otherwise i swap the tuple

jacek: oO

Counterbalance: nice!

Neumann: new CGBenchmark release ! https://forum.codingame.com/t/cgbenchmark-tool/2956/40

Default avatar.png Adu01: Guys can you offer me some help?

Default avatar.png Adu01: I'm struggling at The descent

Default avatar.png Adu01: I just don't know how to make the spaceship target the highest mounatain

Default avatar.png Adu01: Mountain*

Default avatar.png Shivay: where is trivia bot creating tutorial

AntiSquid: a what?

Default avatar.png Shashank28: hi

eulerscheZahl: something like this shivay? https://www.codingame.com/playgrounds/40701/help-center/ghost-in-the-cell---guide

jrke: i have one bookmarked link

jrke: https://www.codingame.com/playgrounds/25775/codingame-sdk-documentation/create-a-multiplayer-game

4DsErfan: yo

michaellan: yo

4DsErfan: have you ever heard anything ordak

4DsErfan: about ordak*

eulerscheZahl: i wonder if i should avoid generating all child nodes at once before expanding them one after another (or not at all) at later times

eulerscheZahl: it happens that I generate 3M nodes (and valid moves) but only ever use 450k

eulerscheZahl: any experience from other players?

Scarfield: which multi, and mcts? (no exp fiddling with it though, i always expand fully=

jrke: what doe inline do btw?

jrke: in c++

martinpapa69: http://chat.codingame.com/pastebin/413ec14a-7598-4d2a-9d4a-2c5f1558535c

darkhorse64: When expanding, I do a rollout for each child. Ofc, this is very bad for games with huge branching. When combined with a solver, smart rollouts, it simplifies code a bit

MaliciouslyCrypticUsername: My bot went from 500 to 2.4k in CSB gold :( after I changed it

Scarfield: i guess either rollout more than once (if you expand fully), to get an early approximation if the path is worth revisiting. Or only expanding one child and rollout once

Scarfield: oh as the dark horse said :)

BlaiseEbuth: Pale horses are better.

MaliciouslyCrypticUsername: ;o

MaliciouslyCrypticUsername: Purple horses are best

MaliciouslyCrypticUsername: Automaton2000 what do you think

Automaton2000: that's what i see in the ide

MaliciouslyCrypticUsername: Exactly

TobiasA: is Automaton2000 a bot?:neutral_face:

Automaton2000: how do you know what you're doing

MaliciouslyCrypticUsername: Automaton2000

Automaton2000: and see if i can get a new one

MaliciouslyCrypticUsername: AutomatonNN

jrke: yeah he is a chatbot

MaliciouslyCrypticUsername: ik

jrke: it*

MaliciouslyCrypticUsername: What happened to NN?

TobiasA: reeeeaaaally?

Astrobytes: Gone thankfully.

MaliciouslyCrypticUsername: .-.

MaliciouslyCrypticUsername: Rip

BlaiseEbuth: Not much purple horses here. Not enough brains to eat I suppose...

MaliciouslyCrypticUsername: Not enough brains on CG?

MaliciouslyCrypticUsername: Not enough brains anywhere then ._.

TobiasA: how do i talk to the bot

MaliciouslyCrypticUsername: Just mention name

MaliciouslyCrypticUsername: Automaton2000

Automaton2000: does anyone has a good bot

TobiasA: Automaton2000

Automaton2000: or at least that's what i did.

Astrobytes: Or, preferably, don't.

MaliciouslyCrypticUsername: ^

TobiasA: hm i wonder why i didn't notice it was a bot all along

MaliciouslyCrypticUsername: "Magus XMPP chat bot" it says

TobiasA: well i never looked at that till now:rolling_eyes:

BlaiseEbuth: Let him do. I want to enjoy my mod privileges while I can. :smiling_imp:

TobiasA: my fault

TobiasA: i don't want to get banned:sleepy:

eulerscheZahl: Scarfield 06:54PM which multi, and mcts? (no exp fiddling with it though, i always expand fully=

connect4 MCTS

eulerscheZahl: "When expanding, I do a rollout for each child." but that requires further child expansion down the tree. doesn't this result in a fully evaluated game tree?

eulerscheZahl: "Gone thankfully." I won't miss AutomatonNN either :)

TobiasA: why?

BlaiseEbuth: eulerscheZahl is there a way you miss AutomatonNN one day?

jrke: finally i gathered my will to rewrite my CSB bot

jrke: i will be trying smitsimax this time maybe

TobiasA: what is your rank?

jrke: TobiasA my rank in CSB?

TobiasA: yup

jrke: 43 currently

eulerscheZahl: if it gets revived and the next loss of it means that the whole CodinGame website died. then yes, I would miss AutomatonNN (or at least the other things I lost with it)

TobiasA: damn

eulerscheZahl: is that just the inoryy bot with a few lines of eval?

jrke: nah

jrke: my own but its quite messy code now

TobiasA: what algorithm did you use at first jrke?

4DsErfan: where's the ordak

eulerscheZahl: i tried hard on mean max. and a bit on poker chip race. that's enough physics for me.

ShakeDatBunda: how do I approve or reject a contribution?

jrke: simulation with that seed one shifting a move for each turn

eulerscheZahl: click the button that says approve

jrke: if its not WIP

eulerscheZahl: then click the checkboxes for the moderation questions (after reading them)

eulerscheZahl: you need 50 clashes for a clash contribution and level 20+ for the rest

Astrobytes: jrke: I think he meant which algo did you use for CSB at first

jrke: at first means the first time or the current submitted one?

Astrobytes: When you started using a search bot for CSB

TobiasA: before deciding to rewrite

eulerscheZahl: the first thing I did at CSB was setting the speed to max (100)

eulerscheZahl: then I promoted

eulerscheZahl: oh and going to (x, y) instead of (y, y)

Astrobytes: *search bot

eulerscheZahl: then the angle part came into play and I left the game for some months

jrke: for first search bot i used bugged engine with MC random search

eulerscheZahl: didn't see a proper definition of what the angle even means

jrke: then i removed bugs and optimizing it

TobiasA: did you guys simulate opponent moves

TobiasA: ?

jrke: yes

TobiasA: eyeee seeeeeeeeeeeeee

eulerscheZahl: your 'e' key looks stuck

BlaiseEbuth: eyeeeeeeeeeeeeeeee up

TobiasA: eye don't think it is stuck:wink:

jacek: :thinking:

Default avatar.png Roham.N24: hello

TobiasA: hoi

Default avatar.png Badre2dine: how to earn coding point ?

eulerscheZahl: https://www.codingame.com/multiplayer anything on this page gives you CP

eulerscheZahl: amount of possible points varies depending on the game http://apps.rezonux.com/codingame/?user=a3fd116ae61d697167cf0c340c95f4e52542244

michaellan: any tips on how to get 40k points on CVZ?

TobiasA: what's CVZ?

HungryBurger: Code Vs Zombies

Default avatar.png Roham.N24:

YurkovAS: xxd https://www.codingame.com/forum/t/code-vs-zombies-feedback-strategies/1089

eulerscheZahl: 40k? well, almost :P https://www.codingame.com/share-replay/574678277

michaellan: Yeah, I haven't tried to perform combos, maybe that will do it

michaellan: WOW

eulerscheZahl: that's pretty bad actually

michaellan: i have no clue how you code it to build up for a combo like that :/

eulerscheZahl: it's just a random search

eulerscheZahl: random moves until either all zombies or all humans are dead

eulerscheZahl: try over and over again, keep the best

michaellan: like GA?

TobiasA: MC

eulerscheZahl: mine is just a monte carlo. fully random each time

eulerscheZahl: but there are better ways

michaellan: how does the rollout work though

eulerscheZahl: read the link that YurkovAS provided

michaellan: gotcha

eulerscheZahl: in the end you have to reimplement how the game works

michaellan: like the zombie aggro right

eulerscheZahl: ash moves, zombies react. ash moves again and so on

martinpapa69: why is my locally compliled code like 30% faster, than the one i let CG compile it :( i have a lot of pragmas

Default avatar.png blazejek123: t.if

reCurse: Different hardware

martinpapa69: cg hardware. i complile the code, encode it and send it. then save and run

reCurse: If you're using native architecture pragma it's possible it enables more optimizations for your machine that wouldn't work on cg

martinpapa69: both my codes run on cg server

martinpapa69: so...its working on cg server

reCurse: Oh for some reason I understood the other way around

reCurse: Not sure why

martinpapa69: one thing i noticed my asserts even pop when i set #pragma GCC optimize("O3")

martinpapa69: so it looks like a fake o3 to me

michaellan: it is okay for me to close the popup when i click "Test in Arena" right? it will still run all the necessary battles behind the scenes?

martinpapa69: yes

Westicles: Different versions of g++? Or maybe locally you are using clang++. that is usually faster?

reCurse: If asserts pop then it looks like you didn't undef _GLIBCXX_DEBUG

martinpapa69: hm thx, I thought o3 would remove them

reCurse: Pragma doesn't change the command line

reCurse: It only changes what it still can

reCurse: Compiler is still invoked with O0 so it does some stuff like define macros

martinpapa69: hm I just tried, Ofast, its actually much...faster

reCurse: It enables a lot of unsafe behavior so be careful

Astrobytes: Yeah, sometimes it's worth just turning on the parts you need individually rather than the full Ofast

martinpapa69: still better than precompiling

reCurse: shrug

3283415: \o/

reCurse: mcts is so fickle it drives me nuts

michaellan: is it easier to code geneetic algorithm or MC for CVZ?

martinpapa69: mc. its like 10 lines max

michaellan: joking?

Astrobytes: Random move, update state, increase depth. Score as you go with a depth malus or just score at the end.

Astrobytes: As long as your sim is correct you should get relatively reasonable results

Astrobytes: Oh it's CvZ, you should be fine just going with end score

martinpapa69: ok i lied

martinpapa69: 20 lines

martinpapa69: http://chat.codingame.com/pastebin/076f6a1e-067b-4577-8b51-2e5e866f02f4

BlaiseEbuth: Solution sharing...

Astrobytes: hardly

michaellan: whats malus

Astrobytes: opposite of bonus

martinpapa69: its probably top10 worthy solution

michaellan: the s.simulate(m) part is the hard part im thinking

BlaiseEbuth: Do it again and that's a ban.

Astrobytes: Latin iirc

KiwiTae: hooo BlaiseEbuth got his mojo back o/

BlaiseEbuth: \o/

martinpapa69: oh thought you were trolling

reCurse: It's a bug when chat server crashes

xxd: wait.. why ban

BlaiseEbuth: reCurse called me a bug...

TobiasA: where is 0, 0 coordinate in tron?

KiwiTae: :joy:

BlaiseEbuth: Thanksfully, since chat crash once a week I'll be mod forever! :imp:

BlaiseEbuth: And banned get unbanned... That's like a mythological eternal fight!

reCurse: yawn

Westicles: paging rockstar555

Astrobytes: double yawn

BlaiseEbuth: I see you still haven't tunned your humor coefficients...

reCurse: There was a joke?

darkhorse64: rockstar555 is a joke

BlaiseEbuth: You aren't sure? You see: your humor detection doesn't work.

reCurse: Or your humor is off-balance

Astrobytes: darkhorse64: A tired one at that

darkhorse64: tired = élimée aka usée jusqu'à la corde?

BlaiseEbuth: Shit, I was betting on a ':rolling_eyes:', I lost...

Astrobytes: yes darkhorse64

Westicles: He's been pretty nicely rehabilitated. On a few months now without trouble, lots of contribs approved

Astrobytes: I just really don't care any more

Astrobytes: With respect to that

Astrobytes: I have larger fish to fry in the grand scheme of things

Westicles: Yeah, I hear that. People everywhere scheming to get papers so they can buy groceries

michaellan: can someone spoil what the achievements are on the optimization quest path

BlaiseEbuth: Optimization ones.

darkhorse64: Lots of hard work

michaellan: alrighty

Astrobytes: You gotta put in the effort

Westicles: top 2% golf and optim

FalINTOblivion0112: niceee

FalINTOblivion0112: im top 1500 in clash of code now

FalINTOblivion0112: gotta reach the top

michaellan: gratz

FalINTOblivion0112: thank you!

BlaiseEbuth: Do a clashbot, ezpz.

FalINTOblivion0112: everyone is so encouraging, i like this environment

michaellan: is it viable to write MCTS in python or should i switch to C++

ZarthaxX: it will be very slow in python

Astrobytes: For testing purposes sure, but not for a performant bot

ZarthaxX: you wont get good results in the long run

ZarthaxX: well, what astro said :)

michaellan: thanks guys. my prorotype was in python but now i'll try re-writing it in C++

Astrobytes: ZarthaxXamang how are ya dude

ZarthaxX: good

michaellan: also, one quick question about backpropagation

ZarthaxX: but just do a simpel version xxd, without optimizing

ZarthaxX: remember, make it work first, then optimize

michaellan: in many of the tutorials I see that wins are backpropagated like +1 to all parents

ZarthaxX: and always check after each optimization that you didnt break anything

michaellan: but wont this bias the MCTS towards choosing dumb moves for the opponent? what I did was I did +1 if the winner of the rollout == the state's players move

BlaiseEbuth: Well, going to c++ from python is already a huge optimization...

TobiasA: when you guys use bitboards in c++ do you use bitset or some other method?

michaellan: so it basically only adds +1 at skipping every other level

michaellan: effectively

darkhorse64: :scream:

ZarthaxX: i dont get your question xxd

ZarthaxX: BlaiseEbuth ye :P

ZarthaxX: TobiasA use a 64bits variable

Astrobytes: uint64_t

ZarthaxX: bitset is good for debuggintg tho

darkhorse64: no bitsets, please

TobiasA: :cry:

michaellan: Here is my backpropagation code http://ix.io/3vp3

ZarthaxX: but dont use it :P

michaellan: is this wrong?

Astrobytes: true ZarthaxX, and for learning I guess(?) - I rather go all in

KiwiTae: ZarthaxX :*

KiwiTae: o/ Astrobytes

TobiasA: i should have asked earlier

TobiasA: i have to start everything again

michaellan: mostly concerned about the part where I only add 1 if the winner of the rollout is actually the player moving

ZarthaxX: you hardcoded the 1 xxd

michaellan: oh is that wrong too

ZarthaxX: it could be 0 if its a draw, but forget about it :P

michaellan: yeah w is None on a draw :)

michaellan: thanks for mentioning that

ZarthaxX: Astrobytes oh sorry i didnt answer to your prev comment D: hello astromoditus

ZarthaxX: im doing fin,e got an email that finale xam for wed is cancellled

ZarthaxX: so not good

ZarthaxX: :(

ZarthaxX: i will have to do it on september

ZarthaxX: xxd ah ok

ZarthaxX: hey kiwooo

michaellan: My main question was whether the above code is correct or this: http://ix.io/3vp4

ZarthaxX: i gotta go there areis a storm here

michaellan: the difference is in how score is altered at each parent

BlaiseEbuth: You'll be OK for the escapegame contest in october

ZarthaxX: f

TobiasA: wait so the fall contest isn't bot programming?

Westicles: Anyone made a run at Snake optim? That looks like it will be fun

ZarthaxX: second one is wrong xxd

KiwiTae: Westicles im starting

ZarthaxX: you want to do win = -win

ZarthaxX: at each iteration

michaellan: whew

michaellan: thanks for confirming

ZarthaxX: so the next parent gets the opposite

darkhorse64: when backpropagating the result of a rollout, +1 if it's a win for the player that plays the move in the current state, -1 otherwise

ZarthaxX: it's basically what you are doing in the other code

michaellan: so basically the strongest opponent moves are selected

ZarthaxX: but without ifs

ZarthaxX: wdym?

michaellan: thanks thanks thanks

michaellan: i mean like if you just add 1 at every stage, you will select for the "weakest" opponent moves instead of the strongest

ZarthaxX: you are just adding on each node the game result based on that node perspective

michaellan: i think

michaellan: yeah i get it now intuitively, thanks

BlaiseEbuth: Tank tank tank - Guderian

ZarthaxX: you are just colleting data when doing mcts, like some stadistic method

michaellan: sadistic ;)

ZarthaxX: lmao

LelouchVC2: Its always a nice surprise to see that I've been unbanned from world chat

reCurse: Thanks for pointing out, fixed

BlaiseEbuth: Monster...

Astrobytes: That one STILL hangs around?

michaellan: damn, what did he do

BlaiseEbuth: Waiting for any chat crash, ready to come back!

BlaiseEbuth: He never wait long though

Astrobytes: ZarthaxX sorry bout the exam, take care in the storm dude

ZarthaxX: Astrobytes thanks, i hate that subject exam :(

ZarthaxX: didnt leave yet as the storm is not so bad yet haha

Astrobytes: ah cool

Astrobytes: What subject?

ZarthaxX: language paradigms

Astrobytes: KiwiTae: hey hey :)

BlaiseEbuth: Don't go to run outside with your arms raised, and that's gonna be fine.

Westicles: xxd, looks like he made a south park reference

ZarthaxX: i wanted to do it asap to be free tho, so it's not good news

Astrobytes: ZarthaxX: BO-RING :)

ZarthaxX: but the maximum amount of people for the exam was 4

ZarthaxX: pretty lame

Astrobytes: Ah damn.

ZarthaxX: ;(

ZarthaxX: will give another shot on september :P

ZarthaxX: and yes, super boring

ZarthaxX: on the other hand, there is going to be a subject called game theory this semester

ZarthaxX: looking up for it

Astrobytes: You'll pass no problem when you do it. Oho game theory eh :)

ZarthaxX: i hope i do, trying to do the exam with the easy teacher...

ZarthaxX: there is like the ying yang teacher here lmao

ZarthaxX: the other dude just destroys you

Astrobytes: Same everywhere man

ZarthaxX: i know but this one iseven worse

KiwiTae: ZarthaxX patpat

TobiasA: i can't find a good place to learn the 64 bit variable thing

ZarthaxX: patpat? :P

ZarthaxX: TobiasA wdym?

Astrobytes: He's patting you on the head to comfort you ZarthaxX

BlaiseEbuth: Emperor patpatin

TobiasA: like a website

ZarthaxX: you just use the bits of the binary representation of an integer variable as a bitset

PatrickMcGinnisII: uh oh

Astrobytes: TobiasA: are you familiar with bit manipulation?

TobiasA: yea

ZarthaxX: TobiasA maybe you will enjoy this one https://core.ac.uk/download/pdf/33500946.pdf

Astrobytes: So there is nothing different, you are just operating on a 64 bit type is all

TobiasA: i just came from there :grin:

BlaiseEbuth: That's not a website! That's a f**** pdf!

BlaiseEbuth: (╯°□°)╯︵ ┻━┻

ZarthaxX: Astrobytes oh okey :rofl:

ZarthaxX: thanks kiwo :*

ZarthaxX: TobiasA huh, didnt work out for you?

ZarthaxX: also this https://www.chessprogramming.org/General_Setwise_Operations

BlaiseEbuth: That's a website

Westicles: Pat, you have to fight for your honor

ZarthaxX: lmao BlaiseEbuth

ZarthaxX: TobiasA search for bitboard, maybe that yields better results

ZarthaxX: https://www.chessprogramming.org/Bitboards

BlaiseEbuth: The emperor's patpating anakin

TobiasA: i know how to use bitboards but i don't know how to use the 64 bit variable one

TobiasA: i only know how to use bitset

ZarthaxX: so

BlaiseEbuth: So you don't know

ZarthaxX: you just need to learn how to make the bitset operations witha variable

ZarthaxX: integer one

TobiasA: yea where do i learn it?

PatrickMcGinnisII: http://chat.codingame.com/pastebin/ff945d51-f523-414e-8fb2-aafbbd20c0c8

Westicles: reCurse: There are only two things I can't stand in this world. People who are intolerant of other people's code. And PHP programmers.

ZarthaxX: for example, to set a bit to 1 with bitset: value[i] = 1 with variable: value = value | (1 << i)

Astrobytes: You pulled that one up quick

BlaiseEbuth: Don't try humor, he gonna ban you Westicles! :scream:

ZarthaxX: TobiasA you can just look at the bitwise operations available

TobiasA: how do i set the size?

ZarthaxX: you just need or, xor, and and negation for now

Astrobytes: value |= (1 << i)

ZarthaxX: you cant

TobiasA: :0

ZarthaxX: Astrobytes didnt want to confuse him

ZarthaxX: :P

reCurse: Blaise it's getting old

Astrobytes: UglycodethaxX

ZarthaxX: TobiasA you operate with a variable that has a fixed amount of bits

TobiasA: what if i want 600 bits

BlaiseEbuth: Nope, that's just always true... Sadly

ZarthaxX: it can be 8, 16, 32 or 64 bits

TobiasA: i have to write 600 zero's

ZarthaxX: TobiasA ah you need an array of integers

reCurse: Why don't you just ignore me and save us both some time

ZarthaxX: that just gets complicated :P

reCurse: So obnoxious

ZarthaxX: Astrobytes dude i wanted to be cleaaaaarrr

FalINTOblivion0112: i am about to pass the top 1000 clash of code mark

FalINTOblivion0112: this is gonna be historic

ZarthaxX: i got a 10/10 in software design subject this semester, in your face astrolameius

ZarthaxX: :P

KiwiTae: FalINTOblivion0112 playd 1000games? or top 1000?

FalINTOblivion0112: top 100

FalINTOblivion0112: top 1000

Astrobytes: ZarthaxX: I defer to you good sir :D

xxd: you use bit shifting to "index" with a mask

TobiasA: thanks for the help ZarthaxX

ZarthaxX: TobiasA https://stackoverflow.com/questions/47981/how-do-you-set-clear-and-toggle-a-single-bit

Astrobytes: :kissing_heart: ZarthaxX

ZarthaxX: that's a start

BlaiseEbuth: Well. You're the one who come to yawn after me when I spoke earlier reCurse. I'm not pinging you... :roling_eyes:

Westicles: 10/10? What is that, some hippie grading scale?

ZarthaxX: Astrobytes :rofl:

ZarthaxX: love you my astroboi

Astrobytes: Yeah, and write a function to print your bits out so you can double check your results

ZarthaxX: Westicles our marks go from 0 to 10

Astrobytes: It's mostly quite intuitive TobiasA but it can get a little tricky

ZarthaxX: ah astroadviser

ZarthaxX: good one

xxd: astroconsultant

ZarthaxX: :rofl:

reCurse: You really don't know when to drop it

Westicles: Huh. Just shoot for a 4 and it will look great on a resume

ZarthaxX: Westicles yeah that's the minimum mark to pass :P

Westicles: My university had a 5 point scale. A plot of med school admissions vs GPA had this spike around 4

ZarthaxX: oh interesting :thinking:

ZarthaxX: so strange that there isnt a standarized way

xxd: in the US at least many schools use 4.0

xxd: but many codingamers seem to be from europe

xxd: or other... assorted places

FalINTOblivion0112: you guys are already in university?

FalINTOblivion0112: im only in junior high lmao xd

xxd: that explains your username

FalINTOblivion0112: yes

ZarthaxX: FalINTOblivion0112 people here are in all possible stages

ZarthaxX: school, uni, finished uni

FalINTOblivion0112: i know

ZarthaxX: :P

FalINTOblivion0112: im just suprised

BlaiseEbuth: Do you, reCurse? On this interesting question, I go to sleep. Kisses everyone!

darkhorse64: even finished for a very long time uni

michaellan: even in hell

TobiasA: jeez

michaellan: does C++ get boost on codingame server compilation?

michaellan: boost as in the library

darkhorse64: No

michaellan: aw shucks

michaellan: Is it not recommended to use iostream

darkhorse64: for input/output ? iostream is OK

michaellan: i think i heard someone say it was slower than cstdio

michaellan: i'll just use cstdio

darkhorse64: I have heard that too but it's OK for use on CG

darkhorse64: I mean, it's not like we need a lot of I/O

michaellan: that's true

Westicles: It is better to be cool than fast. Things like bitboards lack panache

darkhorse64: :rolling_eyes:

Astrobytes: there might be some performance benefit from printf vs std::cout but I think it's negligible

michaellan: should i do anything about the fact, that for example in Tic-Tac-Toe or Connect 4, that many differnet nodes will have the exact same state? should I somehow have them point to a singular copy of the state or have them be duplicated throughout the tree

darkhorse64: I leave them duplicated and do not try to deal with transpositions

TobiasA: now i am wondering why exactly shouldn't i use bitset?:thinking:

michaellan: do you guys like to store the state of the player who (1) just went; or (2) is about to go, in your MCTS state

Astrobytes: It's a little bit slower since you're using std

darkhorse64: (1)

Astrobytes: Feel free to use it to get used to stuff and check your manual operations

Astrobytes: (1)

darkhorse64: I do not think you can use pdep/pext/avx with bitsets

TobiasA: is the (0, 0) coordinate of the grid in tron top left or bottom left?

darkhorse64: CG likes top left as origin

Astrobytes: darkhorse64: No but he's trying to get used to just bit manip in general so he may be a little far off from that stuff rn ;)

TobiasA: when i use bitset coordinates are mirrored on the x-axis

Astrobytes: You're trying to bitboard Tron?

TobiasA: i used to get to gold

TobiasA: but it doesnt work anymore

Astrobytes: You may be overthinking a little

TobiasA: and i don't understand what is going on

Astrobytes: sorry, distracted, watching Daze Between festical

Astrobytes: *festival

TobiasA: will it be fast enough if i used just vector

Astrobytes: I don't know what you are doing, what you're using bitsets for or anything

TobiasA: in tron

TobiasA: to represent the grid

TobiasA: the state

TobiasA: whatever they call it

TobiasA: so i should use vector or no?

Astrobytes: Whatever is easiest. Straightup array is fine, vector is fine.

FalINTOblivion0112: YES

FalINTOblivion0112: LEVEL 20

FalINTOblivion0112: AND CODING SPEED LEGEND LEVEL

FalINTOblivion0112: now i need to have a contribution approved

AllYourTrees: damn lvl 20

FalINTOblivion0112: yes

AllYourTrees: im still lvl 19 lol

FalINTOblivion0112: i am also coding speed legend levl

FalINTOblivion0112: im so happy

AllYourTrees: congrats!

Astrobytes: yay for clashers.

FalINTOblivion0112: thank you all

darkhorse64: gg

Astrobytes: (don't mind my bitterness) grats dude

FalINTOblivion0112: yes

FalINTOblivion0112: now im gonna spam contributions

FalINTOblivion0112: one of them is gonna be approved

FalINTOblivion0112: and dont worry i will make them good

darkhorse64: Please no. Just make a good one

Astrobytes: How to attract haters on CG: Part 3

FalINTOblivion0112: YES

FalINTOblivion0112: such a good idea

FalINTOblivion0112: i will make that

FalINTOblivion0112: in fact

AllYourTrees: has anyone heard of the game Tak from the kingkiller chronicle by rothfuss?

Astrobytes: (Reasons to be cheerful Part 3 - Ian Dury & The Blockheads in case anyone wondered the reference)

Astrobytes: AllYourTrees: It exists irl now it seems? I know nothing about it

AllYourTrees: yeah they made it real, ive never played it though

AllYourTrees: i don't really know much about it either lol, read the books though so thoguht it was cool that they made it

Astrobytes: Well, you could do 3D chess à la TNG too

AllYourTrees: whats tng?

Astrobytes: Star Trek: The Next Generation

AllYourTrees: oooooh

**Astrobytes is not ashamed

AllYourTrees: hmm chess too hard, what about 3 d tic tac toe? you could call it... TTTT

Astrobytes: reCurse: please intervene

Astrobytes: hehe, no.

AllYourTrees: :scream:

AllYourTrees: a 3d game would be cool though

AllYourTrees: i like that idea

Astrobytes: There are a couple no?

Astrobytes: Dice Duels

AllYourTrees: oh really? which ones

Astrobytes: And another one of eulers... gimme a sec

AllYourTrees: dice duels this looks super interesting

Astrobytes: https://www.codingame.com/training/hard/blockout

AllYourTrees: oh wow

Astrobytes: ;)

AllYourTrees: ty for sharing :D

Astrobytes: np, they are both 'on my list'

AllYourTrees: ;) what are you on now?

Default avatar.png Retroblaze: hi

gilufin: how do I send red messager

Uljahn: messages containing your nick are automatically highlighted for you (and only you) in red

michaellan: IRC space-age technology

Astrobytes: AllYourTrees: FB and some other shiz

Default avatar.png Yuri123: Hi guys

Default avatar.png Yuri123: Anyone got problem with Chess board analyzer? i can explain what issue i've got so i use C# and even my program is passing all test at the end when i submit my solution he complains about one test and says that's for avoiding hard coding solution but i don't have any hard coding solution, anyone can help me for that?

Default avatar.png Yuri123: So i can't have 100% i'm stuck at 93%

Westicles: which validator?

Default avatar.png Yuri123: Escape by overtaking

Westicles: http://chat.codingame.com/pastebin/41116fed-5c4a-4d53-966d-6fad7e41033e

Westicles: Output: N

Default avatar.png Yuri123: bro i know

Default avatar.png Yuri123: i said it works

Default avatar.png Yuri123: the test doesnt fail

Default avatar.png Yuri123: i mean when i submit all my solution it complains about this test

Default avatar.png Yuri123: after having say it pass

Default avatar.png Yuri123: so all my test pass no problem at that moment

Westicles: validators are different than tests

Default avatar.png Yuri123: but when i SUBMIT

Default avatar.png Yuri123: it complains

Westicles: you need to figure out why your code fails with the validator I pasted

Default avatar.png Yuri123: it doesnt fail omg do u understand english?

Westicles: You aren't listening, 'bro'

Westicles: Good luck and gfy

Manjius: kek

michaellan: do u understand english even bro

Manjius: validators are test+hidden tests u cant see

Manjius: wait i mean, when u submit

Manjius: welp

michaellan: is it possible to have a global RNG generator in C++

michaellan: These are the first lines of my file, and this doesnt work: http://ix.io/3vpN

michaellan: ah, i use .discard() in main it seems

michaellan: nope thats not it

michaellan: ... why is random numbers in C++ so complicated

Default avatar.png IISadNightII: Vamos a hacerlo :)

michaellan: I'm not sure if I'm using shared_ptr correctly cuz none of my nodes are getting updated

michaellan: oh it probably has to do with going out-of-scope and deleting

Default avatar.png Yuri123: Finally i found the solution fucking chess analyzer

Default avatar.png Yuri123: how can i know hidden test lmao

Default avatar.png Yuri123: we should know all test even hidden ones

Default avatar.png Yuri123: ???

clashofcodeisbad: if you know all test cases anyone can just hardcorde the answer..

Default avatar.png Yuri123: yea i know that but even without hard coded the solution some test required "special indication" that aren't explained but are verified on hidden test

Default avatar.png Yuri123: i mean on all test that i passed none of them verify if the king can actually attack another piece of chess

Default avatar.png Yuri123: but this thing was actually verified on hidden test

Default avatar.png Yuri123: so all my test passed but fail on hidden test after submitting the solution

Default avatar.png Yuri123: doesn't make any sense for me

Default avatar.png Yuri123: also no indication of WHY it fails

juice0: after almost 2 hours of complaining about it, i still don't understand what exactly the problem is. but somehow you have a problem with less than perfect information?

Default avatar.png Yuri123: i can send u my solution and u will see what i'm talking about if u want?

juice0: that's not going to help either of us

Default avatar.png Yuri123: u will see that my solution pass but at the end it complains about hard coded solution which wasn't the case but another thing was the real problem

Default avatar.png Yuri123: on the hidden test it verify another thing that actual tests don't

Default avatar.png Yuri123: codingame is not perfect

Default avatar.png Yuri123: that's what i've learn

juice0: gotcha. yep i did a puzzle and it passed the tests but didn't get 100%. i learned a similar lesson. my next lesson should be to stay out of general chat while i'm trying to be better than 0% at the current task...

Default avatar.png Yuri123: hahaha

Default avatar.png Yuri123: good luck my friend i gtg have a good night cya

Wontonimo: the hidden tests (validation) are supposed to be different than the visible tests. The tests are just to help you get on the right track and give a bit of feedback. The validators are to see if your solved it.

Default avatar.png Yuri123: yea but how can we know what it expect from us?

Default avatar.png Yuri123: that's the question

Default avatar.png Yuri123: if it randomly fails without saying anything to guide u

Wontonimo: welcome to data science

Default avatar.png Yuri123: give me the keys i will launch the debugger of Codingame directly on the web host XDDDD

Wontonimo: this is part of science and testing in general. There are things you know, so you come up with a theory

Wontonimo: and then you test that theory on new data

Wontonimo: that's just basic science.

Wontonimo: if i show you how to add numbers from 1 to 20 except the 12's. I can test your understanding of addition by asking you the 12's

Wontonimo: if your response is "you didn't show me the 12s yet" then you don't understand addition

Wontonimo: you only memorized

Wontonimo: same thing

Wontonimo: as the validation

Default avatar.png Yuri123: well u can't apply that for all data science cuz some got infinit rules

Default avatar.png Yuri123: u will reach 100 years old without finishing it

Default avatar.png Yuri123: lol

Default avatar.png Yuri123: also as a programmer u are depending of ur technology and time and expectation and safety but can't predict all the cases cuz some can come from unpredicted sources that's why game devs are in front against bugs

Default avatar.png Yuri123: at high level nothing is perfect

Default avatar.png Yuri123: u have to admit it

Wontonimo: what coding challenge was it that you were working in?

Default avatar.png Yuri123: im asp.net developper

Default avatar.png Yuri123: i'm creating web app

reCurse: It's your job to predict all cases to the best of your ability

Wontonimo: +1

Default avatar.png TheDoomedLongcoat_ddf8: is this page for gud coder or begginer can understadn as well?

Default avatar.png Yuri123: yea at the end u can't predict all

Wontonimo: hey UCHIHASHISUI, how beginner?

Default avatar.png Yuri123: simple exemple if someone come with a modified linux core and browsers that's a thing u will never think about

Default avatar.png Yuri123: and ur web app is fucked up

Default avatar.png TheDoomedLongcoat_ddf8: as in basic, like understanding def function or module stuff, does this website highlight this? or it expect me o already know?

Default avatar.png Yuri123: nah here i'm coming just to play with some exercises

Default avatar.png Yuri123: i don't learn anything here

Wontonimo: http://chat.codingame.com/pastebin/77e53a7d-113e-4c53-80fe-250f2b322017

Wontonimo: oh, that's good enough for here UCHIHASHISUI. Check out this game and tell me what you think as far as difficulty https://www.codingame.com/training/easy/the-descent

Default avatar.png Yuri123: ok i will check out tomorow i need to go sleep sorry bro

Default avatar.png Yuri123: it's a bit late

Default avatar.png TheDoomedLongcoat_ddf8: wontimo, idk how to use thhis website, am i suppose to edit the code at the side or add stuff?

Wontonimo: yeah

Wontonimo: there is starting code

Wontonimo: and you read the problem statement on the left. make sure to scroll down on the left and read the whole thing

Wontonimo: it explains what your program should output and what inputs it is given

Wontonimo: If you want more instruction, there is this onboarding tutorial with popups to explain each button and area https://www.codingame.com/training/easy/onboarding

Default avatar.png TheDoomedLongcoat_ddf8: oh thx sir, for helping out

Wontonimo: let me know if you have any questions

Default avatar.png Turntlane: @codergirl05

codergirl05: @turntlane

Default avatar.png Turntlane: how are you doing these so fast omg

Wontonimo: cuz she's a coder. it even says so in her name

Wontonimo: :rolling_eyes:

Default avatar.png Turntlane: @codergirl05 can you plz explain how youre doing these

codergirl05: keep practicing

codergirl05: i've been coding for 5 years

codergirl05: and still learning

codergirl05: im sure you'll get there :))

Default avatar.png Turntlane: no like i actually want you to explain to me your process

Default avatar.png Turntlane: i wanna learn

Wontonimo: hey you two, try out the bot battles. They are way more fun than clashs. Here is the 'intro' one https://www.codingame.com/multiplayer/bot-programming/coders-strike-back

codergirl05: thanks

codergirl05: i'll give it a go some time

Wontonimo: Check out this replay of some bots https://www.codingame.com/replay/574000496

codergirl05: trying to complete these achievements

Wontonimo: look at how they arn't even trying to turn before the checkpoint, or slowing down properly

Wontonimo: The link I sent will start you on the AI achievement path

codergirl05: oh thanks a lot

codergirl05: i've saved the link

Default avatar.png hoangjunokatte: hi

Default avatar.png hoangjunokatte: hi

Wontonimo: hi hi

Default avatar.png hoangjunokatte: vietnamese and you

Wontonimo: for that game, a first good control system would be to not aim for the checkpoint, but instead aim for the checkpoint - velocity

Wontonimo: Here is an example of a different game where the bot aims for the target

Wontonimo: https://www.codingame.com/replay/574732847

Wontonimo: and here is that same bot aiming for the target minus its velocity

Wontonimo: https://www.codingame.com/replay/574732878

michaellan: urghhhh

michaellan: my C++ port of my python MCTS is going crazy

Wontonimo: ah

Wontonimo: you got any test cases?

michaellan: like this is the output for the first move of tic tac toe

michaellan: the square at the bottom is the chosen square

michaellan: but like that first square is just broken AF

michaellan: oops i never sent link

Wontonimo: what kind of probabilities/scores is it outputting for each move?

michaellan: http://ix.io/3vqi

michaellan: here is pastebin

michaellan: not sure if you're willing to read my code or not.. that is a big ask

michaellan: i'd appreciate any areas which might be causing this bug, any suspicions u might have

Wontonimo: yeah sure, i could have a look. sent me a PM

Wontonimo: from the probabilities, it looks like it may be that the backprop isn't updating the node that the backprop starts on, or the UCB calc is wrong