Snef: if your search have random it's logic
Snef: the fails tho..
Astrobytes: YurkovAS was wondering about the CPUs on the machines
Illedan: Yeah, some have said they have to use 35 ms to "fix" the fails..
YurkovAS: Snef yes, use randoms with fixed seed
Snef: yea me too
BeardedWhale: I use fixed seed, but since the searchtime is somewhat random, the results still vary somewhat.
Snef: struct u there ?
struct: now I am
cegprakash: top 20 on oware xD
cegprakash: i go till depth 15! what a pruning!
struct: nice cegprakash
jacek: see, and it was without that pruning array
jacek: wait, in this situation is draw? https://www.codingame.com/share-replay/469018118
jacek: i thought you take the seeds from the board
Doju: Hey, python question here:
Doju: So i have a vector class that inherits from a numpy array
Doju: i'd like that class to have vec.x and vec.y
Doju: i can do that by doing self.x = self and self.y = self in the __init__
Doju: but that breaks when i add two of them together for example
Doju: if i go (vec1 + vec2).x it throws an error
Doju: So how should I go about fixing that? Getters and setters?
Uljahn: why don't you like to use recarrays?
Uljahn: mixing python and numpy is generally not the best idea
Doju: Oh huh, that's pretty much exactly what i want
Doju: you're right
Doju: thanks Uljahn
cegprakash: I had a 13-8 lead in early game
cegprakash: and it doesn't seem to matter
WINWINWIN: Even your CoC puzzles are CSB based MadKnight?
WINWINWIN: You must output the length of a race track, given a number of coordinates, using the Euclidian distance . If the track has 2 coordinates, they represent a start and finish line. If the track has 3 coordinates or more, the first coordinate represents the start/finish, while the others represent checkpoints that are required to be passed through in order.
cegprakash: the problem with minimax is it assumes there is only 1 optimal strategy for any game state
jacek: could be more if 2 moves have the same score
cegprakash: I rekt eulerscheZahl on Oware.. feels awesome xD
cegprakash: 10-4 ahead still lose
cegprakash: he won with 1 freaking seed
cegprakash: 20-5 ahead and lost
darkhorse64: Won by overfeeding
cegprakash: i seriously don't understand if I should take it as bad luck or I should predict 100 turns ahead
jacek: ponies like seeds
jacek: what is your eval? only seeds count?
cegprakash: just score
cegprakash: and avoid getting 0 seeded
jacek: think of something more
cegprakash: my brain is dead
cegprakash: 3 days no sleep
cegprakash: only oware
cegprakash: I was trying to figure out obvious winning states
cegprakash: but I couldn't find any
cegprakash: may be MCTS is better
cegprakash: and top 20 has probably more than 20 depth
cegprakash: I only gave a depth = 9 to 15 based on no. of seeds left on board
jacek: the one when you get 25 seeds is obvious winning
cegprakash: my bot with timeouts do perform better than bot without timeout
Uljahn: top oware are NNs i guess
darkhorse64: 2 NN, 2 MCTS
Csipcsirip: nn like alpha zero ?
Csipcsirip: hm gotta make my own
jacek: with blackjack and hookers
cegprakash: is it good to have too much seeds?
darkhorse64: Wealth is good
darkhorse64: "Greed is good". Gordon Gecko
cegprakash: when I'm greedy
IAmNoob: what is this arena?
WINWINWIN: that was oware abapa IAmNoob
IAmNoob: k thx
Uljahn: jeez, there is "Did you like this game? PLAY" at the bottom, do you even read?
IAmNoob: i havent seen
IAmNoob: why u so toxic
WINWINWIN: calm down, this kind of talk is discouraged on the chat IAmNoob
IAmNoob: but why is he so toxic?
Uljahn: why u stupid
Uljahn: lvl26 should know such things :relieved:
Uljahn: unless you're a cheater
WINWINWIN: which emoji is that Uljahn?
IAmNoob: no i am not
WINWINWIN: cant find it?
IAmNoob: and i dident know because this is my first contest
Uljahn: hover over it
IAmNoob: the Pac man
IAmNoob: i mean
WINWINWIN: Ah thanks
WINWINWIN: Just found the magic formula for Code vx Zombies :)
cegprakash: WINWINWIN u spend small time on lot of games instead of lot of time on a single game?
WINWINWIN: cegprakash Yes a very short time on a lot of games
WINWINWIN: I just get till bronze and change focus like a goldfish :P
WINWINWIN: But some of them held my interest for a longer amount of time
jacek: those look nice for next cg multi https://en.wikipedia.org/wiki/Cylinder_chess#Horizontal_Cylinder_chess_and_Toroidal_chess
Alshock: that sounds like a good way WINWINWIN
Alshock: I should try that
wlesavo: im actually used to do that, because it was hard to improve and my motivation would drain out pretty quickly
wlesavo: NS was probably the first game i was really commited to and it felt much better trying different stuff on and on
wlesavo: number shifting
wlesavo: since it is an offline game you can try a lot of things, which normaly you wouldn't do, especially with python
wlesavo: i think i tryed mcts, GA and SA with a lot of different modifications, all of these algos for the first time
AntiSquid: on number shifting
Jasperr: I'm going to try to create my first working GA on SR tonight
Jasperr: Looks like a suitable competition
WINWINWIN: Jasperr seems like a good idea, no collisions so very easy to simulate
wlesavo: AntiSquid yes, but the only solid implementation was sa, other ones failed misserably
eulerscheZahl: can you elaborate on it? which puzzles? which kind of bug?
Astrobytes: They're all bugged. So is the leaderboard. The whole site is one huge bug.
eulerscheZahl: pretty sure it's a bug in your code. ASCII art is well tested by thousands of users
eulerscheZahl: for the first line of output that is
eulerscheZahl: missing or extra spaces?
Scarfield: missing a newline?
eulerscheZahl: share code
jacek: is there some state of the art software for GA parameter optimization? like i give it set of parameters and it calls my fitness function and it does all the genetic shit on its own and give me the optimized parameters
eulerscheZahl: you are missing a space in the first line of output
eulerscheZahl: and even more spaces for lines below
eulerscheZahl: replaced expected spaces by dots for visibility
Alshock: errr of course that depends on the ascii-art alphabet you're given as input
Astrobytes: You can't just hard-code the answers btw
Alshock: well technically on CG it's possible
Alshock: yeah, and be the lamest guy there's ever been. But it's still possible :P
eulerscheZahl: you won't pass with hardcoding
eulerscheZahl: the alphabet is different for the last testcase
eulerscheZahl: because of optimization?
Astrobytes: Your code is so highly optimised, it fails?
eulerscheZahl: there's always a way around the expected solution. but what's the point if it's harder than solving the problem itself?
eulerscheZahl: it's part of the input
eulerscheZahl: no, that's not the full input
Trice: Have any idea, what is this? input: 123 output: 2 Input: 2314 output: 3
Astrobytes: ArianPunk yes
eulerscheZahl: only 100% solutions are visible
Trice: input: 1111 output: 0 Input: 33278110 output: 5 Input: 112233445566778899 output: 8
Scarfield: you are clashing Trice?
Trice: yes :D
eulerscheZahl: Trice https://www.codingame.com/contribute/view/751a429755126a612e1e5aea3feedd8889f
Scarfield: you are in world chat
Trice: But I can't understand this problem
Trice: Thanks :D
Scarfield: lol i guessed it was largest int minus the first int xD
Astrobytes: Komby, I haven't done that puzzle but can't you just use a 2D array to represent the grid?
eulerscheZahl: and i spoiled the clash. not even feeling sorry for the other players :imp:
Trice: I exited from the clash when I asked for help.
Astrobytes: It shouldn't be Komby
eulerscheZahl: I approved that clash :o i don't even remember
eulerscheZahl: chat? triangle at to bottom
eulerscheZahl: now that I revealed this secret, will we lose more chatters?
MostComplicatedUsername: You see...
MostComplicatedUsername: Pressing the triangle takes WORK
eulerscheZahl: at least it's an actual link, vimium detects it. unlike switching chats
eulerscheZahl: meanwhile I just got a clash of code invite :rage:
Zenoscave: Is TvC still live?
eulerscheZahl: what is tvc?
Zenoscave: the JBM troll game
eulerscheZahl: toad vs chat?
Zenoscave: troll vs castle or something
eulerscheZahl: there's a link somewhere on the forum
Zenoscave: ok thanks
Zenoscave: found it!
Scarfield: how can you improve what is already perfect?
Astrobytes: That troll in fullscreen is the best thing I've seen this year.
MostComplicatedUsername: Wow that's one abundant puddle
AbundantPuddle: Just saw the time stamp. A bit late to the party I suppose. Lol.
AbundantPuddle: I like the irony of your username. Lmao.
MostComplicatedUsername: My code for bandas is pretty simple, every turn I calculate the side with the most opponent pawns, and move in that direction. But for some reason, it only ever chooses left
Astrobytes: You have a bug in your code.
Astrobytes: I also have a bug in my SR
darkhorse64: Don't write an eval. You'll have no bug. Only TrueAI
Astrobytes: Dammit, that's what's wrong with my GA, it requires no fitness function and no sim.
darkhorse64: Yeah remove the sim
Astrobytes: Add some if-elses
darkhorse64: Kidding aside, my bandas bot has no eval at all, only random rollouts
struct: who needs eval
Astrobytes: Not tried Bandas properly yet
Alshock: @struct interpreted languages do
Alshock: Ok I'm out :door:
darkhorse64: I think my SR fitness sucks. I am stuck slightly above 12k
struct: No yavalath rating is ruined :'(
struct: 2 stars
darkhorse64: Only 5 stars allowed
jacek: good thing it wasnt 3 stars
Astrobytes: I ported some of my CSB code today, and it sort of works but doesn't. Will check the sim properly tomorrow.
darkhorse64: Beware that pos and speed are truncated
eulerscheZahl: 2 stars, cute... got 1 star on tower dereference and onitama :sob:
Astrobytes: Yup, saw those darkhorse64
darkhorse64: Collisions are also a bit different. The referee helps for that
struct: damn :/
Astrobytes: Yeah, I'll go through it properly tomorrow morning
struct: you have as many 1 stars as BOTH
darkhorse64: Do you know of sensible parameters for CSB aside fro checkpoint count and distance to next checkpoint ?
struct: angles maybe?
eulerscheZahl: i think it was 1 user giving 1 star ratings to 6 community multiplayers
Astrobytes: blocker distance to enemy CP, to enemy runner, angles for sure
struct: Like enemy blocker angle to your runner
struct: or what astro said
Astrobytes: there's literally *loads* of parameters you can try
struct: I never tried any though
struct: blocker distance to enemy cp is good
Astrobytes: It's balancing them all that's the issue (like always)
darkhorse64: And for runner ?
Astrobytes: distance from enemy blocker, enemy runner is covered I guess, angles again
Astrobytes: Play around with it.
darkhorse64: angle is what is missing from my SR fitness function
Astrobytes: I would definitely include that
darkhorse64: Although it's hard to define what's the right value
Astrobytes: Yep. You don't want it going straight towards the checkpoint at all times ofc
darkhorse64: I feel transportation illness when seeing SR replays
Astrobytes: Even on slow speed?
Astrobytes: (called motion-sickness btw ;) )
darkhorse64: Slow is not an option
Astrobytes: Yeah I watch on max speed too
darkhorse64: forgot the right words. thanks
Astrobytes: no prob, ttyl
jrke: hey is this search race new
Scarfield: SlowstroBytes ;)
Astrobytes: hahaha Scarfandgloves :D
Scarfield: lol xD
Scarfield: never realised scarf was part of my name, now i cannot unsee it
Astrobytes: sorrynotsorry :P
jrke: hey search race same like CSB
jrke: just copying
Astrobytes: little bit different
jrke: whats change
jrke: just alternative output
Astrobytes: The engine is a bit different, so your sim will need to take that into account
jrke: seems recurse will be first
Astrobytes: He *is* first
jrke: whats your progress astrobytes
Astrobytes: 19th/44 16,322.67 100%, no search
Astrobytes: Currently porting my GA from CSB
Illedan: jrke, change is that there is no enemy :P
Illedan: Just to practice the racing part
eulerscheZahl: and that CSB never killed any other games
eulerscheZahl: wait, promotion... forget what i just said
jrke: now something else to add going
jrke: hey how give rotation angle output
Astrobytes: "EXPERT angle thrust"
Astrobytes: It's at bottom of the statement ;)
BeardedWhale: Yay! Finally below 11k on Search race!
Zenoscave: Second in TvC!!! JBM your game is fun
Illedan: :clap: BeardedWhale!
Zenoscave: gz BeardedWhale are you doing a sim-less ai?
Illedan: -4 vel is the magic formula
Zenoscave: not again
Illedan: You didn't try yet?
351062: -3vel > -4vel
BeardedWhale: Zenoscave, no I am using GA
Astrobytes: gj BeardedWhale
Astrobytes: Debugging my sim + GA rn
Astrobytes: Think I fkd something up earlier. I blame my pain relief ofc
Astrobytes: In Java too BeardedWhale :clap:
Astrobytes: Very well done
AntiSquid: left CG open and when i come back i see a bunch of clash invite, wtf? is there no cure for that shit?
AntiSquid: what's the current game topic? speed racer?
BeardedWhale: Yeah, I'm doing speed racer at least
PatrickMcGinnisII: what is TvC?
PatrickMcGinnisII: oh, the pew pew is gone
Illedan: Troll vs Castle PatrickMcGinnisII
PatrickMcGinnisII: tx Illedan, didn't know
BeardedWhale: I cannot find Troll vs Castle, where is it?
caillouthebuilder: yo guys
AbundantPuddle: Hey caillou
cegprakash: 16 points diff bw #18 and #1 in oware