Yasho25: print("Hello World")
MadKnight: print("Hello Automaton2000 ")
Automaton2000: but you are welcome :)
Marchete: new ideas, jrke?
jrke: jo not new ideas just trying to fix bugs
jrke: no*
MadKnight: hey Marchete
dbdr: jrke: fixing bugs is a good idea :)
dbdr: MadKnight, are you working on a bot now?
jrke: but 0 bugs fixed till now or maybe some new i introduced
dbdr: cannot you reproduce those endgame mistakes that MSmits mentioned?
jrke: endgame bugs are too diffucult to fix in my current bot
dbdr: why?
jrke: code became messy
dbdr: time to refactor
jrke: yeah
jrke: dbdr:do you made any other change or resubmit
dbdr: changed
jrke: k
dbdr: do you resubmit just to see if you get lucky? ;)
jrke: no im changing small things and checking will it work or not
dbdr: probably you would get more benefits from fixing the endgame bugs
jrke: yeah the major bug or you can say weakness of my bot
jrke: did you went heuristic in endgame?
Vasyakin: 5051
dbdr: yes
jrke: means turns > 56 for game and 23> for a single player right?
dbdr: not precisely, but that's roughly when endgames starts, yes
jrke: ok
Sevi1: hello
dbdr: hi
jrke: hey
jrke: dbdr nice submission
dbdr: thank
dbdr: s
Sevi1: what you doing on this page
dbdr: nobody knows
Ethan.B: hello?
Marc19: hey
AntiSquid: hi
AntiSquid: it's a hello world page, now make your own Sevi1
Marchete: dbdr_: 8-41 against jrke?
dbdr: it happened
dbdr: thanks for being here Marchete
so we don't need to be last at 0% ;)
dbdr: played: 34
lost: 0
played against #1: 0
- D
dbdr: .
Marchete: :S
Marchete: good bless debuggers
dbdr: and those writing them?
Marchete: D&B seems a top3 game right now
dbdr: and top 3 make some stupid mistakes
dbdr: it's unusual
m0rsmordre: hello there
Marchete: maybe people tend to go with simple heuristics
Illedan: My bot is dead stupid, no idea why you are not getting 100 % wins vs it :P
darkhorse64: No double dealing, chain/loop stuff ?
Illedan: Double dealing?
Marchete: there is a simple way to know if a line create two separated chains?
Marchete: lol Illedan_ :D
darkhorse64: OK, really dead stupid bot
Marchete: so: D&B seems a top2 game right now
Marchete: no offense :D
Illedan: I know
Illedan: But I don't get how they don't crush my bot
darkhorse64: double dealing is a trick to keep control when farming boxes in the endgame
Illedan: To split a 4 chain into 2 doubles?
Marchete: it's about control
Marchete: give boxes for a greater gain
Marchete: "sacrifice"
Illedan: Yeah, I do those things. But I never check if it is worth it :P
Marchete: then you are using it
Marchete: maybe you call it "Illedeal", but it's probably de double dealing
Marchete: the*
darkhorse64: When you farm boxes, you sacrifice the last two boxes in the chain to force your opponent to open a new chain and leave the hand to you
Illedan: Yeah, that is easy to do with a simple IF
darkhorse64: Actually, when there is no single box with three borders left, you are in the endgame and there are algos to ply optimally. The real deal is building chains so that you win in the end. Othersise said, it happens between move 1 to 56, after it's strictly determinisitic
darkhorse64: *Otherwise*
Illedan: Yeah, I was expecting the top to create those chains already :P
Marchete: :unamused:
Marchete: you are the -3vel of D&B Illedan
Illedan: Most advanced stuff I have is a floodfill :P
Marchete: wow!
darkhorse64: Have you heard about string and coins ? It's the dual game compared to B&D and it provides a data representation that allows esay chain computation
darkhorse64: No floodfill needed
darkhorse64: Unless you need for other purposes
darkhorse64: *need it*
Illedan: I just save the chucks and the number of edges, using it to determine my logic :P
Illedan: Didn't even try to model this stuff correctly
Illedan: Still using both edges provided for 1 edge :D
darkhorse64: No search ? Just reasoning on topology ?
darkhorse64: I mean no deep search
darkhorse64: No MCTS or Minimax have been harmed during this multi
JBM: you like that phrasing
darkhorse64: a chat meme ?
JBM: like we need more
darkhorse64: OK. Twice the same day is too much but I used two different chats
Marchete: depth = 0 bot :D
darkhorse64: TrueAI bot
MSmits: I gathered some statistics of the nature of simple loony endgames in 7x7 D&B
MSmits: Turns out that by the definition of simple used by those papers with heuristics that are written for 5 x 5, about 0,3% of endgames are simple
MSmits: meaning that if I generate 1 million endgames, I get 3k simple ones :P
dbdr: randomly?
MSmits: yes
jrke: so what you discovered from that Ms?
MSmits: that the heuristics as formulated are useless unless expanded. That will be my project :P
MSmits: also there is one heuristic that says if the controlled value is >= 2, you can just take it as the real value
jrke: how many possible outcomes could be?
MSmits: out of all the simple games, that is almost always true
MSmits: so the more complex heuristics are useful in about 0,02% of cases
jrke: so you will go with heuristic approach?
MSmits: sry, 0,002%
jrke: with expansion
MSmits: I will try it, it's a fun exercise. Not lookign good though :)
MSmits: bte
MSmits: btw
MSmits: http://chat.codingame.com/pastebin/1894abf5-98c9-473b-af72-d3062e9bb88e
MSmits: thats one of those super rare simple loony end games where complicated heuristics are useful
MSmits: controlled value is negative, so you got some control switching back and forth
MSmits: people will actually take the boxes in the loops, instead of giving them away
jrke: yeah i think i will lupdate my bot to turn heuristic at endgame
MSmits: yeah, but that's not easy at all and as i said, almost never useful unless you expand on the current research knowledge base of D&B
MSmits: so if you succeed, you could publish a paper on it
MSmits: it's hard for me to even test if i implemented the heuristics correctly, because I have to look so hard for a suitable game
MSmits: I'm going to try and code a minimax to solve the end games that way and see if it lines up with the heuristic result
MSmits: then I can cook up some new heuristics, run a million games and see if they still agree
MSmits: fun fun
jrke: million games offline?
darkhorse64: If there are still 3+ connections, it's not yet an endgame but you have to search until the board comes to that point (or find an heuristic for these) right ?
MSmits: yeah, what i actually mean is, try 1 million seeds (takes a few seconds) and only take the games handled by the heuristics and minimax those
MSmits: darkhorse64 that's not true, a 3+ connection thing can be an endgame
MSmits: it's just not a simple endgame
MSmits: because you can't place a line on any of the 3 places that are empty due to neighbors
MSmits: (that already have 2 lines)
MSmits: even a 4 way split is possible in an endgame
MSmits: I found 2 end games with 12 splits in a single floodfilled structure, out of 1 million games
MSmits: so that's lots of crossroads
MSmits: lemme find one and share it
darkhorse64: OK. You are in the endgame if any move results in giving away a box
darkhorse64: something called a grid lock state
MSmits: http://chat.codingame.com/pastebin/58fd10f9-6c63-4017-9f5d-43e2068077c2
MSmits: right
MSmits: look at my example
MSmits: everything is connected
MSmits: still an endgame
MSmits: very non-simple
dbdr: single chain :)
MSmits: not this one, but sure, single chain is possible
MSmits: single chain is actually a simple end game
dbdr: right, chain is not the right word
jrke: https://www.codingame.com/replay/484083437
dbdr: yeah
MSmits: yeah jrke, my heuristics work for that one :)
MSmits: but it's trivial
dbdr: MSmits, what year can we expect your bot? ;)
MSmits: 2020 hopefully :P
MSmits: honestly not sure
MSmits: 2 things can happen. Either I keep trying to do this heuristic thing and finally succeed. Or I give up and just make some basic search bot and give up.
MSmits: thats giving up twice :P
MSmits: I like a challenge though, we'll see
reCurse: Giving up is writing an opening book
MSmits: can't write an opening book without a good bot
MSmits: or anyway, not one that wins
dbdr: also I think OB is useless here
MSmits: depends on how others do their bot. if they dont use appropriate randomness, then it might not be
MSmits: but should be easy to do
dbdr: useless against decent bots ;)
MSmits: should be yeah
dbdr: initially, I meant rather useless for the general case
dbdr: as opposed to beating a few specific bots
MSmits: ah yeah, it sure is useless for that
MSmits: not in ever game btw.
dbdr: definitely
MSmits: btw, I do have two expansions on the basic heuristics that might work. If you want to discuss them, I am not 100% sure
dbdr: there must be a typology of games
dbdr: or we must create one
MSmits: sure
dbdr: does branching go up, down, or up-then-down, for instance
MSmits: that exists
MSmits: is called diverging or converging
dbdr: that has a name?
MSmits: yeah
dbdr: ah nice
MSmits: read that ina paper on solving games
MSmits: othelo diverges, then converges
dbdr: right
dbdr: D&B converges
MSmits: so, I was thinking, you always open 1 chains first, then 2 right?
dbdr: yes
MSmits: so isn't that also the case when the 1's and 2's are connected to a split?
dbdr: what's "that"?
MSmits: that you would take them first
MSmits: so if they arent isolated
MSmits: but connected to a 3-way or 4 way
dbdr: I haven't proved it's optimal
dbdr: just seemed logical
MSmits: me neither, but I did some examples and it did seem logical
MSmits: another one is
MSmits: say a split has only chains attached to it
MSmits: , for example 3 chain, 4 chain, 5 chain and 6 chain
MSmits: all connected by 4 way split
MSmits: would it ever be worse to open the smallest first?
MSmits: since the rule for simple games is to always open the smaller chains before the larger ones
dbdr: I always do smallest first
MSmits: ah yes and thats certainly optimal for simple games
MSmits: I am just wondering if it's true for games with splits also
dbdr: but if you open, you have lost control, and it's not often that you can win
MSmits: true, but there might be a bunch of loops you took earlier
MSmits: loops are beneficial for the defender (non-controller)
dbdr: yeah, and 2s
dbdr: well, 2s you switch
Stilgart: MSmits, dbdr: do you use a hardcode solution for CvZ ?
MSmits: sure, but 2's go back and forth so it only helps for an odd amount
MSmits: the penalty for short chains is at most 1 or 2
dbdr: Stilgart: offline, yes
dbdr: why?
MSmits: i do too, I even used unity for this one
Stilgart: i am trying to figure out what is the best non-harcoded score in the leaderboard
Stilgart: 700k sounds reachable without hardcoding
MSmits: Robo's non-hardcoded
MSmits: rank 24 with 463k
MSmits: my fallback solver has like 200-300k points only. It's a basic MC
dbdr: it's a thin line between offline and online spamsubmit
MSmits: I think the difference between nr 1 and lower ranked players might just be 1 validator
dbdr: theorem: as the number of submits tends to infinity, it is undistinguishable from offline search
MSmits: infinity yeah
MSmits: but the calculation time is used less efficiently
reCurse: Theorem: infinity does not exist
darkhorse64: You have to love cars, stairs, traffic lights and busses
dbdr: "tends to", not sure if that's the right word
MSmits: the difference between doing a meta mcts or just using the same calculation time for MC
MSmits: when you just do MC, you lose all information gained from previous iterations
MSmits: nothing is stored
dbdr: reCurse: of course infinity exists :)
MSmits: I dont even understand what "exists" means in this case
darkhorse64: AG, HC should be better in this respect
dbdr: has a precise meaning
MSmits: whats AG and HC?
MSmits: also what do you mean by precise dbdr? That exists means it has a precise meaning or that infinity has precise meaning
darkhorse64: AG->GA= Genetic algorithm, Hill Climbing
MSmits: ohh ok
darkhorse64: AG is french
MSmits: I see
dbdr: I was trying to come up with a definition of "exists". not saying it's the ultimate one
MSmits: ahh ok
MSmits: also it's your definition
reCurse: So dragons exist? I have a precise meaning for it
MSmits: we need to all share the same definition to use it in a conversation effectively :)
Stilgart: yeap, dragons exist
dbdr: there has been debates whether 0 exists, or negative numbers, or complex ones, ...
dbdr: at least in math
dbdr: if you can give a precise definition of dragons in math language, then yes they exist in math :)
MSmits: physics is the subset of math that exists
dbdr: troll :D
MSmits: maybe :)
dbdr: "what exists is what I understand" ;)
MSmits: solipsism?
reCurse: He's kinda right though
darkhorse64: phisics is to math what military music is to music
Stilgart: dbdr: it's a cylinder with spikes that spits fire
dbdr: isn't that ""what exists is I only" ? :)
MSmits: yeah
dbdr: darkhorse64 I heard that about military justice
darkhorse64: It's valid too
darkhorse64: Or you can say the same for Strauss and Mozart
dbdr: the mandelbrot fractal exists, but you cannot make a physical object of it
MSmits: the idea of it exists in your mind, but part of the idea is the infinity part.
dbdr: you can define it and calculate properties, you can visialize any part of it on a computer, ...
MSmits: yeah
reCurse: Exist in your mind is contradictory
dbdr: that's exactly the definition of "towards infinity": it's not that you can see it all at once
dbdr: it's that there is no end
MSmits: reCurse that hinges on the definition of "exists" again
dbdr: and you can always see further
reCurse: If that's the definition then everything exists
reCurse: Not very useful
MSmits: noone said it had to be useful
MSmits: well you just did I suppose
dbdr: there are very few things that can be defined precisely
dbdr: so it's a very useful concept
dbdr: infinite but few. like natural numbers compared to real ones
reCurse: So everything exists
MSmits: I had a discussion about that with a professor when I was in high school
MSmits: got a little heated
MSmits: I said there were more real numbers than natural ones and he claimed they were both infinite so the amount is equal
dbdr: define "everything" ;)
dbdr: you were right MSmits
dbdr: I thought you were the one who didn't get this
LinhT.Nguyen: indeed
MSmits: no I get it, it just didnt make sense to me that all infinities are the same
dbdr: and there is a precise definition of "more"
SydM107: all infinities are not the same...
dbdr: they are not the same
MSmits: right
dbdr: reCurse derailed the chat with his obscurantism ;)
MSmits: allright gonna try to do some coding on this minimax, been a while since i coded one
dbdr: ok, report back in 10 minutes
MSmits: 10 hrs probably :P
dbdr: for minimax? ;)
MSmits: nah it's mostly the move generation
dbdr: right
MSmits: instead of playing lines i need to pick chains and loops to play
dbdr: or lines
MSmits: not in endgames
MSmits: simple ones anyway
dbdr: ah, only for endgame
dbdr: right
reCurse: Obscurantism pfft
MSmits: i need the minimax to check if my heuristics are correct
dbdr: "infinity does not exist" is basically rejecting science
reCurse: lol ok...
MSmits: what if you're saying the earth is flat and infinite ?
reCurse: Nice troll
MSmits: :)
JBM: and of homogeneous resistivity
MSmits: damn now i have two things to google
**MSmits googles obscurantism and homogeneous resistivity
dbdr: the earth is fractal, it has finite volume and infinite surface
JBM: got it wrong, it was "uniform density" https://xkcd.com/669/ sorry MSmits
MSmits: still doesnt make sense :P
MSmits: infinite flat earth violates gravity pretty much though
JBM: you don't say
JBM: just make gravity a constant uniform field, duh
MSmits: oh, right
JBM: it's how *my* physics teachers did it
MSmits: that's actually an easy solution
JBM: then again you might have your own tricks
MSmits: no no, that's what we do
JBM: see
dbdr: do we even know what causes gravity?
JBM: your mom
reCurse: lol
JBM: there, done.
MSmits: perfect burn
dbdr: :clap:
MSmits: if you're asking why stuff is attracted to other stuff it's answerable
MSmits: it's the curvature of spacetime
MSmits: if you want to know why mass curves spacetime, we dont know
SydM107: more of a how than a why
MSmits: yeah, the how is hard enough
MSmits: without worrying about the why
dbdr: yeah, why question ftw :)
Marchete: nice output
MSmits: there's much simpler questions physicists can't answer
MSmits: like, what is electrical charge?
MSmits: we know perfectly what it does, but not what it *is*
MSmits: true for any elementary property I suppose
Marchete: I think any N chains of degrees < 3 are solvable with theorems
Marchete: but not checked
MSmits: < 3 or <= 3?
Marchete: < 3 like your example
MSmits: yeah they are for sure in simple end games
MSmits: thats what the papers say
Marchete: ok then :D
MSmits: but what if they're connected to a split
MSmits: 3-way or 4-way
Marchete: then it's not a endgame for me :D
MSmits: your definition of an endgame is different from what the papers say then
MSmits: arbitrary endgames are games where you cant place a line without giving away a box
MSmits: http://chat.codingame.com/pastebin/9ac30b07-fa27-45e7-82db-0530e056389d
MSmits: is this an endgame for you Marchete?
dbdr: MSmits: https://www.youtube.com/watch?v=2rjbtsX7twc
MSmits: lol
dbdr: you know it?
dbdr: "these infinities are real" is also dedicated to reCurse :)
Marchete: not an endgame at all
reCurse: My point didn't make it through so forget it
reCurse: Easier to just say I'm a troll or "obscurantism"
reCurse: w/e
MSmits: Marchete you cant play a line anywhere though
Marchete: there is a special case of a loop joined to a ground chain
Marchete: but with a T
MSmits: without giving away a box
Marchete: I can't say a lot
MSmits: with your definition, you might have to wait till the last few lines till it becomes an endgame
MSmits: there's no reason those exceptions will be opened first
dbdr: reCurse: I spoke about "approaching infinity", which is the basis of calculus. saying "infinity does not exist" misses the point
dbdr: it's not "reaches infinity"
jrke: the thing you imagined is not infinity
MSmits: dbdr i understood about 30% of that song :P
MSmits: enough for it to be fun
dbdr: me probably 5 or 10% :)
dbdr: but still fun, I can see what it's doing
MSmits: whats cool is that the guy shared his masters thesis in the description
MSmits: so he's not just singin it
dbdr: yeah, you need to know a lot just to write the lyrics
MSmits: yea
MSmits: it's really nicely edited too
MSmits: could be a professional music video easily
dbdr: https://www.youtube.com/watch?v=i6rVHr6OwjI
this one is good too
JBM: it's one of his earliest
JBM: they got much better
JBM: i love the hamilton one
dbdr: he had to learn to speak that one backwards. definitely huge work
MSmits: yeah
MSmits: he sang it backwards
MSmits: I think
dbdr: well, pretended at least, for the lips
dbdr: I'm sure the sound track is still recorded forward
MSmits: i'm not sure, some guy in the description says there's a making-of video where he sings it backwards
dbdr: for the lips
MSmits: but i cant imagine singing backwards
jrke: i just made my 200th submission in D&B
jrke: gn
MSmits: nice jrke :)
dbdr: really? 200
MSmits: I have 4 submissions of which 2 failed in wood 2 :P
MSmits: cuz the bot was too good
dbdr: classic
dbdr: I have 40
MSmits: he sang it backwards
MSmits: https://www.youtube.com/watch?v=54nAOaEe_Fw
MSmits: dude is crazy
JBM: i love him
dbdr: right, but if you reverse this track does it give the final one?
JBM: he's one of the very rare youtubers i've actually sent money to
dbdr: also, the final track has many voices at once
MSmits: no idea dbdr, maybe he explains it in the video, still watching
EveryOS: On the home tab, how did i get the collaboration badge under "Unlocked Certifications"
EveryOS: ?
MSmits: maybe they thought you were EricSMSO, that's what keeps happening to me
MSmits: https://www.codingame.com/profile/f5f8fd68a484329ecd629d610c8d05755689841 this guy
EveryOS: Ah gotcha, that's weird
MSmits: i dont actually think thats why you got the badge, but i do get confused a lot
IcedCrescent: does anyone have "Activating C# language assist" when choosing C# in the editor? It seems that intellisense never works
MSmits: I don't code much in the IDE, but when I do, the assist annoys me (using c++)
MSmits: sometimes the error reporting helps a bit
MSmits: I suggest you code locally for anything bigger than a puzzle
MSmits: especially in C# when you can run VS
IcedCrescent: Yeah, I retorted to using RoslynPad to write snippets (It's like LinqPad but free)
DoxLee: Hi
reCurse: I tried it a bit and it worked very well with C++ actually
reCurse: Light years ahead of the old editor anyway
MSmits: I guess I am just annoyed that it slows things down when you're not using it
MSmits: loading the IDE with it takes more time
jacek: reCurse light years are units of distance, not time
jacek: common mistake
jacek: :v
MSmits: I'd say you're trolling, but I am blown away by the correctness of your physics
jacek: so, no trolling?
JBM: how far away?
reCurse: I was talking about a distance
reCurse: Should go without saying
**JBM torn between "that's not how common mistakes work" and "that's not how analogies work"
**reCurse torn between sticking to Discord and still giving webchat a try
JBM: any action on discord nowadays?
reCurse: No but the signal to noise ratio is much better
JBM: i've given it a read this afternoon to marke much as caught up
jacek: why dont bridge discord and webchat?
JBM: so i disagree
jacek: that solution would be light years ahead
JBM: but i suppose it's self-selecting
JBM: jacek: i'll tell the devs about it
reCurse: Webchat is mostly filled with trolls anyway
JBM: or
JBM: webchat is mostly filled with trolls *when you're there*
JBM: hard to tell
reCurse: Case in point
MSmits: wait... are you saying you're a troll when you're here, but not on discord? :confused:
JBM: that's my reading
reCurse: No I'm saying this is another prime example of trolling
reCurse: Or pointless discussion
MSmits: there's more points on discord?
JBM: after the infinities' inexistence, that's rich XD
reCurse: Alright you guys win
reCurse: Bye
MSmits: I blame this on you JBM
JBM: there's probably some gamification thingy i'm missing on discord
JBM: MSmits: hey discord doesn't want me ok
MSmits: possibly
JBM: by a very personal choice of him, dare i add ^^'
MSmits: honestly though, I don't mind light banter with no point and I like discussion of the game mechanics also
MSmits: of D&B for example
jacek: or UTTT
MSmits: are you trying to make sure re curse is really gone now jacek
JBM: what? he's got a problem with uttt?
MSmits: sort of i guess
MSmits: not sure how deep it goes
jacek: deeper than your book
MSmits: oof
JBM: i'm surprised, i'd have thought that's the kind of multi he'c crave
JBM: readjusting framing... complete
MSmits: nah he doesnt like it and he does kind of have a point
jacek: its love-hare relationship
jacek: hate even
JBM: i'd follow on the uninterestingness aspect of the game
MSmits: it's very hard to get a reading on that game as a human
JBM: but there's a MIPS aspect i'd expect him to shine in
MSmits: some weak heuristics you can use, but other than that you just try all sorts of random enhancements on mcts to see what sticks
MSmits: he was hoping NN would help, but that failed also
MSmits: I still have one thing on my list for uttt
MSmits: wanted to try to solve it with proof number search
MSmits: probably out of reach, but the statespace is far smaller than most people think
MSmits: maybe if the advantage of p1 is large enough, you can put lots of heuristics in that preserve the win for p1, you dont need to explore the entire tree that way
MSmits: at worst, I end up with a nice solver to add to my meta mcts. Anyways, thats a long term idea
AntiSquid: what NN did you do msd9126
AntiSquid: MSmits
MSmits: i didnt, recur se did
MSmits: for uttt
AntiSquid: always pinging wrong person lol
JBM: and unpinging someone who left
AntiSquid: ah i thought you did one for the dots games
MSmits: he also played millions of games in a meta mcts to find patterns to help design his NN I think
MSmits: nahh, i just ran a random sim to generate end games to test heuristics for
AntiSquid: how come there's so many people to ping but hardly anyone typing stuff in chat ?
MSmits: turns out the heuristics from papers cover 0,3 % of 7x7 endgames
JBM: they're all on discord
JBM: alternatively: "blame chat autojoin"
JBM: 7x7 endgames?
JBM: are we still on UTTT?
MSmits: thats the D&B size on CG
MSmits: papers were written for 5x5
JBM: yeah that's what i thought
JBM: i just missed the topicchange
MSmits: the paper results are general to any size, just not useful
MSmits: yeah I am lightning fast
JBM: and subtle
MSmits: so anyway, this kind of means my bot is 0,3% done :P
MSmits: needs more heuristics
dbdr: 7x7x7x7 uTTT would be a monster
MSmits: how would the board switching work?
eulerscheZahl: AntiSquid 07:59PM
always pinging wrong person lol
i didn't see this disadvantage of being a mod coming
eulerscheZahl: too many users online to auto-complete with just 2 letters
MSmits: also dbdr... uttt? with 7 cells per miniboard?
jacek: EveryOS?
MSmits: I think I know why you're not top 10 in uttt :P
AntiSquid: what's the secret to climbin in UTTT these days? quite sure my bot was around 60 at some point, #136 all of a sudden @_@
MSmits: nothing new
MSmits: I think top 20 or so have some heuristics
MSmits: though, the teccles heuristic is quite powerful, maybe that caught on
MSmits: if you put that in, you might get a good amount of ranks
MSmits: if you dont know what it is, it's basically an early game heuristic for player 1
AntiSquid: i remember it
MSmits: if you arrive at an empty board, play the cell with the same index as the board itself, so the opponent is sent to the board you're already in
MSmits: kk
MSmits: put it in
MSmits: test it
AntiSquid: send opponent back to same cell or something
MSmits: yes
MSmits: my bot has it, even though it's not perfect
AntiSquid: nah don't feel like it @_@ didn't even get the new games done yet
jacek: my bot went from ~15 to ~7
jacek: when i added this
MSmits: nice
AntiSquid: salad tomatoes #3
MSmits: it's mostly because in the early game the rollout-statistics are very bad
jacek: why tilde is more like minus here
MSmits: so you need a heuristic that almost guarantees a good move
AntiSquid: looks more like a tilde on ubuntu
MSmits: there's some for player 2, but they are worse, might not be worth it
AntiSquid: or maybe it's the screen, not sure
jacek: https://i.imgur.com/w2MnNTf.png
MSmits: yea it looks like a minus for me too
MSmits: at first i thought it was jacek-humor
AntiSquid: https://imgur.com/a/B6HkjEi
jacek: nyoro~n
AntiSquid: or maybe there's different font installed
AntiSquid: https://www.youtube.com/watch?v=A5Fpr7g7plE
Hjax: recently when im home, the chat doesnt load on this site
Hjax: anyone experienced that before?
Hjax: theres just a white box where the chat is supposed to be
jacek: thats racist
Hjax: :thinking:
eulerscheZahl: black boxes matter?
Ace-Code: :D
Ace-Code: XD
eulerscheZahl: on a more serious note: chat is working fine for me
Hjax: its only my desktop it doesnt work on
Hjax: i have no idea why, same browser / settings as other places i log in
eulerscheZahl: any errors in the console?
Hjax: ill check when im home later
jacek: suddenly im loading #3859980 channel o.O
eulerscheZahl: is that a direct message?
jacek: dunno, it only says the chat is loading
jacek: MSmits have you seen the yet another hexagonal game?
MSmits: nope
MSmits: but i just randomly generated 1 million D&B endgames out of which 3322 games were simple endgames and tested my heuristics compared to a negamax and they were identical each time. Yay endgame 0,3% solved :)
MSmits: only took 10 seconds or so
MSmits: on c++ debug mode
jacek: you could be the guy who spent 8 years to solve 5x5
MSmits: sure
MSmits: for now I seem to be the guy that spends that much time trying to beat your oware bot
MSmits: some branches go 70 plies deep now :P
jacek: youre just trying to outbook it
MSmits: of course
MSmits: it's fun though
MSmits: worked on everyone else so far, your bot is quite something
MSmits: http://chat.codingame.com/pastebin/95b1bcfa-502a-439f-b0bc-cd9b3e3169a6
MSmits: deepest so far, you're p2
MSmits: that one is actually close to the endgame book of 15 seeds
MSmits: but not that useful if I am getting wrecked on that branch :P
jacek: maybe i should learn about avx after all
MSmits: from what I can tell it's mostly useful for those physics games
MSmits: and othello
jacek: nn is just vector dot product. i think its autovectorized in my case anyway
MSmits: oh right, nns should use it too ofc
MSmits: it seems annoying to avx it by hand only to find out there is no improvement due to autovectorization
MSmits: in othello i did get an improvement, it was not autovectorized
MSmits: just a tad too complicated to do that I guess
MSmits: getting some stuff from the store bbl
jacek: for bt i do incremental updates which sped up quite
jacek: though for oware most moves change board drastically
jacek: (still not bitboards)
jacek: here at 63 frame i solved the win https://www.codingame.com/replay/482774691
jacek: its not rocket science when at the end only 1 move is available
jacek: for alphabeta i would need to use search extensions and whatnot, but in best first minimax i have this for 'free'
MSmits: ah yes
MSmits: mcts solver does this too, though probably not as efficiently
MSmits: what helps me a lot in these cases is a 6 seed endgame db
MSmits: could have done the lookup at the depth of frame 79, which means I would have solved it much earlier
MSmits: probably around 63 also
Memo12334: in coders strike back. North is actually South right
Memo12334: nvm
Memo12334: stupid question
MSmits: it depends on whether you mean the geographic north or the magnetic north
Memo12334: http://chat.codingame.com/pastebin/e41803ae-12f2-48be-a1cc-7344407cd165
Memo12334: MSmits sir do u understand this function, I understand everything fully except when doing the 360 - a thing
Memo12334: I know a negative angle can be seen as positive by virtue of rotation
Memo12334: and the dy part
Memo12334: sry I should have asked this in discord , is any admin online
PayItForward: first day here trying to learn c. where should i start
PayItForward: i have no experience
Jrix: Go on the doc
PayItForward: I dont know what that is lol can i search for it
mr.mister: doc is short for documentation
PayItForward: ah thanks
Jaek: Can i share my contribution here ?
MSmits: Memo12334 sorry, that angle stuff always makes my head hurt. I usually try a few things until i find something that works
MSmits: sometimes i am able to work it out on paper, but i lose it fast
kvitebjonn: i cant find where to change my avatar, help
norbi97: hover it on your profile, then you can click change picture
itzblinkzy: markyy is it okay if you share your code? I failed miserably :pensive:
itzblinkzy: alright then :cry:
markyy: @itzblinkzy sure, just tell me which clash
markyy: I did
markyy: it's not readable though cause it's shortest code mode :sweat_smile:
itzblinkzy: Yeah it's fine
LinhT.Nguyen: yasuo change cai mode di
LinhT.Nguyen: reverse kia
markyy: did they fix the clash of code chat bug?
markyy: it used to disappear during the contests
zile151: i dont think they fixed the bug, but im not sure if thats a bug or a anti-collaboration feature lol
LordSahandii: hi
thanhphuoc: hello
Sevi1: hello
