Chat:World/2021-03-11
AGroatEatingGoat: C# can be good for speed, if that's what you're most comfortable with
AGroatEatingGoat: Python is one of the best, though, in both the speed and golf categories
AGroatEatingGoat: I'm (re)learning a lot about python from doing these!
AGroatEatingGoat: Python is literally executable psuedocode, it's just beautiful <3
void_main_sucks: Nooo not using namespace std; > : (
AGroatEatingGoat: bruh, void_main_sucks but you don't include a return statement in main
void_main_sucks: forgot to lol i just use the starting code
Aa_Ku_Ttp: hello world
CaptainKitten_322e: kim bor. yaxshimisizlar?
CaptainKitten_322e: qayerdansiz
Redbladexix: theres is nothing better at a boring job than to train ur brain by playin some pod race :D
void_main_sucks: sometimes i think some players are bots
derjack: oO
magaiti: do you sometimes think you are a bot too?
derjack: how many symmetries clobber has from starting position? just normal and rotate180?
MSmits: 28 moves
MSmits: 112/4
MSmits: simplest way is to just move each piece to the right and ignore down, left and up
MSmits: then you have all unique moves
derjack: im talking about board
derjack: if you rotate by 90, youll get white in left down corner. it seems not the same board anymore
MSmits: me too
MSmits: ah, then dont do the 90
MSmits: you can do 180
derjack: right now im doing 180
MSmits: you can also mirror diagonally
MSmits: I think
derjack: :tada:
derjack: so i may get 4 positions per move to label
MSmits: you may get 8
MSmits: the starting board has 4
MSmits: but maybe later boards have more
MSmits: you should just use the othello scheme
MSmits: its the same 8x8 squares
MSmits: so the same transformations apply. Just most of them aren't reachable
derjack: i think it may mess up the 'real' representation of game
derjack: because whites will be on entire different squares
derjack: maybe near endgames it doesnt matter
MSmits: yes they will when you transform them, but that doesnt mean the transformation is wrong
MSmits: it just means its not a reachable state
MSmits: this is true for every game, all states in the treespace are reachable, but not all states in the statespace are
MSmits: well almost every game I guess
MSmits: for example oware has many states where all pits have seeds
MSmits: but they are all unreachable except the start
MSmits: because after every move there is an empty pit
derjack: still gotta work on better inputs. fairly trained agent wins 15% vs mcts
darkhorse64: I am surprised that mcts works so well. tric gave up on ab and started his first mcts
derjack: mcts works better if you dont have any good eval
derjack: i wonder if clobber will become performant mcts-fest like uttt
darkhorse64: true but it turns out that all papers that I read were based on heuristics.
darkhorse64: However, these look quite difficult to bitboard. There will be a tradeoff between number of light/heavy rollouts
derjack: welp 10x10 is difficult to bitboard anyway
darkhorse64: In retrospect, 8x8 was a wise decision
derjack: until i solve it, yes
darkhorse64: I trust NN to bambozzle MCTS
MSmits: if you find the right heuristics for ab, you can also use them for a mcts with eval, or a jacekmax
MSmits: either way mcts could come out ahead
MSmits: or you could just keep the random rollout and use heuristics there
MSmits: depending on whether they are pruning heuristics or heavy rollout heuristics
MSmits: NN might not bamboozle mcts here depending on how important the math is. NN can simulate mathematical heuristics of course, but it doesnt do it more efficiently than the math itself.
MSmits: or you could combine the math with an NN... that would probably be best
YurkovAS: how to find 4 in row with pdep? for example: horizontal line from left to right. we check 3 left and check 3 right separately (i'm understand hot to do) or need only 1 time check?
MSmits: which game?
MSmits: yavalath or connect 4?
YurkovAS: connect4 or yavalath
YurkovAS: yes
MSmits: well its different
MSmits: i wouldnt use pdep/pext for connect 4 so...
MSmits: in yavalath what I do is this:
MSmits: in the first 1s, for each of the 61 hexes I generate a star pattern around it in bitform
MSmits: the start pattern goes 3 steps in each direction
MSmits: at max this will be 18 bits around it
MSmits: these 18 bits are all located somewhere on the 61 bit board
MSmits: Then I generate all possible combinations of these bits
MSmits: set/not set
MSmits: when I play a hex, i PEXT the star pattern
MSmits: and then do a lookup to see if its a win
MSmits: so the lookup handles all 6 directions at once
derjack: YurkovAS i used this to generate patterns https://github.com/skeeto/yavalath
YurkovAS: MSmits, derjack thanks!
MSmits: yeah, there's various levels of calculation vs lookup
MSmits: you can also do 3 lookups, one for each axis, then you use 6 bits, but have to do 3 PEXTs
MSmits: some players do this
MSmits: also YurkovAS, if you still want to know how to do it in connect 4. Use SIMD to check 4 directions at once
MSmits: so no lookups, no pext/pdep. Full calculation
darkhorse64: ^
MSmits: and if you're afraid of simd, just do it without simd first to get the bot working. That may motivate you to experiment with Simd after. It's not the hardest game to parallelize (othello is far harder)
YurkovAS: yes, easy rules, i'm try simd, thanks!
MSmits: np
darkhorse64: YurkovAS: write bitboarding for C4 first, do not jump straight to simd if you have no prior experience. Once your bitboarding code is OK, you will see more easily the parallel code patterns
MSmits: also lots of unit testing :)
MSmits: use your working normal win checking and then your simd version
MSmits: run them both for every move your bot makes, see if they are ever different and if so, produce an error
MSmits: I didn't actually do this, but I am an idiot and don't follow my own advice
MSmits: my method: Just do the simd thing, see the bot fail and stare at code for hours
darkhorse64: improved method: print the bit patterns and stare at them for a few minutes
MSmits: yeah
magaiti: staring intensely at machine code is the popularized method of debugging
MSmits: indeed
MSmits: yoyo
AntiSquid: MSmits how do you define you vars for bitboarding ? typedef unsigned long long U64; ?
darkhorse64: uint64_t
MadKnight: yea there's a built-in type for that
AntiSquid: ok looking at the differences
darkhorse64: it works also for VS
AntiSquid: so much info on bitboards for chess specifically, but not in mood to do chess right now
MadKnight: they added chess multi to CG ?
AntiSquid: community made
MadKnight: Automaton2000 we gotta go rekt everyone in chess
Automaton2000: now that's a good thing
AntiSquid: well recurz made and community approved
MadKnight: well if recurz made it....
MadKnight: no it's still the same
AntiSquid: odd though he didn't submit anything
MadKnight: probably wasn't interested
MadKnight: too common and too much solved
MadKnight: so actually not a surprise
AntiSquid: ah so you already have the #1 bot in your mind but you need to write it
AntiSquid: as expected of MadAsAJazzer
AntiSquid: MadKnight *
MadKnight: wait wat?
MadKnight: what is that about
AntiSquid: referring to how you improve your bots when the time comes ...
AntiSquid: hello new here ! welcome !
MadKnight: well i can probably analyze game strats better but #1 ?
MadKnight: ez
AntiSquid: MadKnight it's just a joke, ignore me. And yes your exact words from another time, long forgotten now.
MadKnight: but #1 ?
MadKnight: AntiSquid i mean i didn't understand the joke
MadKnight: still no idea where it is
derjack: :drum:
derjack: and this is chess960
derjack: https://www.codingame.com/multiplayer/bot-programming/chess
derjack: bitboarding is relatively easy. king checks and castlings are meh
AntiSquid: oh my D bot still wrecking noobs
derjack: oh your
6cdh: My bad English even can't understand the problem description.
derjack: oO
AntiSquid: which problam
derjack: and hes from usa. hmmm
AntiSquid: he migrated and didn't learn the language i guess
zedd001: Hello everyone, I just want to ask where do you learn advanced programming?
AntiSquid: on your own? you can do it here too
zedd001: Thanks for your sharing so much !
MadKnight: what is advanced programming zedd001 ?
MSmits: it's a natural follow-up on basic programming MadKnight
MadKnight: but to what degree?
MSmits: 7
dirty_code: 42
MadKnight: well that's easy just go finish 6
MadKnight: Automaton2000 imagine asking what goes before 7
Automaton2000: it's better to use a neural net
MSmits: anyone here that solved bender 4 100%?
MSmits: I have an approach I want to bounce off someone
MSmits: before i code it
AntiSquid: no but i think i have the solution figured out ? :P
MSmits: allright, so I want to do something like this:
MSmits: the boulders are walls, at least at first
MSmits: to make it easy
MSmits: you can change the map with the levers
MSmits: so i make a tree of map changes as opposed to a tree of movement
MSmits: the parent is the first state, the children are new mapstates that are reached by doing the toggles that are within reach
MSmits: (without actually crossing other toggles)
AntiSquid: ah i am thinking of a tree of moves, just keep looking till solution is found
MSmits: thats too expensive i think
MSmits: you get cycles and such
MSmits: my way only requires a tree with a few 1000 nodes at most
MSmits: then when you find a path, all you need to do is bfs from one node of the tree to the next
MSmits: if you know a toggle is reachable by bfs without crossing other toggles it doesnt really matter what the route
MSmits: is
AntiSquid: damn ... you make me want to look into it and solve it now lol, i am quite sure that in worst case scenario i can use some conditions to ignore some moves
MSmits: so basically i have a high level BFS where i chain toggles together instead of squared
AntiSquid: or break the path into multiple searches
MSmits: then a low level BFS to connect the toggles
MSmits: i do a floodfill to find out which toggles are reachable from a current toggle position
MSmits: (no matter the route used)
AntiSquid: my reason for thinking this will work is the ooc bot i have which is highly unoptimized and the amount of stuff it can look for, think this problem is less demanding
MSmits: oh I guess it is less demanding, but i want to expand it to also use boulders later, which are also map changes, like toggles. But they branch much much more
MSmits: this is the approach i used in hypersonic and it was very fast there
AntiSquid: i mean your way is basically same as breaking path into multiple ones to find otpimal one
MSmits: in hypersonic, my bomber dude is every where at once and only gains a position when he changes the map
MSmits: AntiSquid yes it is
AntiSquid: and actually you can have tree of map states where each state has trees of paths or whatever ugh ... then prune / ignore less promising trees
MSmits: imho the way to go for bender4 is basic solution no boulders -> add boulders to get better solutions -> add functions
MSmits: AntiSquid thats the thing, i dont generate the paths
MSmits: i only generate the changed map states
MSmits: i dont care how bender gets to the next spot, only that he can (found by floodfill)
MSmits: i only generate the path later, when fry is reached
MSmits: so it's really a simple tree, not a tree of trees
AntiSquid: ok how about finding all paths for all possible switch / bolder / stuff change and then finding out which combination of paths leads to solution ?
AntiSquid: i mean you precompute paths then look for combo
MSmits: that may work, but i dont see why you need a precomputed pat
MSmits: path
MSmits: maybe a precomputed floodfill for each map state
AntiSquid: less floodfills
AntiSquid: yes that's what i mean
MSmits: ah ok, well thats a neat optimization
MSmits: I want to try to at least solve it this weekend. I have some other stuff to do as well but hope to find some time
MSmits: only 41 people have solved it 100%
AntiSquid: is that your motivation ?
AntiSquid: low solver count ?
MSmits: also i told a student to do it
MSmits: without realizing how hard it was
AntiSquid: ah that changes everything
MSmits: so i kinda feel like i have to. And I think it is interesting
AntiSquid: idk, when i finally get around doing it i will start it with just first test case and then add rules for 2nd ... then 3rd etc .. and then a big rewrite at the end if needed :P
MSmits: yeah that often makes sense
MSmits: but i got an algorithm in mind i am sure works, so i want to try it as a whole
MSmits: then optimize it, your suggestion is good
MSmits: i did that in kutulu
MSmits: calculate all that stuff beforehand
AntiSquid: which suggestion ?
AntiSquid: ah
MSmits: I think your way is what i suggested to the student
MSmits: plot a direct path to the target
AntiSquid: i only said that because you said it takes too long, will try it headon without precomputation first probably :D
MSmits: if there are obstacles, make a direct path to the toggle you ned
MSmits: if there are obstacles to that, make a direct path to anothe toggle etc.
MSmits: until you get there
MSmits: thats another way to do it
MSmits: but i think he may get stuck on cyclic toggles, where you untoggle something important when moving to a different toggle
MSmits: sometimes they are in the way
AntiSquid: this one has functions
MSmits: that too, but thats like loops in code of the ring
AntiSquid: i think it was the function bits you can encode reason why i left it for later
MSmits: its just intended to make the solution shorter
AntiSquid: seems very time consuming
AntiSquid: 889 CP max, it was lower back then too, maybe 200 max
MSmits: I think i can do the simple solution in a few hrs, then i'd be rank 42 at least
MSmits: so it's worth it
MSmits: brb
AntiSquid: i stopped improving my fireworks / pengu for now, need full rewrite to progress so kinda aiming for lower hanging stuff instead
AntiSquid: ah right unoptimized solution shouldn't take that much, but with only rank 42 you get a drop in a bucket as CP
AntiSquid: top 10 min @ 2,243
MSmits: thats a lower limit, assuming the worst possible solution though
MSmits: i can easily adapt to find shorter solutions and get somewhat better rank
MSmits: if i include boulders (lots of coding) it should get a lot better
MSmits: then with functions, you can get any rank you want if you do it well enough
MSmits: probably the validators are known as well, so you might be able to make use of that
AntiSquid: "you can get any rank you want if you do it well enough"
ya just this . how much time will you spend :P
MSmits: till i get bored
AntiSquid: or are willing to spend
MSmits: i did code of the rings till top 20, which is quite good for a game with so many players
AntiSquid: not asking you superficially, just generally speaking
MSmits: ah yes
AntiSquid: maybe i should do another hex game ?
MSmits: do you like boardgames?
AntiSquid: on CG ? no
MSmits: ah ok
MSmits: well coders of the carribean is not really a board game and is hex, but you probably did that already
AntiSquid: maybe i should at least try to do something on my own game
MSmits: yavalath is a boardgame so doesnt qualify then
MSmits: oh cool, did not know you were working on something
AntiSquid: pengu is different because you eliminate tiles and that changes a lot
MSmits: yeah
MSmits: but boardgame-wise it's just a bit mask
AntiSquid: also check the amount of hex / line
AntiSquid: they are not the same, that cause me some headaches
AntiSquid: figured something new out just to make it work :P
AntiSquid: initially thought i have to check if row is odd or even
MSmits: makes sense
AntiSquid: and i have the board in one array, so if i rewrite i should use uint64
MSmits: probably
AntiSquid: ideally .
MSmits: unless you do a minimax
MSmits: then do/undo works fine on any type of board
AntiSquid: no won't do that
MSmits: assuming you dont use transpositions, hashing complicated gamestates is expensive
MSmits: uint64_t helps a ton with hashing
AntiSquid: i have minmax on some games and i feel like i could have done much more with mcts
MSmits: i guess it depends on where your skill/knowledge lies
MSmits: i suck at minimax, compared to what i can do with mcts
Astrobytes: It depends. If minimax is performing well but you've run out of ideas, some kind of EPT MCTS might be better.
AntiSquid: even make some custom selection of my own, doesn't matter if it's weird
Astrobytes: You do have to invest some effort in an a/b to make it good
MSmits: yeah
MSmits: btw gotta go, class starts in 5 mins (as a student, not teacher)
MSmits: dont want to make teach mad :P
Astrobytes: FP?
MSmits: nope
MSmits: educational research
Astrobytes: Oh right
MSmits: FP is every other friday, next friday again
MSmits: ttyl!
Astrobytes: later
AntiSquid: see you next friday :P
Astrobytes: doesn't quite work as well as Tuesday or Thursday eh
Astrobytes: working on penguins?
AntiSquid: no
Astrobytes: yava?
AntiSquid: he was comparing coders of caribbean to pengus and i was saying it's not the same
AntiSquid: tiles keep getting removed that's major diff
Astrobytes: Oh, yeah, quite different indeed
Astrobytes: And penguins is a board game
AntiSquid: and really odd board shape for pengus
AntiSquid: you get 8 hex for odd 7 for even
AntiSquid: very important detail
Astrobytes: Only in your indexing
Astrobytes: I like penguins, good game. Must get back to it soon.
AntiSquid: maybe yinsh if i do another hex game, it looks weird and i don't really get it
derjack: are yo hex excited
Astrobytes: It's quite complicated compared to some others, but not to hard to get if you spend some time learning it. Haven't a clue for a good algo though
derjack: still need to make sim for it
Astrobytes: Yeah same
derjack: no boss so it cant become potw
Astrobytes: Not necessarily a bad thing - do you think explaining yinsh to every Tom, Dick and Harry on cg every 5 minutes for a week? :D
Astrobytes: *think you'd enjoy
Redbladexix: looking for the most beginner friendly bot games. To advance on my questmap need another one to reach bronze league.
derjack: tryangle is potw and yet no one asks about it
derjack: coders strike back
Astrobytes: True... are many people playing it?
Astrobytes: TC I mean not CSB
derjack: 24 now. before it was less than 20
derjack: not much boost
derjack: :(
derjack: AutomatonNN where is the creator
AutomatonNN: with page in the top right
Astrobytes: That must be the Disappointment aspect.
AntiSquid: Redbladexix botters of the galaxy
AntiSquid: oh wait it asks for legend at the end
AntiSquid: probably CSB and you can't play any community games for this lol
AntiSquid: CSB easiest legend do that Redbladexix, rules change in gold
Redbladexix: CSB already at bronze need another one for this quest
AntiSquid: what quest is it ?
Redbladexix: reach bronze in 2 games
AntiSquid: oh you need later silver in 2 games
AntiSquid: hmmm
derjack: ultimate tic tac toe :?
AntiSquid: is it 3x3 until silver ?
derjack: is 3x3 only in wood
AntiSquid: in silver it's more complicated, more headaches
AntiSquid: do botters of the galaxy until silver should be very easy
Redbladexix: ok, will try. thx for the help
AntiSquid: mean max for silver also very little coding actually
AntiSquid: and maybe code royale too? but can't remember anymore for sure
KelvinAndHubbles: tron bike till silver isn't too hard too
MrPizzadur: yes
MrPizzadur: Hi everyoone!
MrPizzadur: Pas très sympa mec
Taubeth: hi everyone, am I allowed to use snippets(or bigger parts) of code from stackoverflow during clash of code challenges?
MrPizzadur: yeno
BoatBuilder: you can, but it's less fun than doing all by yourself
dirty_code: is tychkord bot?
LiudasStaniulis: yes
Butanium: dirty_code: you can check it on it's profile
KelvinAndHubbles: beat the boss 9/9 times and won 29/32 battles, yet in a wood league still says I'm worse then the boss :/
jacek: what game
KelvinAndHubbles: Botters of the Galaxy
KelvinAndHubbles: just in wood league 5
KelvinAndHubbles: have a basic bot that pretty much just buys damage and hits
KelvinAndHubbles: just kept submitting and eventually got it :D
AntiSquid: nice
ill-be-waiting-outside: hi
HunterEhrenfeld: hi
ill-be-waiting-outside: how are you
Saphine: #clash
ill-be-waiting-outside: thaks
FrancoisLetourneau: did anyody ever had a technical assement fof job using codingame?
ill-be-waiting-outside: kinda
ill-be-waiting-outside: mad it was hard for me
FrancoisLetourneau: I have to do 1, I got the link by email, it says there are 40 questions in the test and it takes on average 25min to do.
FrancoisLetourneau: How can this be? It is for a junior position
ill-be-waiting-outside: idk
HunterEhrenfeld: youre have a technical assessment using this site? lol ask antisquid for answers lol
FrancoisLetourneau: what/who is antisquid?
BlaiseEbuth: Why :no_entry_sign::squid: should know the answers ?
ill-be-waiting-outside: idk maybe he knows
BlaiseEbuth: Anyway, FrancoisLetourneau there is a lot of mcq in the 40, so taht is pretty quick.
BlaiseEbuth: *that
ill-be-waiting-outside: true
AntiSquid: no worries, i wouldn't tell you
AntiSquid: https://www.codingame.com/ide/puzzle/botters-of-the-galaxy
AntiSquid: best way to prepare FrancoisLetourneau
FrancoisLetourneau: yeah that's what I thought
BlaiseEbuth: Oo
jacek: ?
BlaiseEbuth: Sorry, that's your catch phrase.
ill-be-waiting-outside: how do you do Pythagorean Theorem it is hella hard
BlaiseEbuth: a²=b²+c²
jacek: you need it in tryangle catch?
AntiSquid: a^2 + b^2 = c^2
jacek: why xoring by 2?
AntiSquid: that's power
BlaiseEbuth: xoring is powa!
ill-be-waiting-outside: ik that
CryptoZenith: xoring vs power
ill-be-waiting-outside: but its something else and thats it thatnks
BlaiseEbuth: clear question -> clear answer
CryptoZenith: unclear question -> unclear answer
AntiSquid: why is clear question a node
CryptoZenith: what
CryptoZenith: what's the lightning symbol next to your name
CryptoZenith: oh wait
CryptoZenith: nevermind
AntiSquid: pokeball ?
CryptoZenith: horseshoe?
CryptoZenith: i have a small monitor
AntiSquid: woah
BlaiseEbuth: I choose you AntiSquid!
CryptoZenith: takes up like 3 pixels
BlaiseEbuth: It totaly can be a pokemon name.
CryptoZenith: anti-squid
CryptoZenith: like a squid, but not
CryptoZenith: whoa anitsquid is 11th on the global leaderboard
CryptoZenith: *anti
BlaiseEbuth: tentacles...
AntiSquid: "The true motorcyclist – the Anti-Squid – examines these crashes, and attempts to fix things."
AntiSquid: quote from article linked
ill-be-waiting-outside: cool
jrke: is there any way i can identify traps in crystal crush?
JLukeSkywalker: i dont think so
JLukeSkywalker: entityType: http://chat.codingame.com/pastebin/63c86544-ede3-4d3b-a7af-7228cd3e0b5b
JLukeSkywalker: that is whats given, in addition to just ore values
ET_BB-8: if the enemy stops at the base for 1 turn it could mean the have picked up a trap
darkhorse64: the whole game revolves around traps and efficient farming
jrke: ET_BB-8 what if he picked radar not trap?
ET_BB-8: thats the problem with that system
ET_BB-8: but its better not digging up a radar position than digging up a trap position
ET_BB-8: unless your going to kamikaze
jacek: is that utg?
BlaiseEbuth: ya
BlaiseEbuth: name changed for the multi
darkhorse64: or if you dig a potential trap, make sure you will destroy one opponent bot at the same time
jacek: or delete your account
BlaiseEbuth: Oo
ET_BB-8: in later leagues some players wait at the base for 1 turn also to throw off those systems
jacek: it was fun to watch the replays
BlaiseEbuth: Advanced kamikaze strategy was my greatest idea during the contest...
darkhorse64: even better, they stop at several holes to puzzle you when trying to guess traps
AntiSquid: have to fix that bot at some point . was close to legend
AntiSquid: oh i dropped ...
AntiSquid: now free pushes ?
AntiSquid: no*
ill-be-waiting-outside: cool
JohnnyHotBody: Hello
JLukeSkywalker: hello
JohnnyHotBody: how u doin
ill-be-waiting-outside: good
JLukeSkywalker: decent
JohnnyHotBody: cool
JLukeSkywalker: stuck at work, watching the other cyclists out the window on the road
JLukeSkywalker: wishing i could take the day off
ill-be-waiting-outside: me too
JohnnyHotBody: Same
JLukeSkywalker: at least this website counts as work some days
JohnnyHotBody: Well I got to go
JohnnyHotBody: Bye
ill-be-waiting-outside: ok
SoumyaCodes2020: hi
ill-be-waiting-outside: hi
Lachrymosa: hi
ill-be-waiting-outside: wyd
Lachrymosa: Rock Paper Scissors Lizard Spock
ill-be-waiting-outside: holy shit i love that game
jacek: are you winning son
Lachrymosa: Theres a nifty little puzzle in easy for it
Lachrymosa: I'm close
ill-be-waiting-outside: really
ill-be-waiting-outside: i have a test to do
ill-be-waiting-outside: its more like T-T
jacek: UwU
ill-be-waiting-outside: yep
JLukeSkywalker: lol, just demolished the wood leaderboard in crystal rush
KelvinAndHubbles: is it possible to solve the puzzle of the week?
jacek: you need to advance the league
JLukeSkywalker: darn, almost made it to silver completely ignoring the existance of emps
Westicles: Huh, somehow my cout << "0" << endl; advanced to wood1 in connect4
JLukeSkywalker: yeah, i think he is working on fixing the boss
jacek: hm?
JLukeSkywalker: think it might have been darkhorse, said he was trying to fix the connect 4 wood bot
JLukeSkywalker: the current one is really bad
jacek: blame the approvers
darkhorse64: No, I did not write the boss
JLukeSkywalker: ik, but the other day thought you said you were writing a new one
Astrobytes: I think that was a joke
darkhorse64: It was a half serious joke
JLukeSkywalker: maybe was a joke, but the boss is a joke
JLukeSkywalker: someone does need to fix it
struct: connect 4 is just depth 3 minimax
struct: iirc
struct: At least is what I remember from when I saw the code
Astrobytes: It's a what, depth 2?
Astrobytes: Oh 3
JLukeSkywalker: and people are beating it with random
JLukeSkywalker: maybe there is a bug in it
Astrobytes: I can't remember. I remember it looked really inefficient.
darkhorse64: tbh, I did not checked it
Astrobytes: If it's that bad then I personally nominate darkhorse64 to write a new one :P
darkhorse64: You know, it's hard to nerf a bot
Astrobytes: This is true. Just copypaste your alphabeta from Clobber
Astrobytes: (not the eval perhaps :P )
darkhorse64: I could also copypaste euler's C# bot
Astrobytes: Is that any good? I'm sure I beat that by hand
ZarthaxX: rude
darkhorse64: MCTS with a solver.
Astrobytes: Yeah, but it was weak iirc
Astrobytes: ZarthaxXamaton
darkhorse64: If it beats cout << "0" and random, it's OK
ZarthaxX: Astronaut
ZarthaxX: howdy
Astrobytes: darkhorse64: that's fair
Astrobytes: How's it going ZarthaxX?
imathwee: how you gunna find the highest rated playa here
MadKnight: js even crashes when u start a number with 0
ZarthaxX: all gud
MadKnight: Automaton2000 imagine crashing from x=025;
Automaton2000: i'm pretty sure you can do all that stuff
ZarthaxX: trying to learn new stuff, was cleaning the house until now
ZarthaxX: you astrobyto
MadKnight: hey ZarthaxX
Astrobytes: Cleaning the house is new Zartho? :P Yeah, same shit different day here lol
ZarthaxX: hey mado
ZarthaxX: Astrobytes NO LOL i just said 2 different things
Astrobytes: ZarthaxX I know! :P
ZarthaxX: *today i learned what a broom is guys!*
Astrobytes: #winning
ZarthaxX: so same shit eh
ZarthaxX: fair fair
Astrobytes: Yep
ZarthaxX: what's up with that clobbeer game?
Astrobytes: Not my bot anyway. It's last :D
Astrobytes: Interesting game.
ZarthaxX: real or invented
jacek: whats the difference
jacek: all games are invented
ZarthaxX: wanted to know if it was a port
darkhorse64: It is
Astrobytes: Yeah, abstract board game, been around for 20 years
ZarthaxX: aight ty guys
ZarthaxX: also looks super simple, nice
Astrobytes: Yeah, very easy sim.
ZarthaxX: i bet right
Astrobytes: The rest however...
Tuo: looks can be desieving :D
Tuo: my python bot was so slow it was useless
ZarthaxX: also, random question but where is toad?
ZarthaxX: has he been here the past couple days?
Astrobytes: Long story best saved for another day ZarthaxX
ZarthaxX: bad or nah?
Astrobytes: Well, he's fine but it's a crappy thing.
ZarthaxX: can you tell me another day?
ZarthaxX: ;(
Astrobytes: Yes. Not today.
ZarthaxX: ty
ZarthaxX: hope he is doing well
Astrobytes: All good afaik
ZarthaxX: and another random thing, i got banana plants in my house, and the last one had 110 bananas approximately
ZarthaxX: 27kg of bananas
Astrobytes: Dude
darkhorse64: He is OK but he has stepped down for a while
ZarthaxX: aight aight , good then
Astrobytes: 27 KILOS OF BANANAS
Astrobytes: GROWN IN YOUR HOUSE
Astrobytes: Wtf am I doing in this country
ZarthaxX: LOL
ZarthaxX: it's hard to keep them like.. not rotten
ZarthaxX: eating all of them before they rot is hard, we freeze them but it's still a fuckton
Wilster: you can make banana n.ut bread with ripe ones
Astrobytes: Yeah, make some banana pureé and stuff, less space when you freeze it
Astrobytes: And sell the skins to vegans, they like that kind of thing.
Astrobytes: Or throw them in front of unsuspecting people and film the cartoon slip
jacek: bananas with coke?
Astrobytes: (the last two were jokes)
jacek: its so hard to slip on banana
ZarthaxX: oh we never tried the banana bread one, nice
ZarthaxX: neither the pureé, we did the ice cream one
ZarthaxX: Astrobytes LOL the skins are thrown but to make compost
Astrobytes: Ah that's cool
Astrobytes: Can you make some kind of alcohol from bananas? Or at least use as a flavouring.
Astrobytes: You can make banana wine and beer apparently.
Astrobytes: And therefore banana brandy
Astrobytes: Stop this computer nonsense and open a brewery + distillery
ZarthaxX: :rofl:
Astrobytes: :grin:
ZarthaxX: isnt it kind of hard to achieve a good wine/beer?
Astrobytes: Well, yeah. Takes practice. You have many bananas.
ZarthaxX: and who will be the victims that try it?
Astrobytes: :P
ZarthaxX: family? :D
Astrobytes: Ofc!
ZarthaxX: *and that's how you poison all your beloved ones*
ZarthaxX: i think the ice cream one is safer hehe
Astrobytes: Start with beer and wines, don't try and distill anything if you want to keep your eyesight/loved ones
ZarthaxX: i bought a chocolate licour in the holidays, now that i reemember
ZarthaxX: good one
ZarthaxX: okey Astrobytes :rofl:
ZarthaxX: bananas is the future
Astrobytes: Sell them to Elon Musk
ZarthaxX: DaciaN0 noob here
ZarthaxX: Astrobytes i will make him tweet about my bananas
ZarthaxX: easy money
ZarthaxX: bananastocks raising
Astrobytes: BananaX
ZarthaxX: gotta find a good cryptocoin name
Astrobytes: SparthanaxX
ZarthaxX: :rofl: wtf
Astrobytes: ooh wait: BanX
ZarthaxX: nailed it :O
Astrobytes: Yeah, that's actually decent lol
ZarthaxX: ye :P
Astrobytes: Then when you're uber-rich you can get a gf called Dirt or something like him
Astrobytes: (Elon's gf is called Grimes, grime = dirt, if anyone needed an explanation)
Astrobytes: btw I am not beating the boss in Connect4 with random
ZarthaxX: oh i didnt get it at first lol
ZarthaxX: deception
Astrobytes: Maybe I random wrong
ZarthaxX: change your distribution
jacek: you can beat the boss if you write only one column
Astrobytes: really? lmao
ZarthaxX: astro :sweat:
UnnamedCodinGamer: I have been looking at bitboard stuff for connect4 and all of them refer to 6*7 board
jacek: 63 bits still fits
UnnamedCodinGamer: I would be nice if there was a site with basic implementation for different games
UnnamedCodinGamer: but i have to do it myself
Astrobytes: Yeah, all the stuff you see about using the spare bits for this or that you can just do with some masks
UnnamedCodinGamer: I am thinking in this direction
Astrobytes: Don't overthink it
Astrobytes: It's just the same as every other bitboard game
Astrobytes: *bitboardable
UnnamedCodinGamer: I lack experience
UnnamedCodinGamer: I guess at some point it becomes relatively easy
Astrobytes: Some stuff yeah, but it always requires a degree of care!
Ordonnateur: We can register to the codinggame spring challenge 2021 up to when ? up to last minute ?
jacek: yes
Astrobytes: Up to any point throughout the challenge Ordonnateur
UnnamedCodinGamer: the other thing about bitboards is that once you try them it seems there is no way back
Ordonnateur: (I will register later so, better than forgetting about it then being sad having a 0score on my history ;p)
Astrobytes: :)
Astrobytes: Yes UnnamedCodinGamer, for some games to be competitive it's really a requirement anyway
jacek: its easier for me to bitboard stuff than to use good old 2d array nowadays
UnnamedCodinGamer: jacek yes, especially move generation
UnnamedCodinGamer: in terms of performance I was actually surprised to see that in breakthrough my array implementation was a bit faster than the first bitboard implementation
KiwiTae: ZarthaxX o/
jacek: move generation is one thing. eval stuff like counting is also fast
ZarthaxX: hey kiwo :*
UnnamedCodinGamer: then I implemented the move extraction using ffsl and it got 2 times faster
UnnamedCodinGamer: jacek, that too yes
jacek: ffsl/
jacek: hmm looks like ctz or clz
UnnamedCodinGamer: yes, they are similar
ZarthaxX: first set bit for long
UnnamedCodinGamer: I did not know about them then
struct: Returns one plus the index of the least significant 1-bit of x, or if x is zero, returns zero.
UnnamedCodinGamer: struct, exactly
UnnamedCodinGamer: in terms of performance i did not find it worse than __builtin_ctzl, but I may be wrong
Astrobytes: oh nice
UnnamedCodinGamer: actually darkhorse64 and you Astrobytes suggested that
jacek: probably came from msmits
MSmits: nope
jacek: oO
MSmits: always use ctz
Astrobytes: ffsl? No that's a new one on me
UnnamedCodinGamer: I meant move extraction, not the particular method
Astrobytes: Ohhh right, gotcha
MSmits: yeah i use ctz/ctzl for move extraction
MSmits: ffsl also works
Astrobytes: Never used ffsl
MSmits: probably on compilation it works out to be equally fast somehow, compilers are smart
Astrobytes: Any particular advantage to it?
MSmits: i think it wont break on 0
MSmits: clz has undefined behavior on 0, not sure about ctz
jacek: well code is written it will never reach 0
MSmits: exactly
MSmits: thats why i dont bother with it
ZarthaxX: those instructions compile to a single assembly one
MSmits: i like my stuff to break when i screw up
Astrobytes: I guess could be useful in some situations but I'll stick with ctz/clz
Astrobytes: on here
UnnamedCodinGamer: it is nice if it breaks, and tells you so
MSmits: c++ telling you whats wrong?
Astrobytes: What sorcery is this????
Astrobytes: You must find out where you were wrong yourself and hit yourself with a c++ branch 100 times.
ZarthaxX: c++ branch LOL
Jake_Dubz: hi guys i am new to coding:grin:
Astrobytes: :P
MSmits: arent we all, to some extent
MSmits: hi
ZarthaxX: hi new to coding :grin:
Astrobytes: Hi new to coding, we are not new to coding
ZarthaxX: worst mod :P
Astrobytes: I am not!
Astrobytes: AntiSquid is way worse, and don't forget about BlaiseEbuth!
darkhorse64: defensive programming is a very bad practice performance wise
ZarthaxX: :rofl:
ZarthaxX: defensive? what
Jake_Dubz: i am new to coding, probably not you, but meeee
Astrobytes: Defensive in #debug
darkhorse64: checking always whether assumptions are wrong
ZarthaxX: ah ye branching is super bad :D
MSmits: I start doing that when i cant find a bug after X hours of staring
AntiSquid: Astrobytes ?
MSmits: i drop in like 10 conditional cerr statements and thats usually enough
Astrobytes: asserts and if (I'm fking up here) go in #debug mode
darkhorse64: that's debugging. It's different
ZarthaxX: astro knows :D
MSmits: well it's all i do
MSmits: never used an assert
Astrobytes: darkhorse64: the point being that's not the code you use in your main program
ZarthaxX: use asserts D:
darkhorse64: I know
Astrobytes: You never used assert MSmits?!
MSmits: nope
ZarthaxX: this guy is crazy
ZarthaxX: someone hold him
**ZarthaxX brings the c++ branch
AntiSquid: why use asserts? you should know what your program does
**Astrobytes holds MSmits and lets ZarthaxX beat him with the C++ branch
MSmits: I dont have any strong reason not to us eit, i just never tried
MSmits: lots of things i never did with c++
**ZarthaxX starts beating Msmits with the branch nonstop
MSmits: i never made a file with .h at the end
Astrobytes: AntiSquid: we're only hooman
ZarthaxX: MSmits :I
ZarthaxX: *shit*
jacek: oO
Astrobytes: Oo
AntiSquid: MSmits if you already have some efficient way to to debug you probably still don't need it
MSmits: possibly i could save some time if i did use it, i dunno
MSmits: I am very noob with C++
AntiSquid: oh this again ...
Astrobytes: It's in c too
ZarthaxX: lmao
Astrobytes: Honestly it's a good sanity check
MSmits: it's just a very tiny part of c++ i use, 1 file solutions with very few features from c++
ZarthaxX: im tired of doing this already astro
AntiSquid: remember that c++ meme with a guy learning c++? and then discovering immortality elixir, then time travel and then going into the past to kill his former self before he could waste time on learning all of c++ ? MSmits
**ZarthaxX gives c++ branch to astro
MSmits: nope, dont know that meme
AntiSquid: ya
**Astrobytes beats the **** out of MSmits to probably no real positive outcome
MSmits: torture won't make me code better :P
Astrobytes: :D
AntiSquid: https://img-comment-fun.9cache.com/media/aKxWNeb/akm0d6Nr_700w_0.jpg MSmits
ZarthaxX: GIVE IT TO HIM ASTRO, YE
ZarthaxX: MSmits we will see
**ZarthaxX brings the c branch
MSmits: I think this is true for any language btw. Euler is horrified about how i use C# also
Astrobytes: AntiSquid: but that's a meme about the 'teach yourself C++ in 21 days' books
AntiSquid: same end result
ZarthaxX: LOL squido
ZarthaxX: MSmits how do you use it?
ZarthaxX: i get it tho, the use you give to those lnaguages is purely for bots
MSmits: like a c++ without segmentation faults, very simply, using very few features
ZarthaxX: doesnt require you to know a lot
Astrobytes: Let's just say, generics don't get a look-in
ZarthaxX: thought so haha
ZarthaxX: linq, what is that? :D
MSmits: yeah i barely ever use linq
MSmits: when i do, its because i copied some code from somewhere
MSmits: but doing FP class, so maybe i'll start using it more
Astrobytes: And you declare strange lists
MSmits: how do you mean strange?
MSmits: it just normal lists
MSmits: objects, ints etc.
MSmits: but generally I do it the same way as in C++ and just declare arrays that are large enough
jacek: weird names
Astrobytes: List<T>
AntiSquid: well higher contest rank results means either linq is bad or best to focus on what matters, not how ugly the code is
MSmits: linq is good for when coding speed matters and performance does no
MSmits: t
MSmits: would be great in code a la mode for example
Astrobytes: List<T> is what you don't use
MSmits: i dont, no
MSmits: i just know the type of the list
jacek: :scream:
Astrobytes: You miss out on the whole power of generics
MSmits: i dont template in c++ either
MSmits: yep
Astrobytes: Heathen.
MSmits: its not like i dont understand it, i just dont have practice with it
struct: in cg it doesnt have much use tbh
MSmits: these bots are all fairly short and at some point it just works
AntiSquid: wait c++ template programming is like a programming language of its own
Astrobytes: Practice! It can only improve your skills ;)
Astrobytes: AntiSquid: C++ templates don't have to mean full on template metaprogramming insanity
therealbeef: compile time limitations mostly prevent any serious template meta programming
therealbeef: (ive tried)
Astrobytes: Masochist!
therealbeef: i'd be unfair anyway
MSmits: why is it unfair?
AntiSquid: you take a step in that direction, why would you ever stop
MSmits: i dont see how i would get more performance using templates?
therealbeef: you can do calculations at compile time, which other languages cant do
MSmits: hmm
MSmits: I see
MSmits: but in many cases that doesnt matter
Astrobytes: because you can simply template functions and simple things without going nuts AntiSquid
therealbeef: i calculated part of the uttt eval function at compile time
MSmits: the first 1 second in most bots is already overkill
Astrobytes: There is templates and then there is template metaprogramming
AntiSquid: i know
MSmits: because you need calculation time later on in the game, when moves have been made, not in the first second
MSmits: exception is games where you use the full 1 second to precalculate endgame db or something
Astrobytes: <insert time banking comment here>
AntiSquid: 1 second to decode an entire book of most important moves and positions, think about it
MSmits: actually in oware i dont decode, i calculate
MSmits: i decode the opening book though
MSmits: 9 seed db in 500 ms or so and a little bit of time for the 2-3k moves from opening book
Tuo: uint64_t f = 1 << i; such a simple line of code, the problem must be somewhere else
AntiSquid: does this count as obfuscation ? it does right?
MSmits: what does?
jacek: Tuo 1ULL << i
darkhorse64: tuo 1ULL << i
Astrobytes: Tuo: 1ULL << i
AntiSquid: books MSmits
AntiSquid: tuo : 1ULL << i
MSmits: my oware book is very readable atm
Astrobytes: :rofl:
MSmits: havent needed to compress
AntiSquid: ah k
MSmits: {5{5{1{2{4{2{2 ... etc
Tuo: i figured it out in the end, it just took way too long
MSmits: it's a tree
AntiSquid: what are you working on tuo
MSmits: start with move 5, on move 5 from opponent, you play 1 etc.
Astrobytes: Tuo: we've all been there
MSmits: yeah
MSmits: i still do that
Astrobytes: I would bet that tuo is working on Clobber
MSmits: it's really annoying
MSmits: VS warns me though
darkhorse64: Read: we all made the mistake
MSmits: get a green squiggle
Tuo: y globber but im not that far :S
MSmits: clobber is a nice game to start doing bitboards
UnnamedCodinGamer: I wasted 10 minutes on it the other day
MSmits: i wasted hours on this before
MSmits: you're lucky UnnamedCodinGamer
CryptoZenith: its not wasting if you enjoyed it
MSmits: but then again, i dont assert so..
Astrobytes: MSmits :P
UnnamedCodinGamer: I was trying to print a bitboard and it printed half of it
MSmits: ah yeah, typical
MSmits: thats sometimes how i find otu
Astrobytes: lol, there should be a list of common bitboard issues
darkhorse64: LeRenard joins the party
Astrobytes: Yay nice one
MSmits: i sometimes shift a 32 bit 32 places to the left and then or it to a 64 bit
MSmits: thats a similar error
MSmits: does nothing
ZarthaxX: what party¡
MSmits: fox party ?
darkhorse64: Top 4 Clobber in 0.5
AntiSquid: they call clobber a party
MSmits: Sweeper, no swiping!
MSmits: :fox:
jacek: clob clob clopper
ZarthaxX: hell of a party brothers
Astrobytes: I think LeRenard just needs another submit, boss is probably boosted. His game history looks good
LeRenard: I try a second submit with a updated weight
Astrobytes: Looks a lot stronger LeRenard
jacek: alpha beta?
MSmits: a boardgame leaderboard is not the same without LeRenard on it
Astrobytes: ^
LeRenard: MCTS
MSmits: cool, you usually do ab right?
LeRenard: First good result with MCTS
MSmits: ahh ok
LeRenard: I try a MCTS on UTT but with bad result
MSmits: your ab bots tend to be pretty strong. Yavalath one is hard to beat
Astrobytes: It really seems MCTSy this game
MSmits: LeRenard your uttt problem is probably optimization + lack of heuristics. Add some teccles magic
LeRenard: But I copy paste may UTT MCTS in Clobber.
LeRenard: Yes I could retry on UTT
UnnamedCodinGamer: what is teccles?
MSmits: play 0 on empty 0 board
MSmits: 1 on empty 1 board
MSmits: 2 on empty 2 board
MSmits: etc.
MSmits: send opponent to same board
BlaiseEbuth: Oh. I was pinged an hour ago...
darkhorse64: It's a Clobber chat only. Where is your bot ?
Astrobytes: sorry BlaiseEbuth, just saying that you're a worse mod than me to ZarthaxX :P
UnnamedCodinGamer: MSmits, thanks
MSmits: sometimes teccles is a losing move though, if you check karliso vs recurse, recurse often wins/draws as player 2, because karliso uses teccles in a late 20s play move
MSmits: ply
MSmits: i solved this move as a loss
BlaiseEbuth: That's true. And probably THE worse. Pretty proud of that.
darkhorse64: evil mod
ErrorRazor: Othello boss timed out :>
jacek: who made it!?!?!?
darkhorse64: don't know
Astrobytes: :D
Astrobytes: BlaiseEbuth: :japanese_ogre:
Westicles: Is the MILA clobber code published? Be interesting to see how it does against you guys
darkhorse64: won't fix if it does not happen too often
ErrorRazor: First time I've seen it happen
ErrorRazor: Usually I'm the one timing out
Astrobytes: oh it was played as an 8x8 I see Westicles
Astrobytes: darkhorse64: It's rare
darkhorse64: I usually don't have safety margins: I search full time
Astrobytes: Ahh, there I go, back to my safe place at the bottom of Clobber Wood 1. The balance is restored!
Astrobytes: (didn't finish porting to mcts yet)
struct: this is all I can find
struct: https://dke.maastrichtuniversity.nl/m.winands/
struct: and this, is from lines of action bot
struct: https://dke.maastrichtuniversity.nl/m.winands/loa/
struct: Called Mia
jacek: the author of mcts solver paper?
struct: Is it him?
struct: Well I guess he is also the creator of Mila
jacek: he wrote many mcts papers
Astrobytes: Yeah winands
Astrobytes: https://research.cyber.ee/~janwil/publ/clobreport05.pdf
struct: Westicles
struct: https://dke.maastrichtuniversity.nl/m.winands/loa/download.html
UnnamedCodinGamer: this guy is the mcts master - look at his Ph.D. Students supervision section
struct: "GameMaster 2.0 is an abstract board-game playing computer environment. It plays LOA, Breakthrough, Clobber, Domineering, Knightthrough, and Surakarta."
struct: Not sure if the code is public though
ZarthaxX: holy f
Westicles: Looks like it is just an executable
Astrobytes: there are java decompilers
ZarthaxX: mcts-minmax hybrid papers too
AntiSquid: sounds like some generalized ML
Astrobytes: Also, UnnamedCodinGamer, Rémi Coulom is also on this site (CG that is)
jacek: quite old. before that nn hype
Astrobytes: Crazy_Remi = Rémi Coulom
Astrobytes: Uh, he invented MCTS for starters :)
AntiSquid: NLP goes very far back, not necessarily NN indeed
MSmits: he's on the D&B leaderboard. Miklla beat him though (i did by a small margin as well)
Astrobytes: Nice! I hope he gets into some other games
UnnamedCodinGamer: I think I have come across the referred as originator MCTS paper from 2006 I think
UnnamedCodinGamer: I do not remember his name
UnnamedCodinGamer: I might be wrong
Astrobytes: "In 2006, inspired by these predecessors, Rémi Coulom described the application of the Monte Carlo method to game-tree search and coined the name Monte Carlo tree search"
UnnamedCodinGamer: really nice
Astrobytes: Incorporation of the UCB formula was later, but still. The guy's a legend
jacek: remi, winands and Teytaud are common names in mcts papers
Astrobytes: Very strong Go engine too
Astrobytes: Yes jacek
struct: I skip the name part
AntiSquid: ya same
AntiSquid: but maybe worth keeping track who writes the good papers :P
jacek: there are many references
Astrobytes: You would do well to look at names so you can read the references.
AntiSquid: he gets a lot of citations ?
jacek: you read paper and see [name, 2009]
Astrobytes: They all do. Read the names and check the references that are in every paper :|
Astrobytes: *every paper related to the subjects we speak of
jacek: i wonder have many cites that aerodynamic paper has
jacek: dat grammar
MSmits: you wonder about that paper too much
Astrobytes: Poor jacek needs some (o)(o) in his life
therealbeef: an owl?
MSmits: yes, for the sake of the children, lets assume it's an owl :P
Astrobytes: I'm no good at these things
MSmits: :owl:
Astrobytes: Assume owl for safely
Astrobytes: *safety
Astrobytes: Other than that: he needs BREASTS
Astrobytes: Women, etc etc
MSmits: ah i was going to assume chicken next, protein etc.
therealbeef: what are the etc etc?
Scarfield: xD
Astrobytes: He's into anime and catgirls
MSmits: ponies
Astrobytes: Hence the etc
jacek: who isnt?
Astrobytes: Aaaaand the ponies
Astrobytes: I'm not into anime. Or ponies.
Astrobytes: Catgirls...
Scarfield: to each their own
jacek: you have cat avatar
MSmits: and she is a girl
Astrobytes: Both of my cats are
Scarfield: and you say you are not into them :thinking:
Astrobytes: That doesn't mean I'm into human-woman-cat hybrids
AntiSquid: they can wear a costume
Astrobytes: They might be aesthetically pleasing, they might not. Either way they'd be fking terrifying
Scarfield: meh frog girls are where its at
AntiSquid: hybrids are creepy
Astrobytes: Elon can take his cat girls to MArs
Scarfield: he likes car girls?
AntiSquid: elon is playing along with the memes for popularity
Scarfield: cat *
MSmits: probably car also
Scarfield: xD
AntiSquid: someone joked about elon creating cat-girls and then elon reacted that it's an interesting idea, forgot exact story
AntiSquid: he already sent a tesla roadster to space MSmits, but you probably already know about it
MSmits: nope
MSmits: do you mean a toaster?
AntiSquid: oh
AntiSquid: no, a car
MSmits: that doesnt make sense
MSmits: space has no roads
AntiSquid: it does if you look at the story behind it
MSmits: ok
AntiSquid: well it achieved multiple things
AntiSquid: including advertising
AntiSquid: point was to send a heavy load into space and prove his rockets can deliver this task
MSmits: I see
AntiSquid: so he used one of his cars to advertise ...
Nachosauce: Or maybe space is one large road
Scarfield: road theory
AntiSquid: oh and it was orbiting Mars at some point MSmits, comon you should know Mars' satelites
MSmits: phobos and deimos
AntiSquid: and tesla roadster
AntiSquid: for a while
MSmits: i refuse to memorize the tesla toaster
AntiSquid: https://www.youtube.com/watch?v=msn7xbfEHoA
AntiSquid: car is displayed at 8:39
AntiSquid: has an astronaut dummy too
Nachosauce: Are you an octopus?
AntiSquid: one moment Nachosauce
MSmits: I see
AntiSquid: Nachosaucehttps://ultimatemotorcycling.com/2015/03/02/motorcycle-squid-on-the-rise-join-the-anti-squid-movement/ read this @_@
MSmits: dont trigger him about the motorcycle thing
Scarfield: you seem impressively unimpressed smitsi :)
MSmits: it's not very scientific
AntiSquid: until ! an alien species finds it, equips it with AI and fancy gadgets and the car returns to earth looking for its master next millennia
AntiSquid: referencing star trek here, voyager returned to earth on its own :P
Astrobytes: I sent a heavy load into space once, that was a great night, she was... I never did find it in the morning anyway so I assume it went to space ...
MSmits: I'm thinking first contact will be a traffic accident :(
AntiSquid: tunguska style ?
AntiSquid: not saying it was aliens, could be a car equiped with AI from another civilization
MSmits: that's just a meteorite :P
struct: The car is still travelling right?
therealbeef: the squid article mentions 'google+', so quaint
AntiSquid: yes, don't think it hit anything, maybe he should have mounted a satellite on the back of the car?
Astrobytes: Wait til it collides with my load, AB2014
Astrobytes: Anyway, without further innuendo I shall say goodnight
AntiSquid: do you know about starlink btw?
ZarthaxX: cya astro :P
AntiSquid: gn8
MSmits: gn
struct: gn
Scarfield: gn
UnnamedCodinGamer: https://twitter.com/Remi_Coulom/status/1311426326868430849
MSmits: btw, if anyone is interested in the teccles fail at ply 28 for karliso vs recurse. I just reproduced it in IDE and got solved status from my meta mcts
struct: "easy to beat"
UnnamedCodinGamer: is he saying that he used alpha zero?
struct: he is saying it doesnt work
struct: for d&b
MSmits: it really doesn't
MSmits: also hard to use NN, but not impossible ofc. NN can do anything, but might be too constrained in CG
UnnamedCodinGamer: then it is not clear what he used
MSmits: he used some form of nimstring analysis
MSmits: like mikkla and I do
MSmits: it's the most math-y game on CG
MSmits: have to get into combinatorial theory
UnnamedCodinGamer: yes, he mentions it in the reply to the question
MSmits: clobber uses some of this too, but it's less critical there
MSmits: ah ok
MSmits: i cant read the reply, no twitter account
struct: "Dots and boxes is not a game where you build an intuition of who is ahead. You have to calculate exactly. And it is a combinatorial game. It may be possible to make AlphaZero work with the right network architecture, but I have not found it."
struct: You dont need a twitter account
MSmits: ah ok, yes
MSmits: thats it, you do have to calculate exactly
MSmits: any infinitesimal error in a "rough eval" would completely turn the game around
MSmits: it's in permanent zugzwang
MSmits: the only feature you could evaluate is whomevers turn it is basically
MSmits: even the score is useless, because the score has no bearing on what the best move is
MSmits: you can supposedly count chains and get a likelyhood that you will end up ahead, but it's basically a halfassed nimstring analysis
EmbeddedIs77: https://www.codingame.com/clashofcode/clash/1634293c8c92adc0b0de4e5c3845ea517cc6045
struct: MSmits do you think NN would work on STC?
ZarthaxX: shouldnt you ask recurs?
AntiSquid: why would it not struct ?
struct: I think it should
AntiSquid: i am assuming #1 on stc is NN
AntiSquid: #clash EmbeddedIs77
MSmits: it might be a bit complicated. the interaction between player and opponent is weird in stc
MSmits: same as with hypersonic. If you use a NN there, you lose the powerful beam search that ignores opponent
MSmits: using a NN in boardgames is just different
MSmits: might be that it's useful to use a NN partially, in a bot. Do some parts with pure math or brute force and do other parts with NN
Kyuujin: hello to night coders
Nick_Svaki: the problem description is hard to understand.
ZarthaxX: which one Nick_Svaki?
Nick_Svaki: I don't know how to seek the problem that I can't solve.
ZarthaxX: what problem is it