jrke: is there any thing like input().split() in c++
Marchete: hmm has thor codegolf changed?
Marchete: nvm read the forum
jrke: code golf you mean power of thor?
jrke: ah! yes i got up and saw an increase in my rank of power of thor automatically i was shocked and later i discovered the change ;)
jrke: but whats the change?
Marchete: different validators
whay: what do you man
whay: what is your name
jacek: good morning
WINWINWIN: pinarica, I know that it is a late response, but I think the reason for the sudoku error is that you are checking sum of row and column to 45
WINWINWIN: But there may be a validator that sums to 45 but does not satisfy the 123456789 requirement.
DomiKo: good morning
whay: you can tellr me
HDit: hôm nay là thứ 7
HuynhNgocHai: Hôm nay là thứ 7
whay: how can help tyou
Astrobytes: HDit, HuynhNgocHai: please speak English in World chat, you can create your own channel if you like
Edjy: Dear codingamers, I NEED YOU ! I need help to finish my CG logo here : https://pixelcanvas.io/@11161,15 (it is not hard and doesn't take too much time) you just have to place one or two pixels
Astrobytes: Edjy: Please do not spam the chat with off-site links
jacek: thats edgy
DiL: Hi everyone! I think I found a "bugged" clash: my solution doesn't pass all tests (one is failing for timeout) but got 100% score. What's the best way to fix it?
Uljahn: forget about it and go on
Uljahn: timeout could be caused by debug mode of your language or by different CPU
aCat: TonAtHere I suppose there is no a lua course available I could recommend right away
aCat: Depends do you know how to program or not
aCat: I would simply advise doing some stuff and forcing yourself to do it in lua
aCat: Like codingame puzzle - and search lua docs and web for solutions 'how to write some construction'
aCat: lua is quite specific language, if you want to know something specific - or need to know it for writing plugin for some game you can as me some dicrect questions
itzblinkzy: share your code please
jrke: whats the fastest record for testcase 1 in search race ?
abraaobarros: Hi, guys. I posted my code on github. link: https://github.com/abraaobarros/code-rings-codingames/blob/master/src/index.ts :thumbsup: .
abraaobarros: I will try to improve with loops. feedbacks and pr's are welcome
Zenoscave: I'm not sure code sharing in a competitive setting like this is encouraged.
Zenoscave: I absolutely agree. Except that this is an ongoing competition. Sharing small ideas and snippets is fine but an entire solution is generally frowned upon
Zenoscave: as a heads up, if it were an actively ranking current competition that would be a definite disqualification. It has happened in the past
SPDene: I don't know where you get that idea from. A huge benefit of this site s being able to learn from looking at code that someone shared.
Zenoscave: Perhaps the owner's stance on the difference between puzzles and ongoing challenges, and my years spent here where this topic is repeated many times is where I get the idea from
SPDene: ok, next time I share my code, I'll bear in mind that the owner may disqualify me from playing CoC ever again /s
Zenoscave: I don't mean to discourage sharing or learning. Just the modicum in which it was presented. Sharing ideas is a wonderful thing. SPDene this wasn't about CoC...
Zenoscave: I was referring to the Code of the rings code posted earlier
Zenoscave: But do realize that even in CoC people scrape solutions to use as their own with a bot. It is encouraged to share but also be aware of its ramifications
Zenoscave: The forums have many resources discussing it. but, yeah, a T&C would be nice
Zenoscave: F.A.Q. ^^^
DiL: why would someone use a bot to play CoC?
WINWINWIN: To win continuously :)
DiL: and... so? :-) is it funny?
itzblinkzy: how boring
blueman34: I suggest reading lua documentation.
ANONYMOUS42: i suck at ruby
blueman34: I suck at everything.
MSmits: except for the sucking part. He can only get worse at sucking from this point on
MSmits: I think you misunderstand me
MSmits: He sucks at everything. He is the absolute top at it. It cannot get better
MSmits: it can only get worse
MSmits: so he will get better at everything except sucking
itzblinkzy: im okay at coding but im so baad at math its annoys me
blueman34: im the opposite
blueman34: bad coder
blueman34: good mather
DomiKo: you can use https://projecteuler.net/ to improve math skill
ANONYMOUS42: team up and you can be good at both
LoGos: or bad at both
LoGos: I mean if you combined them
blueman34: im so close to sloving this clash of code
blueman34: but i hate lua
blueman34: such a restricted lang
blueman34: i trying to figure out of to round up a decimal
whay: telle me
ANONYMOUS42: are you trying some fancy interval maths or just using an array lol
ANONYMOUS42: don't think it would be better in term of code length like that though
whay: Damn you answer me
blueman34: its a shortest mode
whay: answer me
blueman34: i got 50%
blueman34: even though i managed to get 3/4 correct :/
whay: i got 25
whay: fuk you
blueman34: should've been 75%
DomiKo: that's right
blueman34: i mean 3/4 is 75% so
jacek: so mathy
DomiKo: oh wow just found minus error in MCTS
ANONYMOUS42: what's MCTS
DomiKo: Monte Carlt Tree Search
ANONYMOUS42: ah ok
DomiKo: Monte Carlo Tree Search*
DomiKo: sign error are the funniest
DomiKo: ohhh yes top50 in UTTT
DomiKo: one wrong sign = one day of debuging
DomiKo: oh yes
jacek: what did you say? i cant see that low
MSmits: grats DomiKo
DomiKo: now time for Othello
MSmits: been fitting parameters myself these last few days :P
MSmits: no new submit yet though
MSmits: getting some rare wins vs dbd r now
MSmits: well the emphasis is on rare
MSmits: but before I was winning 0 out of 100 so... progress :)
itzblinkzy: damn it share your code please i've only been coding for a week and some of these questions are so hard :pensive:
MSmits: are you talking about clash?
itzblinkzy: only way I can learn are from other people who solved it
itzblinkzy: yes clash
MSmits: but clashes are easier than easy puzzles
MSmits: i've never done a clash myself
Astrobytes: I fixed some Othello bugs, I'm coming for you all. Very slowly.
MSmits: but that's what I've been told
MSmits: gj Astrobytes :)
DomiKo: clashes sometimes are tricky
DomiKo: or the text is really not clear
MSmits: then they are just bad
MSmits: if not clear
MSmits: should be edited
jacek: well reverse mode are not clear :<
itzblinkzy: well thats the point
Astrobytes: Forgot to change my aspiration window size, was pruning almost everything lol
MSmits: the problem statement should never be the issue
MSmits: whats an aspiration window size
Astrobytes: the size of my aspiration window
MSmits: ohhh right of course
MSmits: thanks for that clarification
jacek: but in aspiration window you re-search when out of bounds eh
MSmits: is this like PVS?
MSmits: or null window
Astrobytes: null window is related yeah
MSmits: I see
MSmits: did you fix your TT?
MSmits: a good way to check if it works is if you solve sooner
MSmits: with yavalath i solve sooner with TT on
Astrobytes: I removed it for the time being until I fix everything else
Astrobytes: Got some time off from the desk job this week so hopefully have some time to work on things
MSmits: Othello has some serious RPS
MSmits: like Oware, very similar
eulerscheZahl: "Forgot to change my aspiration window size, was pruning almost everything lol" have you tried chokudai search? :P
MSmits: I have versions with 60% winrate vs Robo
MSmits: and this:
MSmits: 19:43:39,434 INFO Temporary results for othello.cpp (ETA : 11 minutes) : http://chat.codingame.com/pastebin/407f27f3-5139-4503-bf2e-42816f6857ce
MSmits: so I get to pick who I do good against... once again :P
eulerscheZahl: is that win a rust timeout?
Astrobytes: chokudai... that's that beam searching DFS thingy right?
MSmits: it's actual wins
MSmits: 5-10% vs dbdr
eulerscheZahl: beam search with iterative widening instead of deepening
MSmits: dbdr uploaded a compiled version I believe
dbdr: :scream: :D
Astrobytes: ah yeah
eulerscheZahl: now you pinged him, see what you've done
Astrobytes: nice work MSmits
DomiKo: really nice
MSmits: lol thats ok
MSmits: he doesnt mind
dbdr: I don't mind pings
eulerscheZahl: then invite him to CoC
MSmits: jacek also has some wins vs dbdr
dbdr: using a book MSmits?
MSmits: just fitting params
MSmits: this is not a submitted version
MSmits: I have less than 10 params I am fitting. All the usual stuff you find about Othello
MSmits: corners, edges, stable cells, mobility, frontier etc. Everything has a param
Astrobytes: Got your frontiers working now then
MSmits: No idea
MSmits: I havent fitted it yet
Astrobytes: ah right
MSmits: it's just in there with a param of 100
jacek: use TDL luke
MSmits: maybe it hurts me
MSmits: jacek I will if I get some good example code
MSmits: I have no idea where to start on TDL
jacek: star from the beginning
MSmits: show me the beginning :)
MSmits: how did you learn?
Astrobytes: He did the maffs innit
jacek: started with tic tac toe
MSmits: the maffs?
MSmits: jacek that sounds like you were doing a NN?
jacek: once again i dont like fancy math
MSmits: TDL is also just about fitting a few params, it doesnt have to be a NN does it?
Astrobytes: All those squiggles...
jacek: i have NN only for oware
jacek: yes, tdl can fit other params
MSmits: ok but what params did you fit for TTT?
jacek: aside from N tuples i have frontiers and mobility, and it alaways show frontiers are bad
MSmits: most times i hear about people doing learning for TTT it's a NN or Q learning or something
jacek: i used 3x1 N tuples for tic tac toe
MSmits: ohh ok ntuples
jacek: but for real start, i used direct table
jacek: unsurpisingly, it learned to play perfectly then
MSmits: direct table means every boardstate has a param?
jacek: yes. generaly hashmap
jacek: board state -> win/lose
MSmits: oh ok, well that works too
MSmits: action vs state
jacek: 19683 possible states, while not much, this is not the point to learn every possible state and its value
MSmits: tell me about it, the number 19683 appears A LOT in my uttt bot :P
jacek: with 8 3x1 Ntuples, x2 (for each player), so 8 * 54 = 432 weights it still can play perfectly
MSmits: there are ways to play perfectly with far fewer params I am sure
MSmits: but its a good way to test
jacek: at the very least, you can see if the code is good
MSmits: did you start from someone elses pseudo/example code? Or did you start from explanation/math ?
jacek: from explanation
MSmits: well good job, thats not easy
MSmits: hope I have the patience to figure it out that way as well
jacek: well now i have framework for training, which i used to train oware, othello and onitama
jacek: you can see the results
MSmits: I see it yeah
MSmits: funny thing btw, for a day or two I was convinced you had an opening book because your bot seemed deterministic on oware, for like 20 plies
MSmits: and it also seemed to always make the best move according to my meta mcts
MSmits: then it suddenly changed its 2nd ply move
MSmits: it rarely plays 11 instead of 7
MSmits: as you would say
MSmits: so it's not a book
MSmits: it's just really good
Astrobytes: It is rather excellent
MSmits: I can counter book it rather easily in the sense that it's super deterministic, but I have to find a mistake first and i havent
MSmits: Also, when I put my book in, my bot played worse =/
jacek: it may be a little nondeterministic, i randomly add +/- 0.02 to UCB
MSmits: the reason is, I think, that my book currently finds the 1 reply on 2nd ply better than the 5 and all my params are fitted for 5
MSmits: jacek it really isnt, your bot is super deterministic
MSmits: there are only 5 moves to choose from and usually only 1 or 2 are even worth considering
MSmits: so it's not surprising most good bots will be deterministic
jacek: and yet you pulled me below robo :(
MSmits: Robo's bot is also very very good, I have the same problem when trying to counterbook him
Astrobytes: So narrow
MSmits: I can beat recurse and agade without book currently, or at least around 50%. But nothing works well against robo and you jacek
MSmits: my current version has no book btw. I just tried it once
Astrobytes: Is it helpful?
MSmits: but i got losses against lower players
MSmits: no because its not compatible with my overfitted paramns
MSmits: i will choose a different opening
Astrobytes: Ah ok
MSmits: if the book gets deeper and i refit params, it might work
MSmits: there's a chance my bot just performs poorly in the 7-20 seed region. In that case no book will help me
Astrobytes: Maybe a case of finding the right weight for the book moves vs the rest of your params
Astrobytes: whay, do you have anything useful to say other than 'hi'?
MSmits: Astrobytes there is no weight for my book moves. I simply play them
Astrobytes: That's not always optimal MSmits
MSmits: why not?
Astrobytes: In your case, your eval is sometimes better than book
MSmits: if it were, why would I not have noticed this when playing offline?
MSmits: i use the same eval
whay: i hate all
MSmits: i just spend more calc time and go deeper
MSmits: so I dont see why the 50 ms spent on CG is ever going to be more trustworthy
MSmits: Maybe it's an implementation thingy
MSmits: btw not saying you're wrong per sé Astrobytes. From what I read about opening books, the eval is sometimes preferred over the book result. I think it depends on what you decide to put in your book and what you don't. I set a cutoff at 10k games. If a node has less than 10k games played, it's best move is not put in the book
MSmits: so thats maybe 3 hrs of calculation time vs 50 ms
Astrobytes: Yeah, I guess you kinda don't need to weight your book moves in your case
jacek: i think it also depends how your bot react to the position left by book
jacek: book moves can be really good, but bot too dumb to make use of it
MSmits: yes jacek, that's my problem currently. My bot gets put into situations the parameters are not fitted for
Astrobytes: Yes, that's also true
MSmits: it's not that it's dumb, it's that you can't really deal with all possible board layouts wit just 10 params
MSmits: so it specializes
jacek: so use 16k params like me *.*
MSmits: I use books for that :P
MSmits: just havent been succesful in oware (yet) :)
MSmits: i think it's really cool to see the program tell me all those NN's do indeed play the best moves though
MSmits: i doubt they make more than 5-10 mistakes over a 100 ply game
MSmits: when they solved awari it was also found that usually 5-10 mistakes were made by the best competitive bots
MSmits: and this was a while ago, I think our NN's might be up to that level
jacek: i tested and my alt was 47 with 1-ply search
MSmits: rank 47 in oware?
MSmits: hmm seems good but hard to say
Laminator: how do you guys go about training your models on these games? Is there like an api to call?
MSmits: no support at all
MSmits: completely trained offline
MSmits: in your own framework, you can use tensorflow, keras, pytorch or whatnot
MSmits: then you put in something that fits into the codesize
MSmits: which is 100k characters
MSmits: this includes all the code
MSmits: it's quite an accomplishment to make a NN work on CG
Laminator: So you have to get the source code for the game or recreate it?
MSmits: well most NN's here are written for simple board games, so the source code would be a few 100 lines
MSmits: but you can use a java referee if you want
MSmits: most games have them
Astrobytes: Source code is usually given or simple enough mechanics that you can implement it easily yourself
MSmits: I always write it myself for offline programs
MSmits: usually the java ref is slow to setup
Astrobytes: It's mostly faster if you implement yourself, but that's not totally necessary (as per karlis o)
Astrobytes: And yeah MSmits
MSmits: i remember for locam i was fitting params using brutaltester referees
MSmits: the nr 1 used his own version and did like 10k games per hour
MSmits: because locam did not need much calc time and the limiting factor was the referee
Astrobytes: who was #1? closet AI?
eulerscheZahl: 10k per hour? i remember that differently (as in: much more)
MSmits: he did a total of 100k games or so
MSmits: I guesstimated it to 10k/hour, i dunno
MSmits: i remember the 100k number from his PM
Laminator: Very neat. I just finished an RL class last semester so it'll be nice to try to implement.
MSmits: maybe it was 100k a day?
MSmits: Laminator try it on a simple game. Oware we know is good for it.
DomiKo: is there any game where QL will be good?
MSmits: QL is not useful for games that arent solved is it?
eulerscheZahl: "When testing a change running 100K games was a standard way to decide whether it is an improvement."
eulerscheZahl: so 100k multiple times
MSmits: yeah, but it was 30 day contest, so we'll never know :P
Laminator: As long as the states variables are discrete, I think QL could work.
eulerscheZahl: let's quote recurse instead: "I am talking about self-playing at a rate of around 1000 games per second here."
MSmits: ok... i stand corrected :)
eulerscheZahl: closet did the whole draft order with selfplay. that needs a lot of batches with 100k each
MSmits: yeah, I know, I tested with 1-2k batches and it was useless
eulerscheZahl: did the same
eulerscheZahl: 500 games: yay it's better \o/ 1000: wait a second...
MSmits: when fitting params for board games I usually do 400 games vs 5 different players and it works fine. At least for rough estimates
MSmits: thats 400 total
MSmits: 80 per
Astrobytes: I just fumbled around in the dark for a week lol
eulerscheZahl: i hated the game and downloaded matches to steal the draft from others
eulerscheZahl: also smits was a great help as he gave me some order
MSmits: yeah you did that
Astrobytes: I remember :D
MSmits: i spent 5 hrs stealing it by hand
eulerscheZahl: instantly 75% wins vs my old bot
MSmits: i just watched a ton of games
Astrobytes: Was kinda sad it was reduced to that
MSmits: to see what he preferred
eulerscheZahl: and i told you that you can manually set the cards in draft
MSmits: i went to rank 1 immediately when i did that, i guess the rest of my bot was still better than most at that time... i stayed like a few hrs at 1 :P
eulerscheZahl: even 3 times the same and recurse printing his fake stats to give different score to them :D
MSmits: the sad thing is that games like this have so much unknown info that it's hard to do any smart algorithms
eulerscheZahl: https://www.adrienvannson.fr/cg-ranking/ you plateaued a while
MSmits: everyone just bruteforced it and the draft was so unbalanced that the difference between objective strengths of the cards was greater than the need for balance
DomiKo: MSmits how you play those 80 games per player?
MSmits: automated IDE games basically
eulerscheZahl: or play by hand
MSmits: oh right, i see i was at 1 for a while, nice find eulerscheZahl
MSmits: I think I was when legend opened
MSmits: interesting, I guess my end result at rank 5 wasnt as lucky as I thought it was, i actually went down over time
MSmits: massive submissions distorting the last few days prolly
eulerscheZahl: i went pretty downhill too in the last hours :D
MSmits: yeah you were unlucky, the graph shows a small dive at the end
whay: level up
eulerscheZahl: and you gave me the draft on august 14th, in the evening hours
MSmits: I gave it to you?
MSmits: oh right
MSmits: after my 5 hrs of madness
eulerscheZahl: but you told me not to tell on chat that you did :flushed:
MSmits: sounds like me
MSmits: I guess you helped me a lot too and i felt obligated to share
eulerscheZahl: i guess it doesn't matter anymore, 2 years ago
eulerscheZahl: now the info is leaked
whay: You are racists, do not speak to me
Astrobytes: whay, really. Please go and waste your time elsewhere.
eulerscheZahl: shush, don't talk to him!
whay: talking to me
MSmits: uh oh, you fed the troll
Scarfield: feed the ducks not the trolls, btw has there been made any new mods for chat?
Astrobytes: It's not been done yet Modfield
jacek: arent you mods already?
eulerscheZahl: on the way, delayed
eulerscheZahl: because of homepage redesign
Astrobytes: *'Quest Map'
MSmits: noone here is a mod are they?
MSmits: dont you get a disk or something?
eulerscheZahl: but 2 are about to become one
whay: What did you say about me?
eulerscheZahl: 3 if you dig a little more in the chat history
eulerscheZahl: about 7 in total IIRC
MSmits: how do they decide who to make mods?
MSmits: I mean I am assuming you'll be mod
Astrobytes: 8 I think
eulerscheZahl: thibaud asked here on chat to suggest names
Astrobytes: Yeah, I nominated euler
MSmits: ahh ok
MSmits: good choice
eulerscheZahl: noble Astrobytes
Astrobytes: Thibaud asked me directly, also WIN WIN WIN
whay: im arabic So do not talk to me
Astrobytes: And we suggested some others
MSmits: ahh ok
jacek: i declined
eulerscheZahl: also: robo, dbd r, struct, Neuman n
MSmits: good, then we can more quickly get rid of spammers
eulerscheZahl: and uljah n
Uljahn: im in doubt
whay: Damn you do not talk to me
MSmits: why are you doubtful Uljahn?
Astrobytes: Why so Uljahn?
eulerscheZahl: you need a dictator for #ru
Scarfield: with great power, comes great responsibility
eulerscheZahl: i'm moderator on discord for a year maybe. Deleted less than 5 messages in that time, no kicks or bans
whay: i love nazs
MSmits: your presence is enough
eulerscheZahl: either there is much less trolling or i'm online at the wrong time. or both
Astrobytes: Discord's pretty tame in comparison to CG chat spam-wise
Scarfield: or online at the right time :thinking:
MSmits: CG chat is easy to get into
whay: telle me
Astrobytes: yeah, discord requires more than a click I guess
jacek: and elements of harmony
MSmits: I sense an obscure reference
MSmits: well Idon't know it, obscure to me
jacek: try image google discord
eulerscheZahl: afk cat time
Scarfield: just googled it, its my little pony
Astrobytes: discord... harmony...
Astrobytes: stupid MLP
jacek: voiced by Q
Astrobytes: I was thinking musically
leojean890: eulerscheZahl your bender episode 4 is not easy ;P
MSmits: he has made harder puzzles
leojean890: I believe you :)
Astrobytes: Yeah, try Space Maze. And get your ray-casting A* NN ready
Astrobytes: (last part is a joke btw)
leojean890: "very hard" category I suppose:)
MSmits: leojean890 do you do a lot of puzzles?
leojean890: I began "training" part with puzzles only 2 or 3 weeks ago :P
leojean890: before, I did only multis and optims;)
MSmits: ahh ok
leojean890: but I did all the easy and medium ones
leojean890: only the official CG ones
MSmits: I enjoy them, but when I do a puzzle I always think I am going to be lagging behind on my multis :)
leojean890: for the momeht
leojean890: lagging behind ? you mean it prevents you from improving your multis ?:P
leojean890: because you do something else
MSmits: new multis come out all the time and i cant keep up
leojean890: so it slows your multis improvement down :P
MSmits: thats it
Astrobytes: Haven't done any puzzles for a while, I tend to go on a little spree of doing them then leave it for a while
leojean890: I saw that you have many great rankings indeed :P
leojean890: it's hard compet in the top of each ;P
MSmits: thanks, yeah it is
leojean890: I never reached it for the moment :P
MSmits: i am not particularly fast though
leojean890: (top of multis !)
MSmits: I tend to stay obsessed long enough to get top 3-5
leojean890: I am around 28 in vindinium, bandas, oware..
Astrobytes: You're not #1 in any MSmits? I'm sure you were before
MSmits: yavalath currently
leojean890: but next spots are harder ;P
Astrobytes: Ah right
MSmits: rest, close
MSmits: uttt close
MSmits: onitama close
MSmits: was nr 1 there both of them
MSmits: oware too, for a little bit, but the NN's took over again :)
Astrobytes: lol yeah, indeed
leojean890: I have to learn how to do them ;P
MSmits: me too, mostly for work though. I don't think I'll enjoy it that much beyond academic interest. I don't see myself obsessively try to get nr 1 with a NN
Astrobytes: And you can do pretty damn good without them
Astrobytes: (on CG I mean)
MSmits: yeah :)
MSmits: hi _Akira_
leojean890: spots < #20 are hard to reach, right ;p
MSmits: depends a lot on the multi leojean890
leojean890: for little ones I meant ;P
MSmits: sometimes only the top 10 is strong and the rest is much weaker
leojean890: for populated ones
MSmits: twixt-PP top 10 should not be too hard
leojean890: it's hard to reach #100 :P
MSmits: or paper soccer
_Akira_: have you played the power of thor
MSmits: leojean890 hypersonic top 100 is really hard
leojean890: beam search, right ?
leojean890: I have to throw my python away
leojean890: for this one:)
MSmits: yeah you need C#/java or faster there
leojean890: I already did it for othello and oware ;)
Astrobytes: top 5 for Othello
leojean890: but still #22, #30
leojean890: (minmax both)
MSmits: I think an inexperienced board game player would find it hard to get past you Astrobytes
whay: I'm not joking
MSmits: so top 5 is not the only hard part
Astrobytes: Fair, it does have mini leagues as you can see by the score diffs
whay: I'm not joking خن
MSmits: it's just that these board games have a high concentration of good players in a small league
Astrobytes: It gives a good sense of accomplishment that way I feel
MSmits: I dont need to see the rank to know I did well
MSmits: If i see myself near tric trac, I know i did ok
leojean890: I heard that the first use minmax in othello
whay: I'm not joking
Astrobytes: With multi probcut though leojean890
MSmits: leojean890 most use minmax yes, but with a lot of enhancements
MSmits: multi probcut is a really exotic one
leojean890: oh, using cutting ?
Astrobytes: (just misread that as erotic, wtf!)
leojean890: I should read about multi probcut
MSmits: its a pruning strategy thats almost like a neural network to decide what to prune
leojean890: oh ok
MSmits: Astrobytes has a dirty mind
leojean890: I see, didn't know this one
Astrobytes: It's like a multiple linear regression to tune your alpha and beta (to put it really simply)
MSmits: I use a mcts variant with eval instead of random rollout
Astrobytes: Check it out, lots of info out there, chessprogrammingwiki, some papers
leojean890: thanks for the tips, I will read about that !
jacek: or use jacekmax, the most awesomest algo there is
Astrobytes: hehe ofc
DomiKo: there is that moment
DomiKo: in life
MSmits: things can only get better when you add max to them
DomiKo: when you want to read about sommeting
DomiKo: and first google search
DomiKo: is jacek comment on github
MSmits: which one?
MSmits: link us
DomiKo: ok not the first
Astrobytes: must be benchmark
DomiKo: i was searching for mcts solver
DomiKo: and found jacek github ;)
jacek: oh my
MSmits: Robo uses mcts solver in his nn's
jacek: solver should be part of 'vanilla' mcts
leojean890: he has a good perf in othello as well !
MSmits: agreed, but it doesnt always help very much
MSmits: leojean890 Robo always does well when he tries
leojean890: yeah, like for pacman :)
MSmits: he won a contest before
leojean890: oh nice
leojean890: ow this one
leojean890: not my best multi :)
leojean890: hard to simulate I think
MSmits: it was my first contest
leojean890: maybe will do that one day
Astrobytes: I still hate that game lol
MSmits: yeah not that easy
leojean890: I agree
leojean890: nor botg ;)
MSmits: never tried that
MSmits: but I hear bad things :P
Astrobytes: I didn't do the contest but after all those Wood leagues I got bored
leojean890: very hard to simulate ;)
MSmits: did you try kutulu?
leojean890: I reached bronze in botg with easy code (5 lines ?)
leojean890: but after it's harder
eulerscheZahl: you will even struggle to put the BotG engine into 100k chars :D
leojean890: kutulu yes, I'm ranked #120
MSmits: ahh ok, i found it hard to come up with a search for that
leojean890: MSmits, I read your PM btw
MSmits: oh right
MSmits: i remember drawing lots of circles for that
leojean890: smitsimax :)
MSmits: that was the first iteration
Astrobytes: Smitsimax: First Blood
eulerscheZahl: took a while to get popular among other players
leojean890: to code botg engine with 100k chars, indeed..:P
MSmits: because of CSB
MSmits: it became popular for a while
leojean890: some people tried to use it for pacman, I read
MSmits: even though almost noone pulled it off there
leojean890: but had difficulties
MSmits: for some reason it was most effectively used in xmas rush though
MSmits: probably because there were very little good alternatives
Astrobytes: Didn't seem sensible for Pacman imo but hey
Astrobytes: That was surprising for Xmas Rush
MSmits: sometimes having "a search" is better than no search. These contests are often hard to write searches for
leojean890: OOC for example ;)
MSmits: yeah, i had a beamsearch for ooc
leojean890: a good tracker is a good start :p
MSmits: to path through areas with few mines
MSmits: but it's only a partial search really
leojean890: my mine avoidance strategy was weak, that's why I stayed top gold for 10 days not succeeding to reach legend ;P
MSmits: thats what you have to do sometimes, do one part with heuristics and one part with search
MSmits: ahhh, that sucks, to be stuck behind gold boss
Astrobytes: search is search, doesn't always need to be a full game tree search, you might use different types of searches for different stages/situations
MSmits: I always managed to get past it, but often at the last moment
leojean890: for 10 days it's sucks yeah^^
MSmits: yeah Astrobytes
leojean890: I was like between #15 and #2 gold all the time ;P
MSmits: was there much rating difference with goild boss?
MSmits: sometimes the rank doesnt mean all that much
MSmits: if gold boss is at +3 for example
leojean890: I sometimes had more than its "min"
MSmits: ahh ok
leojean890: he navigates between 28 and 32
leojean890: and me between 24.5 and 28.5 :P
leojean890: need big luck xd
MSmits: I see, then you might have gotten lucky with a RPS spamsubmit thats weak against you and strong against boss
MSmits: thats how people get pushed into a league
leojean890: and with CG bench I had 30-40% winrate against him;)
MSmits: wow mobility param has a really strong effect on the game in othello
MSmits: i lowered it a bit and my bot is crap
MSmits: was at 800, started at 500-600 for refitting and both tests 20% winrate overall
leojean890: yeah sometimes someone spam submit and some people pass because of him:)
MSmits: thats how the servers overload sometimes
MSmits: when someone spam promotes a bunch of players in CSB
leojean890: leagues with 30k people :)
leojean890: it's a huge mess:p
MSmits: it doesnt get much attention on chat these days
MSmits: but before, everyone would always talk about csb
leojean890: the most popular multi ;P
leojean890: -3vel became a legend ;)
MSmits: well mostly because it's connected to the tutorial i think
MSmits: it gets suggested at first
Astrobytes: That and a certain absent advocate
leojean890: often, people talk about it in #fr ;)
MSmits: you mean MK :)
leojean890: I heard about him
leojean890: I was not really present in CG at his time ;P
MSmits: csb is still alive in #fr?
leojean890: sometimes yeah but depends on who pass :)
MSmits: ah ok
Astrobytes: Yes, pretty much the same here
leojean890: often new codingamers
eulerscheZahl: fix ur csb
leojean890: but it's maybe 1 day out of 2 ?
leojean890: idk the exact frequency
Astrobytes: ezpz euler
eulerscheZahl: where did MK go? that was his job to say that
MSmits: yeah I dunno
MSmits: in my first week on CG he kept asking to teamviewer me :P
icecream17: errr... what is csb?
Scarfield: el structo said he was gaming last week, so he is still around, just not advocating CSB anymore
MSmits: pod racing multi icecream17
icecream17: oh that
eulerscheZahl: once i changed my nick as automaton kept pinging me
Astrobytes: He quit CG
eulerscheZahl: MK didn't realize it was me and PMed me to give advice to CSB
Automaton2000: the difference between the two
Astrobytes: literal last words "MadKnight quits"
Automaton2000: a simulation of the game
Astrobytes: And not been seen since then
leojean890: by changing only one character
leojean890: you gets spammed as well
MSmits: Astrobytes any specific reason ?
Astrobytes: dbdr showed the chat log last week
MSmits: I know he hadnt been coding for a long time
Scarfield: untill he unquits, or some other user starts getting people to play CSB
Astrobytes: Looked like he got (wait for it): offended
eulerscheZahl: last contest he tried was mean max i think not counting a placement around rank 1000
Astrobytes: Hard to say, was only mild trolling about CSB from what I could see
Scarfield: i saw the link, but didnt seem to me that it was anything bad (?)
Astrobytes: Or about him being a bot
Astrobytes: Yeah it was very mild trollage
Astrobytes: I think he was fed up with the CSB association
leojean890: so he PMed you to "teach" automaton bots to say "fix ur csb"
eulerscheZahl: he's so easy to parodize spellcheck underlines that word, no idea how to write it
MSmits: no leojean890 AutomatonNN at some point decided to talk to euler all day
AutomatonNN: why is that a thing about checkers?
MSmits: so euler got pinged
MSmits: so euler changed his name to prevent being pinged
Astrobytes: Just parody euler
leojean890: because named automaton ?
eulerscheZahl: and royale had to change his nick permanently because of a contest :D
leojean890: code royale
MSmits: for the best Automaton statement, check eulers profile leojean890
eulerscheZahl: yes :D
leojean890: "eulerscheZahl is a noob" - AutomatonNN
MSmits: i was there when he said it
AutomatonNN: eulerscheZahl is there a way to see a tutorial on what the problem is that the other ones are always
MSmits: could not stop laughing :P
leojean890: nice hehe
Scarfield: wish i was there xD
MSmits: was in the morning when i was on the train if i recall correctly
MSmits: best start of the day
eulerscheZahl: luckily automaton got fixed to some degree and pings me less often now
eulerscheZahl: as seen above
MSmits: oh right, he just did it again
MSmits: britney spears references, really ?
MSmits: i thought we drew the line at my little pony
Scarfield: well you got it
Scarfield: and she is beyond that line :thinking:
MSmits: i dunno
Astrobytes: enough of this Toxic chat
Astrobytes: Moar Britney
Scarfield: its bitney "beeach"
MSmits: imagine MK missing all this
Astrobytes: We don't have to imagine, he is.
Astrobytes: (I know what you mean)
eulerscheZahl: he managed to turn himself into a meme in the end
Astrobytes: An interesting life cycle for sure
Scarfield: 20% helpful 80% "fix it/ ezpz"
Astrobytes: *70% "fix it", 10% rubberducking/help me with my assignment
eulerscheZahl: a long time ago, even before Automaton, i suggested to write a MK bot. seems we should do it now, if the real one is gone
MSmits: yes, do it
eulerscheZahl: Astrobytes these numbers don't add up
MSmits: did you use "+" ?
eulerscheZahl: yes, in base 10
**MSmits is stumped
Astrobytes: 70% fix it, 20% helpful, 10% rubberducking?
terragady: is it possible to have higher difficulty on clash of code?
eulerscheZahl: funny part: it's always base 10, no matter which number system you are using
terragady: the tasks are super easy
Scarfield: 10% power of will
tomatoes: use bash
MSmits: terragady if you want harder tasks, you will need to do something other than clashes
Astrobytes: Do puzzles instead terragady
Astrobytes: Or bot programming, or optimisation
terragady: but i want to compete with friend :P
MSmits: you can compete through bot programming
eulerscheZahl: you can fight them in bot...
MSmits: I type with 2 fingers euler
Astrobytes: And that lasts longer too
eulerscheZahl: i'm sleepy
Astrobytes: You too MSmits?!
MSmits: ah yes, its your bedtime
MSmits: i never did a typing course
MSmits: so yeah
eulerscheZahl: what happened to your other 8 fingers?
MSmits: thats how i learned
Astrobytes: Even darkhorse has upgraded to using thumbs
MSmits: it's just badly learned
Scarfield: wait, isnt it 41 minutes past your bedtime?
eulerscheZahl: i didn't take a typing course either. invented my own system
MSmits: there is a big issue with 2 finger typing
Astrobytes: I learned typing to the soundtrack of... Boney M
eulerscheZahl: i'm staying up longer in summer. my room is under the roof, too warm to sleep
MSmits: your hand makes larger movements, so you hit the keyboard a LOT harder
Astrobytes: Thank you Mrs Hendry :(
MSmits: i wake up my wife and daughter typing sometimes :p
eulerscheZahl: can you type blindly?
Astrobytes: Me or Smits?
MSmits: nearly, i correct my hand placement subconsciously every minute or so
MSmits: if i use a different keyboard, i have problems for a while
Scarfield: you wake them up, do you mean you type with your fists, not fingers?
MSmits: or am sitting in a different position
eulerscheZahl: only 1 way to find out: https://play.typeracer.com?rt=sc2v8x79u
MSmits: Scarfield it sounds like fistst
Astrobytes: I usually type without looking, depending on pain meds status lol
MSmits: ok one time euler
Astrobytes: And I have a loud keyboard
Astrobytes: Satisfyingly loud
MSmits: mmh 55 to 59
eulerscheZahl: we are both slow :D
MSmits: i got stuck on the word forever
MSmits: had to retype it twice
eulerscheZahl: yeah, you had a lead
MSmits: thats what happens when i misplace my hand on the keyboard
MSmits: some words i type super fast and others i get stuck
Astrobytes: Hm, that's a late-morning exercise for me I think lol
Scarfield: how do you type "stewardesse" MSmits?
eulerscheZahl: yet i constantly beat you on the chat Astrobytes
MSmits: Scarfield, poorly
MSmits: with 1 finger
Scarfield: the longest word only typed with the left hand
Scarfield: nice :)
eulerscheZahl: and i have to put more thoughts in my wording, would be easier for me in German
Astrobytes: Don't mock the afflicted eulerscheZahl :P
MSmits: do you think it's possible for me to be better with 10 fingers than with 2 if i typed this way for 20 years?
eulerscheZahl: missisippi is fun to type as well
Astrobytes: Maybe, but if you can type fast enough with 2 then crack on
MSmits: yeah for now
eulerscheZahl: obviously you need to switch to Dvorak first
MSmits: maybe if i get older i could get some repetitive strain injuries
Astrobytes: Works fine for darkhorse using two fingers (and recently thumbs as stated)
MSmits: oh he upgraded recently?
Astrobytes: Yes. Stopped using his ankles too.
MSmits: actually i do use a thumb for space bar
eulerscheZahl: my spacebar is worn up on one side, almost new on the other
eulerscheZahl: always using the left thumb for space
MSmits: Astrobytes what did he do with his ankles?
Scarfield: type loudly
Astrobytes: lol, it was a mistranslation, he'd hurt his wrists typing but stated ankles instead
MSmits: ahh ok
Astrobytes: It was too good to ever forget
Scarfield: lol my spacebar is actually as well, but on the right side. would have thought the left took most of the beating from gaming
MSmits: so far I only get issues with too much mouse clicking
MSmits: keyboard has never hurt me so far
Astrobytes: My keyboard is fine
Astrobytes: And I use it all day every day, well, most
Astrobytes: Needs a clean though
eulerscheZahl: i got pretty good at using autocomplete in the IDE. typing those letters that i can reach easily, omitting the others
Astrobytes: Visual Studio is nice in that way
eulerscheZahl: now it is
Astrobytes: Yeah, indeed!
MSmits: do you mean the CG IDE eulerscheZahl?
eulerscheZahl: MonoDevelop had that earlier i think (to that extend)
MSmits: since it was changed
eulerscheZahl: i don't use the CG IDE
MSmits: i thought maybe for puzzles
MSmits: I use it for that
eulerscheZahl: ssorem => StringSplitOptions.RemoveEmptyEntries
Astrobytes: I use it for some puzzles and quick edits
MSmits: when i load up oware these days, it takes like a full minute
MSmits: the C++ language assist I think ?
Astrobytes: Ah not just me then
eulerscheZahl: illedan reported something like that too
eulerscheZahl: on discord
Astrobytes: SAme for a few multis, even without long text strings
Astrobytes: btw, did any CGers do the ICFP contest?
eulerscheZahl: i registered, decided against after seeing the task
Astrobytes: Looked like some reverse engineering thingy when I checked it Friday
Astrobytes: Didn't follow it
eulerscheZahl: my first thought: implement some kind of interpreter for their own madeup language
eulerscheZahl: like processor opcode
eulerscheZahl: didn't check if that turned out to be true in the end. they only had half the rules revealed at the start
eulerscheZahl: defining numbers, math operators
eulerscheZahl: so i expected branches, conditions and such
Astrobytes: That looked like the thing to do yeah, but the changing ruleset put me off even going near it
eulerscheZahl: that too. and the lengthy statement just to decide if i like it
eulerscheZahl: to me a good game has a short statement yet a high complexity
Astrobytes: Concept was fine I guess, didn't need to be so wordy
eulerscheZahl: i don't even know. TL;DR
tomatoes: easy to learn hard to master
eulerscheZahl: to enter*
Astrobytes: lol, just realised my offline Othello runner was only using 50 out of 150 ms :/
Astrobytes: (in 'CG' mode)
AntiSquid: is python_programerPP rejecting stuff for the sake of rejecting? :D
Astrobytes: Yeah, that's IAmNoob's renamed account
Astrobytes: Cheater + troll
Kariin: Took me forever to notice lol
Kariin: Dude that last line it's freaking dirty xD
PatrickMcGinnisII: your lvl is your bedroom prowess
PatrickMcGinnisII: lvl = level ... lvl looks dirty
PatrickMcGinnisII: place your ears on the e's
PatrickMcGinnisII: l e U e l
pedrosorio: I would also recommend https://www.youtube.com/watch?v=y6120QOlsfU
itzblinkzy: haha very funny..
PatrickMcGinnisII: Go hard yungsters
itzblinkzy: share code cmon!
itzblinkzy: im trying to learn here
PatrickMcGinnisII: ahh, Dlx wants some https://youtu.be/6djKy7CEKYg
PatrickMcGinnisII: i quit, darn crossword puzzle got me brain twister, glhf
Dlx: and why can't I get to finish my code when the clash is over, come on... I bet 95% of people here care most about solving it, not really winning it... so unsatisfying that i copied my code and finished it in a jupyter notebook lol
Dlx: Dang I'm doing this coding style for my next pair programming... will be working so fast and dirty on first pass that buddy will have no chanse to follow what I'm doing, just for fun, and then refactor refactor refactor after ;p
Uljahn: Automaton2000: fix it
Automaton2000: what if you could do a better job