mzbear: wtf is with the turn count in code a la mode ... 600 turns total with 50ms each, that's maximum of 30 seconds per match :O
eulerscheZahl: which is the max duration that CG allows
eulerscheZahl: on RAIC i saw 12min per match
mzbear: oh dear
mzbear: meanwhile, that bitfield thing you helped me with yesterday, ... https://paste.ofcode.org/AQNWhUuF8bzJSuMpCnYLtW
mzbear: i'm having a feeling there's an even better way to do this, but it eludes me
mzbear: ... but i already finished that puzzle, so perhaps i shouldn't bother with it
mzbear: oh, right, i could just make two loops and store the results in an array .. do upper triangle and lower triangle separately
mzbear: every once in a while i need to split a string in standard c++ code, and it makes me cry. c++20 seems to finally have string splitting out of the box via the ranges library
Robin_Hood: hello
Robin_Hood: hii
MSmits: makes me cry too mzbear
MSmits: :cry:
MSmits: it's what kept me from using c++ in contests for a while
MSmits: well one of the reasons
mzbear: didn't test, but apparently this is valid c++20: for (auto token : str | std::views::split(' ')) { /* it's christmas time */ }
mzbear: so you can basically make pipes that filter and transform things
MSmits: hmm ok
MSmits: i just want it to split my openingbook
MSmits: so far i used some unreadable code for it i found somewhere using google
MSmits: it's crazy how hard it is to just split a string into characters
mzbear: using stringstreams and std::getline is kinda okayish, but annoying nevertheless
MSmits: I probably would have much preferred rust or some other modern fast language
MSmits: but i already learned c++ and dont fancy learning another
mzbear: rust is a great language, but super annoying for prototyping anything
Wanwws: i can't understand coding
mzbear: it forces you to be quite specific about lifetimes and ownership in a way that clashes with how i've learned to code
Wanwws: help me please
Wanwws: mime type code
mzbear: every time i've ported any c++ code of mine into rust, i've found hidden bugs in the code because they're compilation errors in rust...
MSmits: you take this one mzbear, I need to finish my class :P
mzbear: wanwws, what's the issue?
Uljahn: 99% of newbs: my code fails because of bugged testcases
SupSon: Is there some doc on how to run a local tournament with local instances for a game?
mzbear: supson, I still haven't gotten to set up any local games myself, but i found this when i was looking into it earlier: https://github.com/dreignier/cg-brutaltester
SupSon: grea thanks that's what I was looking for
jacek: good morning
YusufEmad04: print("Hello World")
coldcodehah: Console.WriteLine("Hello World");
Proprogramer: System.out.println("Hello");
coldcodehah: Good Good afternoon
Proprogramer: morning for me
coldcodehah: hah
kurtesy: Good Night
coldcodehah: where night
coldcodehah: Japan?
MadKnight: not in philadelphia for sure
coldcodehah: Coding 《Shadows of the Knight - Episode 1》
MadKnight: okay
MadKnight: i like your 《
MadKnight: AutomatonNN u gotta learn to use those
AutomatonNN: first the contest comes out
coldcodehah: 《Shadows of the Knight - Episode 1》 complete
MadKnight: WOW
MadKnight: jj
MadKnight: oh wait i mean
MadKnight: гг
MadKnight: Automaton2000 u gotta be proud
Automaton2000: how're u gonna do the same for the rest of the game
MadKnight: u can't keep doing the same thing Automaton2000 it just gets old at some point
Automaton2000: guess i should do that
MadKnight: u can't keep talking to Automaton2000 forever Automaton2000 it just gets old at some point
Automaton2000: seems like a fun game
MadKnight: hmm u right Automaton2000 it will never get old
Automaton2000: should have thought of that as well
coldcodehah: Coding <Shadows of the Knight - Episode 2>
MadKnight: are u a bot
coldcodehah: error Timeout
coldcodehah: :joy:
MadKnight: wur
MadKnight: wut
MadKnight: what is that supposed to mean
coldcodehah: not bot
MadKnight: then why are u saying those things
coldcodehah: practice typing
MadKnight: coding << ctrlv >> what a nice practice, Automaton2000!
Automaton2000: not as much as you want to compare the two
coldcodehah: u are right
MadKnight: aren't u already practicing typing in coding ?
MadKnight: Kirbiby
MadKnight: no #fr Kirbiby
MadKnight: are u also here Kirbiby ?
Kirbiby: I'm everywhere
MadKnight: oh cool so i can ping u anywhere
MadKnight: anyway
MadKnight: what are u doing Kirbiby that u talked about on #fr ?
Kirbiby: AoC17
MadKnight: what is that
Kirbiby: https://adventofcode.com/2020/
MadKnight: oh cool
MadKnight: so what's your strat ?
MadKnight: are u doing it in c++ Kirbiby ?
Kirbiby: it depends the day
Kirbiby: it depends on the day
MadKnight: what is today ?
MadKnight: what is it today ?
Kirbiby: clik on it and you see
Kirbiby: today the input is not a pain to parse so it's C++ day
MadKnight: oh it's a weird random "thing"
MadKnight: what do u mean by pain to parse ?
MadKnight: what is a pain to parse only in c++ ?
MadKnight: Kirbiby
MadKnight: stop disappearing in #fr
Kirbiby: comma separated string for instance
MadKnight: wut
MadKnight: u can just split that
MadKnight: u can just getline(std, separator)
MadKnight: did u not know getline got an external argument for that ?
MadKnight: extra i mean
MadKnight: not external
Kirbiby: not it's shitty
MadKnight: u can just do getline(cin, ',')
MadKnight: just make a function for that
MadKnight: and just ctrlc ctrlv this function into every single project
MadKnight: so that u can use it like it's always there
Kirbiby: in ruby I do $<.map{|k|k.split(","}.map(&:to_i)} and I'm done
MadKnight: just make a function for that
MadKnight: cmon
MadKnight: it'll be readintarray() in both c++ and ruby
MadKnight: c++ is so painful cuz everyone tries to optimize it as much as possible to win any amounts of ticks
MadKnight: just make a wrapper for stuff and just ctrlc ctrlv the entire thing at once
Kirbiby: don't worry, you don't have to convince me
MadKnight: no i have to
MadKnight: i've been paid
Kirbiby: I love C++, but sometimes I just want to parse the input in 20 characters
MadKnight: by Automaton2000 ofc
Automaton2000: you can use it, but it doesn't seem to be a good idea in the long term
MadKnight: no Automaton2000 that is not the time to respond
Automaton2000: when you try to use the chat for the first turn of the game
MadKnight: no Automaton2000 i said quiet
Automaton2000: how many players pass to legend
MadKnight: bad Automaton2000
Automaton2000: not sure if i can figure it out
MadKnight: Kirbiby i just have an #include that gets all my stuff into the project
MadKnight: so i can parse the inputs in 20 chars in c++ too
Kirbiby: yeah, I don't
MadKnight: just fix it once Kirbiby and then u can do all the stuff in c++ painlessly
MadKnight: that deserves a big cmon
MadKnight: that will help u go further
MadKnight: u will be even able to do big things like i did - for exactly, an absolutely universal BFS
MadKnight: so i can now do pathfinding the optimal way in any project
MadKnight: and it doesn't look terrible
MadKnight: and is not a pain
MadKnight: Kirbiby
Kirbiby: Yes, if I was not so lazy
MadKnight: i can just lead u
MadKnight: i know how it's done it will only take like 5min
MadKnight: let's do it kirb
MadKnight: Kirbiby
MadKnight: *i can just lead u through it
MadKnight: u won't have to think hard Kirbiby
Maxim251: Its posible to include file in game from disk?
Hussein4K: guys
Hussein4K: will i ever get better?
AI_says_hi: Sure :)
AI_says_hi: What are you working on?
beingsalman: Hi
mzbear: Kirbiby, c++20 to the rescue! for (auto token : str | std::views::split(',')) {}
mzbear: or if you wanted them as integers, str | std::views::split(',') | std::views::transform(atoi) ... should work, i think
mzbear: oh, c++11 already introduced stoi()
mzbear: i almost never parse strings in c++ either :D
Y4enik: ребят
Y4enik: помогите в третьем
Robin_Hood: use english
Robin_Hood: not another language
net1231: thereis a mistake here
VizGhar: bruh... I dived into UTTT using bitwise operations in C++. I must say, I'm not natural :P
Kirbiby: mzbear, indeed I don't use C++20 yet
Robin_Hood: https://discord.gg/2x5W5t5w
struct: bye
VizGhar: Ah... I used % instead of /
Found it after ~hour :|
Q12: ohh this is annoy
Q12: but I see that you near the boss, you can beat him easily
MSmits: VizGhar replied to your question
MSmits: on the forum
tarikyildiz: Hi how important is math and algorithm for programming
ryantimjohn: important
tarikyildiz: So how can I learn? Will I learn math and algorithm only for python or everything?
darkhorse64: math and algorithms are not tied to a specific language
hanzexi: Hi, guys. How are you doing?
VizGhar: MSmits thanks :thumbsup:
tarikyildiz: Thanks
tarikyildiz: I know almost all the code in Python, but I felt like something was missing, what was missing is algorithm and logic
tarikyildiz: thank you very much everyone
tarikyildiz: Well we have to learn algorithms for puzzles and competitions on this site sounds very complicated questions
tarikyildiz: I will try to learn algorithm
tarikyildiz: See you
MSmits: np!
Q12: What does it - "E128 continuation line under-indented for visual indent" means?
Uljahn: i guess you should add more indentation to the line to match some part of the previous line
mzbear: Here, I googled it for you: https://stackoverflow.com/questions/15435811/what-is-pep8s-e128-continuation-line-under-indented-for-visual-indent
Uljahn: ye, just make it aligned with the previous part
mzbear: specifically, align with the first parameter
Q12: http://chat.codingame.com/pastebin/5d3e4957-6433-43f4-a3b6-1442595adeb6
Q12: sorry, I'm not so used to python :sweat_smile:
mzbear: it's saying invalid paste id for me
Q12: and the yellow marks making me crazy
Q12: wait a sec
mzbear: the pastebin was acting up for me earlier too
Q12: https://codeshare.io/arzNLE
Q12: try this
Q12: this line - "if temp_value > value:" is yellow for me and the lines below it
mzbear: hmmm α
mzbear: i think the style checker is bugged because of the variable names
Q12: this variable worked well for me before, it has nothing to do with it
mzbear: it's interpreting the input in different encoding probably
Q12: also β
Q12: I will try to replace it
jacek: alpha? beta? w00t?
mzbear: well, it definitely flags an unexpected line
Q12: still doesn't work
Uljahn: i turned the style checker off with adblock :pensive:
jacek: oO
Q12: the with this problem is that functions don't close to me so I need to scroll a lot
Q12: How did you do it Uljahn?
Uljahn: Q12: https://forum.codingame.com/t/make-the-ides-language-assistant-optional/187625/3
Q12: Thanks
Uljahn: there was an option to turn off pylint in settings but it stopped working some days ago
Q12: Yes!!! :grinning:
HuynhNgocHai: Hello boiz
HuynhNgocHai: How can i get in dark web?
Q12: I reload and it fix the problem
HuynhNgocHai: =))))
Q12: Thank everyone :slight_smile:
HuynhNgocHai: I'm newbie
HuynhNgocHai: What can i do?
Uljahn: you can practice to stop being newbie
Uljahn: play clashes. solve puzzles, make ai bots in multis
struct: Anyone knows what css I need to change to increase console output size?
Uljahn: .cg-ide-console ?
HuynhNgocHai: Thanks you so much!
i_suck_at_coding578375: ello
HuynhNgocHai: Hello bro
i_suck_at_coding578375: i still trying to get to know coding :joy:
ForBidde[N]-: {
Maxim251: YABADABA HOO , I finish Don't Panic - Episode 1 puzzle beautifully :D And Bigger succes for me is that I make logick in side of struct, where not hapens to me almost never. Logick it self is so small and works :D
struct: Ok, managed to do it, thanks Uljahn
struct: I had to change something else but now it works
Maxim251: First play with thor, this is the simplest game what I was playing
jacek: :tada:
Maxim251: #i_suck_at_coding578375
i_suck_at_coding578375: yes?
Maxim251: play with lightining thor, or somehing like that. There You need simple loops
eulerscheZahl: you just got a channel dedicated to you
i_suck_at_coding578375: ok
struct: If anyone wants, this is what I changed "left-bloc console-bloc bloc with-leaderboard"
struct: I added height: 700px
struct: and margin-top:auto
eulerscheZahl: i'm happy with the default
eulerscheZahl: except that it's pretty small
struct: yeah, that was my problem
eulerscheZahl: https://www.codingame.com/forum/t/gm-script-to-maximize-the-console-and-minimize-ide-in-combination-with-cg-sync/2581
eulerscheZahl: that was later added to CG Enhancer
struct: ty euler
struct: Ill check it out
i_suck_at_coding578375: yup Power of Thor i'm not even good at rip
i_suck_at_coding578375: im not good at Power of Thor
i_suck_at_coding578375: so i think i cant code
Maxim251: #i_suck_at_coding578375 Dont say that, it looks dificult, but there You use only 4 ifstatment
Maxim251: If I pass all statement then why I couldn't unlock achivements?
Maxim251: Oh Yeah, I forget to coment everywhere debuging " cerr "
Maxim251: Now I have achievments
Maxim251: :kissing_closed_eyes:
LastRick: Wait, there's a german channel?
ChrisSkyRo: Are you guys familiar with the CoC problem regarding finding the room Brian is in? Does anyone know the statement? I had it in reverse mode and only scored 90%, so I'm curios what I'm missing.
LastRick: anyone got euler's bookmark handy?
jacek: AutomatonNN
AutomatonNN: code as a minute
LastRick: Search here: http://eulerschezahl.herokuapp.com/codingame/puzzles/
ChrisSkyRo: thanks
LastRick: np
LastRick: Someone kick the CG Clash server. It's stuck on Reverse Mode.
Q12: If I am first in silver in UTTT but below boss would I be in gold or I have to be above boss to be in gold?
[CG]Maxime: you need to be above
Q12: thanks
struct: [CG]Maxime is there any way that I can get Yavalath, I deleted the account that created it
eulerscheZahl: as in: reassign the creator of the contribution
reCurse: Have you tried not deleting accounts?
eulerscheZahl: :heavy_plus_sign:
eulerscheZahl: looks better on discord
Astrobytes: at least we learned that he really is a man of his word
reCurse: But at what cost
Astrobytes: Indeed.
eulerscheZahl: losing contrib ownership, causing extra work for staff if they care enough
reCurse: That was rhetorical but thanks
eulerscheZahl: :(
Astrobytes: hahaha
struct: I dont care much about ownership, is just in case if images stored in imgur get deleted
struct: Icant edit contribution if im not the owner
darkhorse64: What about upvoting solutions thousands of times just to prove there is a bug ?
eulerscheZahl: and you want to add a league
struct: ah and that
reCurse: Isn't that how bug reports work? You break the thing
reCurse: Oh, just in case, /s
eulerscheZahl: and you tried to update yavalath with your current account already?
darkhorse64: Im my world not to the point to give trouble to another team
eulerscheZahl: (note how i wrote current account, not main account or alike)
reCurse: That was a joke darkhorse64
reCurse: :P
struct: I havent tried
struct: I could
eulerscheZahl: are you fine with me testing (download zip, reupload)?
eulerscheZahl: worked so far. but i think for langton the send button was the show stopper
struct: yeah its fine
eulerscheZahl: Contribution updated
darkhorse64: I discussed the matter on #fr and some people disagree with me to my surprise
reCurse: Ah ok
[CG]Maxime: struct : I've open a bug report, I can't tell you when it will be reassigned.
struct: ok, so I guess I dont need ownership
reCurse: Figured that's why I needed to add the /s
struct: Its fine Maxime, dont worry
struct: I have no rush
eulerscheZahl: all we want are community contests :angel:
Astrobytes: God loves a trier
jrke: me too euler
Astrobytes: hey jrke
jrke: hi astro
jrke: How make a statement only visible after level n in multiplayer game?
eulerscheZahl: did you propose a contest yet jrke?
eulerscheZahl: saw you asked lots of questions and didn't reveal any content
eulerscheZahl: https://github.com/CodinGame/FallChallenge2020/blob/main/config/statement_en.html.tpl
jrke: https://www.codingame.com/contribute/view/585178f84eb12adf8e123018a1d424f1dfc9
^^ here is the public content euler
eulerscheZahl: oh :D
eulerscheZahl: i'll keep my game for myself atm
jrke: i saw that fall challenge statement tpl file as well as spring challenge but didn't get how to do
eulerscheZahl: define leagues
eulerscheZahl: then write blocks specific for one league
eulerscheZahl:
...
eulerscheZahl: when you build your project, it will create the individual statements from it
jacek: night of war = fog of war?
eulerscheZahl: first impression: your logo needs transparency
jrke: i know and i will fix it
jrke: https://pastebin.com/8xryWaej - why entrance green box statement still exists
struct: What do you mean jrke?
eulerscheZahl: be has a green box without
and now he's confused why it's always showing
Q12: where does the time start in UTTT? after this - "opponent_row, opponent_col = [int(i) for i in input().split()]"
Q12: ?
struct: after reding first input
Q12: ok thanks
Q12: Wow!!
Q12: I didn't understand why I don't have time to do minmax, but now I understand and it is hella better!
Q12: Thanks struct!
struct: np
Q12: Boom! gold league in UTTT in 50 minuts :grinning:
Q12: minutes*
struct: grats
Q12: Thanks!
jacek: aww
Latage: hello
Latage: i need help
Latage: i need help with c++ code
IfIHadATail: just ask your question and if someone can they will answer you
IfIHadATail: it's the best way to go about it
Latage: its hard to explain, because i need to optimize my code
jacek: oO
struct: Oo
Maxim251: #Q12 congratulations. :( For me, gold lague is as far like distroyer from star wars
VizGhar: Maxim251 you don't have to type # or @ before name... just start typing name and press TAB, it will autocomplete
VizGhar: # is channel reference btw
jacek: good to know, #VizGhar
Maxim251: Oww, thanks
fumio: VizGhar
Maxim251: But color of name is nicer, And is easy to spot if somebody writes to You.
struct: Maxim251 it will show as red
jacek: struct impossibru
VizGhar: jeez, I'll keep my mouth shut next time :D
eulerscheZahl: no, keep talking #VizGhar
jacek: or AutomatonNN
AutomatonNN: go #1 back to the contest
VizGhar: (╯°□°）╯︵ ┻━┻
**VizGhar slaps eulerscheZahl around a bit with a large fishbot
eulerscheZahl: :open_mouth:
struct: :tada:
twitlydoof: somebody sent me a message and i accidentally closed the chat in which i got the message, anyway i can see what it was
struct: refresh
eulerscheZahl: when you close the chat, it won't reopen itself
eulerscheZahl: that bug got fixed :P
struct: Nice a bug that doesnt happen localy
struct: :(
struct: oh it does :D
twitlydoof: another validator bug I think at https://www.codingame.com/clashofcode/clash/report/1507914b9d3360758627028c76130327c91d4c0
eulerscheZahl: we can't see the task or your code that way
eulerscheZahl: http://eulerschezahl.herokuapp.com/codingame/puzzles/
find and link the contribution
struct: I was gonna edit the puzzle that made me leave CoC, but it was a blessing
wholesome: gamer
Josh_Fullr: if someone knows how to code C#, I don't know how to code it and need help. Can you make an introduction to it?
VizGhar: There is nobody here that will teach you, try to find some online course first
Josh_Fullr: ok, thank you
VizGhar: or check the "LEARN" section
Josh_Fullr: I have
albo1337: use codecademy
twitlydoof: what's your motivation behind learning C#?
jacek: so he get rid of the glasses
Astrobytes: badum-tsss
Astrobytes: he's here all week folks
struct: :clap:
VizGhar: I've finally managed to implement "possible moves" in UTTT (took me like 2 hours). Now scoring and MCTS
My brain Isn't working like it used to during university
VizGhar: this is something i was able to do in 30 minutes back then :P
jacek: noice
jacek: keep in mind in CG, if no one has 3-in-row in main board, the one with most won small boards wins, unlike the normal UTTT rules
VizGhar: yeah. I've read that and watched a few replays...
struct: I wonder if its worth to send the states back, from depth 4 to depth 3 once the move is done, this is what it normally looks like
struct: States / Score / Move / Time: http://chat.codingame.com/pastebin/6cff6c07-273d-4542-808b-d0f3e8bdcd55
VizGhar: I also challenged myself and trying it with C++ (I'm android dev so Kotlin is my default)
struct: Not sure its worth just to shave 1ms
VizGhar: What's that struct?
struct: From STC
struct: I try all the moves until depth N
struct: in that case 4
struct: I could save a little time on next turn and choose the Depth 3 from the depth 4 of the previous turn
struct: same from depth 2 and 1
struct: But just to save 1 ms, is probably not worth
VizGhar: (hope you are not waiting for my approval :D)
struct: Guess Ill do something else and leave this for later
Astrobytes: just a struc t braindump, don't worry ;)
struct: Yeah, I like to share my progress
struct: Very rarely I reach depth 8 if board is almost full
struct: But by then its too late to win
struct: Still I can do perfect move sthen
jacek: endtables?
struct: not endtables
struct: because the game might not end in that depth
struct: Also is very rare to get depth 8
struct: Only if I have 4.7 moves average over 8 depth
struct: Which is very rare, since the least amount of moves you have is 2 if both stones have different colors
MSmits: this uttt rule where the player with the most boards wins in the CG version is actually the rule that makes it solved for player 1 as win I suspect
MSmits: because player 2 can block 3 in a row, but not at the same time get as many won boards
MSmits: it's basically just an extra way to win for the player who has advantage
MSmits: played a couple games without this rule on some websites and it was pretty easy to draw as p2
struct: How much would game change
struct: if there was swap rule?
MSmits: fundamentaly
MSmits: +l
MSmits: it would probably be solved as a draw then as i suspect yavalath is
MSmits: and it would be very balanced
MSmits: right now it's quite boring at top level play unless you're like me
struct: Not much to improve on?
MSmits: It took me just a few days to make a bot that wins vs kar liso as p1, 100% of the time and he doesnt even use an opening book
MSmits: thats because the better your bot is the more narrow your line of play and his bot is very good
MSmits: mind you this does not work all that well vs tomatoes, but i suspect it still does if i take the time
struct: So you changed stuff besides book?
karliso: interesting
MSmits: nah, but i am working on that. I do like to be able to say i have the best bot and i currently dont
MSmits: hi :)
MSmits: I studied your rollout karliso, thanks for sharing that
MSmits: i noticed you have a lookup table for picking the mainboard
MSmits: on free moves
MSmits: that seems important
karliso: hi :)
karliso: oh yes
karliso: You mean a differen lookup table..
MSmits: i was guessing it prefers important boards where 3 in a row is still possible
MSmits: since you also pass your own mainboard
MSmits: as opposed to just the whole thing
MSmits: or maybe some preference to middle board if it can
karliso: Probably, I don't rememebr exactly. Also maybe 4 more likely than others?!
MSmits: there could be some weighted random in there that you tweaked
MSmits: isnt this a huge table?
MSmits: i wonder if this gives you cache problems
MSmits: because its [myboard][mainboard][random]
karliso: My bot is very well memory-optimized.
MSmits: yeah but 18 bit * whatever the size of the random part is
MSmits: thats a lot
karliso: Random is not big.
MSmits: ah ok
MSmits: preference for winning miniboards was interesting
MSmits: i always had block + win as equal value and randomed between them
MSmits: seemed better than win, but maybe i didnt test it well enough
karliso: idk, I have not even tested other way.
struct: karlis o should do Yavalath, so you have more competition :)
MSmits: its weird, you do realize a lot of players have experimented so much and you just try this way and win with it :)
karliso: I have tried many things, but not block/win as equal.
MSmits: ohh ok
MSmits: well its nice because you dont need to distinguish p1 and p2
MSmits: it uses the same array
MSmits: because block for one, is win for the other
MSmits: well i suppose it might be different for you, i use a ternary lookup
MSmits: still wondering if i should not just go back to binary
MSmits: ternary gives you a smaller array, but slower computations
karliso: yeah, I never tried ternary as I always felt that binary is going to be faster
MSmits: it is, but it depends on how many lookups you do
MSmits: you do "compute winning moves"
MSmits: I look that sort of stuff up
karliso: These are frequent lookups so they should be in cache.
MSmits: probably i should look again in how to do that fast, when i first wrote the bot i wasnt really good at bit tricks
MSmits: oh but your compute winning moves is not a lookup is it>
MSmits: i think you do a calculation
karliso: It is a lookup.
MSmits: ahh ok
MSmits: couldnt you make the function prettier btw? With the if/else structure for the two players?
MSmits: seems very symmetric
MSmits: except you count a draw for p1 as a loss somehow
MSmits: the mewhite? thingy
MSmits: i heard this before
MSmits: people doing this
karliso: There is often a tradeoff between pretty code and performance.
MSmits: yeah I guess so
karliso: This code is written to be fast.
MSmits: what is UTTT:ENDGAME7 btw?
Skynse: Finished my exams today
Skynse: I can code all I want now
MSmits: grats Skynse, hope you passed
MACKEYTH: I think that's out of date, karliso
Skynse: So far I'm doing great, thanks
MACKEYTH: Modern compilers are hella slick. You're better off writing for legibility
MSmits: well what you can do is, try to prettify it, then just profile
MSmits: if its just as fast, keep it
Skynse: How do I make the chat stay on "World" when opening new codingame pages?
MSmits: no idea
karliso: idk if I want to explain ENDGAME7
MSmits: then dont
MSmits: always keep stuff to yourself if you think it is important
MSmits: I havent shared some stuff also
MSmits: that made my bot a lot stronger
karliso: I put my uttt code in gitbolt and examined generated assembly.
MSmits: nice
MSmits: i think mostly you are good at these multis and contests because your intuition is good btw. I mean performance helps, but it;s not that big of a deal
MSmits: uttt is kind of an exception in this regard, performance helps more for this game than for other
MSmits: but still other things are more important i think
MSmits: choosing what moves to prefer, what to prune, what to evaluate etc.
SupSon: how many rollouts do you get 3rd ply
SupSon: for uttt
QuiqueLuna: Hola Amigos
Skynse: Hola mi Amigo, tu hablo anglais?
MSmits: that that much currently, only like 90k, but i do some stuff that makes the rollouts heavier
MSmits: but before i had 130k or so
ElJona174: Hola a todos
SupSon: wow that's crazy
SupSon: i'm only getting like 60k
ElJona174: Algún consejo?
MSmits: karliso and recurse both have near 200k
SupSon: do they do full rollouts
MSmits: you mean every child?
MSmits: a sim
SupSon: i mean for the sim
SupSon: do they play for the end
MSmits: yeah
MSmits: there's a thing some players do where they rollout from ever child on expansion
MSmits: so thats like 7-8 sims after expansion
SupSon: ah ok
MSmits: i dont do this
MSmits: I *think* karliso does this
MSmits: but recurse doesnt
MSmits: hes got about 180k
SupSon: you use solver extension right?
MSmits: yeah
SupSon: do you propagate proven draws
SupSon: or only wins and losses
MSmits: i do draws also
SupSon: You can only propagate a proven draw if all the children are draws right?
MSmits: draws and losses
MSmits: at least 1 draw
SupSon: how do you differentiate between between a proven draw
SupSon: and a normal draw from a sim
MSmits: you mean on selection ?
SupSon: 0 exact and 0.0001?
MSmits: I simply dont select solved games
SupSon: no when you propogate the result up the tree
MSmits: ohh
MSmits: you do the -inf +inf thing from the paper?
SupSon: yea for proven loss and win
MSmits: i dont do that
MSmits: i just have a separate property
MSmits: "status" = -1,0,1, 2
MSmits: 2 is ongoing
SupSon: i guess you don't have to check the value multiple times then for the node
MSmits: what do you mean?
SupSon: to detect whether its a proven value
SupSon: -inf or +inf
MSmits: well i check this property instead
SupSon: yea cool
MSmits: i do it like that in every board game basically
MSmits: it works, might not be the 100% fastest way, but if not, the cost is negligible
SupSon: how do you store the local boards
SupSon: i'm using a ulong and an int combiend
MSmits: i currently use ternary storage, but i might switch back to binary
MSmits: ternary is nice for lookups
SupSon: what's ternary storage
MSmits: basically a board with 1 mark in the corner is 1 if its a cross
MSmits: 2 if its a circle
MSmits: so 0, 1, 2 for first mark
SupSon: hmm alright
MSmits: 0, 3, 6 for second
MSmits: so it's 3^9 possibilities
MSmits: instead of 2^18
MSmits: its about as packed as you can make it realistically
MSmits: there's even tinier methods
MSmits: but they take too much time to convert between
slvt_c0d3r: wassup noobs
MSmits: in binary there are 262144 possible boards
MSmits: in ternary 19683 boards
MSmits: in the smallest version 8629 boards
MSmits: well thats the array size i mean
SupSon: hmm I don't have any lookups for the complete local board state
SupSon: I only use one players
MSmits: yes and thats fine, karliso does this too and he's nr 1
SupSon: so I only have 512\
MSmits: i went really far down the lookup hole
MSmits: you can do some cool stuff knowing the entire board though
MSmits: for example you could have a lookup array for boardvalues and bias selection
MSmits: i dont do that, but it's possible
SupSon: sounds like it'd slow down everything haha
MSmits: yes, but in many games thats worth it
MSmits: in yavalath, pb4's bot is about 5x slower than mine, but his rollout is more accurate
MSmits: the bots are equally strong basicaly
MSmits: (not counting my opening book)
MSmits: uttt is just very resistant to this. It's hard to find changes that are worth giving up performance to
SupSon: yes
MSmits: do you do anything fancy?
MSmits: in uttt?
MSmits: or just plain mcts?
MSmits: your bot was quite strong
MSmits: so i suspect some kind of biased rollout
SupSon: just solver extension
MSmits: weird, 60k rollouts does not seem enough to be on nagraroks level
SupSon: I'm a little below him
MSmits: ahh ok
MSmits: he's tough to beat even now
SupSon: he used to be #1
MSmits: for a very long time too
SupSon: what got you to beat him
MSmits: hmm several things, i am not sure in what order i found them
MSmits: i'll tell you the stuff thats public
SupSon: teccles seems new
MSmits: teccles heuristic helps a ton and is one thing but i found it after i beat him
MSmits: biasing toward avoiding free move for the opponetn helps
MSmits: i prefer winning and blocking moves in the rollout
MSmits: random playout i mean
SupSon: What's a blocking move
MSmits: make sure 2 dont become 3
MSmits: basically play the move that opponent would play to win the miniboard
MSmits: the reason i do both blocking and winning is not to bias the rollout too much
MSmits: in my case it was weaker when i did just winning
MSmits: but karliso does just winning apparently.
MSmits: but it might be this depends on the stage of the game or something
MSmits: not sure
MSmits: you can go pretty far
MSmits: experimenting
MSmits: its just annoying that so few things actually work
MSmits: also as was said, at the highest level bots sometimes become deterministic enough to just beat them with hardcoded moves
MSmits: but this does not work on 99% of players
MSmits: doesnt really work on nagrarok anyway, i tried :P
SupSon: too stronk
MSmits: too randok
MSmits: random
MSmits: it works fine on recurse and karlis
MSmits: i recommend every player to add a bit of random
MSmits: makes my job a lot harder
MSmits: doesnt work on you either btw, i remember trying it
MSmits: oh of course if i spent a month it would work
MSmits: but you cant do that for each player
MSmits: I should go get some sleep, gn :)
SupSon: later dude thanks
MSmits: any time
ChampionCoder: Good morning, guys!
Skynse: bedtime for me lol
Codeab: morning
atahiri: hello all
mzbear: mmh. started working on Yavalath bot :D
mzbear: first version submitted: takes a random move, except always takes 4 in row, always avoids taking 3 in row.
mzbear: random truly isn't a very strong strategy :D
Robin_Hood: hi
Robin_Hood: Good morning
Robin_Hood: hello
