Chat:World/2021-02-23
MSmits: that's obscene language in Dutch
MSmits: careful, or I will find a Dutch moderator
MuhammadMehediHasan: wth :wow:
reCurse: Enough spam
reCurse: Thanks
MuhammadMehediHasan: How you get a moderator badge?
reCurse: Why do you care?
MuhammadMehediHasan: you're a real mod?
reCurse: I can prove it
MuhammadMehediHasan: It has already proved!
jacek: oO
Kitkat260: is anyone here really good at clash of code/
Kitkat260: ?
Astrobytes: Not a single person, no.
Kitkat260: ok
NOTsteve: Do you need help with something or were you just wondering if anyone feels confident?
Kitkat260: sup
Oeloeloeloe: Whats up
Kitkat260: hru
Oeloeloeloe: Great, hbu?
Kitkat260: ehh
Kitkat260: school
Kitkat260: is crazzy
Oeloeloeloe: How come?
PrinceJordan: this is difficult they don't but real prince's through this lol
Kitkat260: i hate homework
Oeloeloeloe: I understand
Oeloeloeloe: Used to hate it too
Kitkat260: yea
Oeloeloeloe: Try to make it fun
Oeloeloeloe: It helps
PrinceJordan: i dont care for homework i have no prob with it
Kitkat260: wow
PrinceJordan: i work way better under pressure so thats why im ok with it
Kitkat260: nice
jacek: AutomatonNN eh
AutomatonNN: eso in companies offer the link to it and then we use it for you and then the contest is over by the
jacek: AutomatonNN should be called AutomatonMLP
AutomatonNN: what do you mean by the contest?
Astrobytes: Hmph. At least we had peace for a day.
Uljahn: my little perceptron
Astrobytes: Better than multi-layered pony I guess
RustyMysteryFromTheSwamps_fe5b: Hi guys
AntiSquid: drain the swamp!
AntiSquid: hi
Saphine: #clash https://www.codingame.com/clashofcode/clash/16078424e2f0294caf387125ad5c122193e9bf0
Astrobytes: Saphine: The idea of the #clash channel is that you click on #clash and post the link in there ;)
Astrobytes: Then you can say "would anyone like to join my #clash"
Astrobytes: Although I'm not sure why I care tbh.
jacek: you're mod, it's your mission
AntiSquid: i consider this "mistake" to be intentional
jacek: youre overthinking
AntiSquid: tell me what i should think, i am dependant on others to have my own opinion
jacek: jacek is op
Saphine: Oh sorry
PatrickMcGinnisII: jacek is PEBCAC AutomatonNN
AutomatonNN: what do you mean?
AnaPaulinaRamirezSierra: Hi, guys does anybody knows the solution to the game
BlaiseEbuth: Ze solution to Ze game. The only one. My preciouuuus.
AnaPaulinaRamirezSierra: ANEO sponsored puzzle... I think I half the way.. but I cannot get through
BlaiseEbuth: If you stop in the middle of the way you'll probably be hurted when the lights will turn green...
Kitkat260: hello
AnaPaulinaRamirezSierra: I think , I am not understanding the problem at all
AnaPaulinaRamirezSierra: so I already calculated the average distance
AnaPaulinaRamirezSierra: sorry speed
AnaPaulinaRamirezSierra: but when I play all the test cases, it doesnt work
Kitkat260: same
Kitkat260: same
BlaiseEbuth: Do you mean that you're printing the same answer for all test cases ?
Kitkat260: yes
AnaPaulinaRamirezSierra: yeah, depending on the output but I am not getting the correct answer
BlaiseEbuth: That's not what that is expected. Your code must compute the answer of each test case.
AnaPaulinaRamirezSierra: 50 1 200 10 like this case I am just getting 76
AnaPaulinaRamirezSierra: I dont understand then how it gets an answer of 36
AnaPaulinaRamirezSierra: I thought that by following the formula speed=distance/time would get the right answer
AnaPaulinaRamirezSierra: but nope
BlaiseEbuth: Have you read the statement ?
AnaPaulinaRamirezSierra: yeah, but probably I am not getting it right
AnaPaulinaRamirezSierra: I think thats my problem
AnaPaulinaRamirezSierra: I will read again
Imnewhelp: how to solve it lol
AnaPaulinaRamirezSierra: I think this problem involves a bit more math
AnaPaulinaRamirezSierra: and analythical skills
PatrickMcGinnisII: usually people forget to convert km/h and duration to more manageable integers
Imnewhelp: im new to this "world"
Imnewhelp: how do you collect even and odd numbers?
Imnewhelp: and by "sum" do they mean add them together?
PatrickMcGinnisII: try multiplying both by 3600 then compare the computed time float by integer division by each duration...if even...then it hits a green light
PatrickMcGinnisII: if you looop from your maxspeed down to 1
Imnewhelp: why multiply it by 3600
Imnewhelp: do you have discord
Astrobytes: did someone miss their Ritalin dose today?
PatrickMcGinnisII: and at each of these speeds, count how many green lights are valid... if you hit them all, then you have your answer
Astrobytes: Imnewhelp: stop talking rubbish and interrupting
PatrickMcGinnisII: time=dist/spd
Imnewhelp: are we on diffrent code servers lol
jacek: its global chat ~
Astrobytes: Yeah, I think he meant to type in his clash chat
AntiSquid: it's ok
AntiSquid: just stop playing clash
Astrobytes: np, just be careful in future
AntiSquid: hold up
AntiSquid: play some multis
PatrickMcGinnisII: determine an easy way to see if car hits the green window is easily done with (int)(tryingspeed/lightduration/lightduration)%2==0 then car went thru on a green light
AntiSquid: https://www.codingame.com/multiplayer/bot-programming
PatrickMcGinnisII: oops messed up formula
Astrobytes: PatrickMcGinnisII: make sure to tag AnaPaulinaRamirezSierra in your replies or it'll get lost in random chat
PatrickMcGinnisII: lightdistance/tryingspeed/lightspeed
Imnewhelp: how long have you guys been playing this game
PatrickMcGinnisII: AnaPaulinaRamirezSierra make sure to convert the values appropriately km to m and h to s
Imnewhelp: i wouldnt call it a game, its more like a torture chamber until you gett gud
AnaPaulinaRamirezSierra: yeap, I did, but I think my approach was not correct I found a web page which has a fair explanation
PatrickMcGinnisII: damn, still screwd up formula
PatrickMcGinnisII: i meant lightduration instead of lightspeed
Astrobytes: don't do that in an important physics calculation :P
PatrickMcGinnisII: AnaPaulinaRamirezSierra the puzzle is pretty easy, it's like 20 lines
AnaPaulinaRamirezSierra: did you solved it already?
AnaPaulinaRamirezSierra: https://www.xarg.org/puzzle/codingame/aneo-sponsored-puzzle/
Astrobytes: Wasn't there a common pitfall involving decimal places or something?
AnaPaulinaRamirezSierra: I found this solution
AnaPaulinaRamirezSierra: I think my approach was incorrect, I didnt think that much
Astrobytes: Well, it's not the best idea to look for full solutions before you solve it
AnaPaulinaRamirezSierra: yeah, I know but I havent seen the solution
PatrickMcGinnisII: Astrobytes yes, that's why multiplying the inputs by 3600 makes the intdiv work to determine odd (red light) versus even (green light)
AnaPaulinaRamirezSierra: so somebody might be able to help me
AnaPaulinaRamirezSierra: to explain it
Astrobytes: gotcha PatrickMcGinnisII
AnaPaulinaRamirezSierra: please
PatrickMcGinnisII: I can't say much more without giving it away
PatrickMcGinnisII: loop from maxspeed down to 1
PatrickMcGinnisII: in each loop, sim the car ... hit a red light...then that speed isn't a solution
AnaPaulinaRamirezSierra: Ok, I think I was thinking about a linear solution but it is different
AnaPaulinaRamirezSierra: I will do the drawing
PatrickMcGinnisII: there are prolly a 100 different solutions, each lang. delas with prcision differently
PatrickMcGinnisII: deals w/ precision..i can't type
AnaPaulinaRamirezSierra: ok, got it. I need to practice more my analytical skills
PatrickMcGinnisII: ashketchum made sure board is basically 63 bits, so the simple ways of determining a win with bitwise operations takes some head scratching
PatrickMcGinnisII: sorry working on connect 4
Neumann: CGBenchmark has been fixed to work with the new APIs : https://github.com/s-vivien/CGBenchmark/releases/tag/1.2.0
jacek: connect4 eh
PatrickMcGinnisII: I need 72 bits, I only have 63 :cry:
jacek: why 72 bits?
PatrickMcGinnisII: I need a 0 row to prevent false positives.... so 8x9
jacek: use some masks
Astrobytes: thx Neumann
PatrickMcGinnisII: *sigh*, storing the board in 3 64 bit ints is ok, but the easy ways to test for a win become more complex
PatrickMcGinnisII: can do it in 2
PatrickMcGinnisII: the 3rd is basically p0|p1 anyways
PatrickMcGinnisII: sry
PatrickMcGinnisII: <-- talks to himself in chat apparently
jacek: welp 63 bits is ideal for bitboard stuff
jacek: you only needs 2 int64s
PatrickMcGinnisII: a normal connect 4 board 7x6 would be easier
Astrobytes: It's not supposed to be easy.
jacek: but boring. its solved
jacek: maybe solution isnt within reach in cg constraints, but its very close
PatrickMcGinnisII: yea ok, 2^64 ...heh
PatrickMcGinnisII: 2^42 isn't THAT far odd, but yes, there are more online discussions of the normal board
PatrickMcGinnisII: odd=off
jacek: https://tromp.github.io/c4/c4.html
jacek: 8x8 is solved
PatrickMcGinnisII: with an opening book no less
reCurse: Checkers is solved too and yet
PatrickMcGinnisII: yea, i see the need to twist it a little
PatrickMcGinnisII: keeping the width an odd number of columns seems to be an advantage of sorts
Imnewhelp: imagine having a social life:joy:
jacek: :scream:
reCurse: You sound like a quality individual
PatrickMcGinnisII: "Stay away from my coffee or I'll drink it from your skull" ;)
PatrickMcGinnisII: lol PHP_INT_MAX<<1 = -2 ... fml
Astrobytes: Honestly Patrick yourPHP masochism is always impressive.
Astrobytes: *your PHP
PatrickMcGinnisII: ((PHP_INT_MAX-1)<<1) = -4
PatrickMcGinnisII: i'm guessing it's a 2s complement thing
PatrickMcGinnisII: yup, self torture
Astrobytes: CGSM
PatrickMcGinnisII: ahh CG sado masochism, took me sec :joy:
Astrobytes: :grin:
AntiCheap: When will spring game be?
AntiCheap: I'll need to dehibernate
Astrobytes: Spring probably.
PatrickMcGinnisII: yea i don't feel like doing a breadth first alphabetaminmax with a node limit of 1000, with some quarklike eval ... I'm losing my mind
Astrobytes: Just use C ffs! Not like you don't know it!
PatrickMcGinnisII: 9^x=1000 anyone know?
PatrickMcGinnisII: depth 3 i guess
Astrobytes: yes
PatrickMcGinnisII: i have done 20k nodes on some simple games
Astrobytes: Honestly man it's an exercise in futility for a lot of multis. Puzzle-wise sure. You might as well be controlling a nuclear reactor with VB
PatrickMcGinnisII: you're not wrong
Astrobytes: At least use the fortran compiler we have on here :P
PatrickMcGinnisII: haven't done fortran since like '89
PatrickMcGinnisII: If I WAS using VB, i would have a loaded gun in my lap
PatrickMcGinnisII: I like java more than c#
PatrickMcGinnisII: exercise in futility
Astrobytes: You are a true mystery.
Astrobytes: I firmly stand by my CGSM label.
PatrickMcGinnisII: I did alot of C++ and ASM graphics back in the day ... I don't wanna pull the boxes of code out just yet
Astrobytes: Be a helluva lot easier than this torturous process!
PatrickMcGinnisII: I guess I'm trying to do the same speed improvements with php that I would do with C# anyways
PatrickMcGinnisII: but i could be sloppier with more power
PatrickMcGinnisII: hmm
PatrickMcGinnisII: yup, you might be triggering me to do the lang. switch on CG, but I still have other php projects going on non-CG
Astrobytes: Well don't give up PHP! And feel free to use it on puzzles etc but for the love of god help yourself wrt multis :)
AntiSquid: https://www.youtube.com/watch?v=wE-aQO9XD1g seen this yet ? it's 3D too so you can drag and look around
PatrickMcGinnisII: i'll prolly see if I can get into wood1 of connect 4, then make sure my Vis. studio still works
Astrobytes: good!
AntiSquid: when did connect 4 appear ?
darkhorse64: last week
PatrickMcGinnisII: I use eclipse for java, i know that works
AntiSquid: eclipse? wow unreal
Astrobytes: Not changed much since Viking AntiSquid :P
Astrobytes: Eclipse?! I'm gonna label you asa pure BDSM over CGSM if this continues Patrick
AntiSquid: eclipse is closer to injecting bleach into your veins
PatrickMcGinnisII: AntiSquid that YT link is pretty damn kewl
PatrickMcGinnisII: ok, you use VS for java too?
AntiSquid: VSC or intellij
Astrobytes: IntelliJ or VSCode
PatrickMcGinnisII: IJ?
PatrickMcGinnisII: ahh
AntiSquid: i am not a fan of VS, i get it some people like it ...
Astrobytes: IntelliJ is great for java but it's a bit slower than VSC for me, ymmv
PatrickMcGinnisII: so I should be using Cordova?
Astrobytes: You what?
AntiSquid: what do you want to do ?
PatrickMcGinnisII: hmm, trying to figure out what plugin i need for android
jacek: android studio
jacek: forget cordova :unamused:
AntiSquid: ya why not android studio ?
AntiSquid: too popular?
AntiSquid: phone gap worth trying too btw
PatrickMcGinnisII: i'm used to the JDK thru eclipse actually
jacek: cordova is phonegap
AntiSquid: http://chat.codingame.com/pastebin/11b5dd42-b7b2-4279-8417-23c1a73d380a
jacek: huh
AntiSquid: is it? i don't know anymore
PatrickMcGinnisII: Android Studio only supports java for programming. On the other hand, Eclipse supports C, C++, PHP, Java, Perl, C#, and many other languages as well. ... When it comes to user interface Android Studio is much easier and quicker. Eclipse is not a native Android IDE and thus, is more complicated than Android Eclipse.
AntiSquid: the why do i find articles titled "Adobe PhoneGap vs Apache Cordova" jacek ?
Astrobytes: Yeah it's Apache Cordova now
PatrickMcGinnisII: Wow, I am into BDSM
AntiSquid: so it closed down and got bought ?
Astrobytes: screenshotted for future reference PatrickMcGinnisII
PatrickMcGinnisII: The screen layout looks SOO much easier in AS
AntiSquid: you can at the very least use kotlin in android studio last i checked
AntiSquid: and with some plugins you can use other languages ...
PatrickMcGinnisII: ok, yall put me in a google hole
Astrobytes: Test failed. Abort.
Scarfield: Astraborts
Astrobytes: Scabortfield
Scarfield: any new plans to escape
Astrobytes: Don't think so, they'd have announced if there were new scenarios I guess
Scarfield: oh i thought someone mentioned there were new ones available
Astrobytes: Oh?
Astrobytes: I still have no access so can't check
Scarfield: they are no new ones
Astrobytes: 'there are not' or 'they are not'?
Scarfield: xD
Scarfield: there are no new ones available
Astrobytes: aw fk even I fked it, 'there are no' fml
MSmits: you're both crap at language
MSmits: just fyi
Scarfield: xD
MSmits: :)
Astrobytes: Thanks man
MSmits: yw :grin:
Astrobytes: :rofl:
Astrobytes: Did you try the escape thingy MSmits? I can't remember if you did or not
MSmits: nope
Astrobytes: It's kinda fun tbh, you need voice/video though. I did the first one with just text from my end - not good
MSmits: just submitted my 4th (?) completely rebuilt random rollout just now... hope this does something, prolly not :P
Astrobytes: A rebuilt random rollout ...
MSmits: oh teamwork game you mean
Astrobytes: Yeah
Scarfield: :v: (imagine the fingers crossed)
MSmits: the worst thing is that most of them get stuck between 5 and 10 or something. So not only did i fail to improve anything, I dont know which one to continue with :P
Astrobytes: Sounds like you need: ML!
jacek: and messing up with the leaderboard
jacek: my little what?
MSmits: no thanks, if i do ML it wont be uttt
Scarfield: xD
MSmits: lol
Astrobytes: Scarfield: how can there not be a fingers crossed emoticon thing
Astrobytes: My little problem
Astrobytes: jacek
Scarfield: yea i know, its ridicoulus
Astrobytes: there's also no salt one on here, but discord has it. I feel salty about that :salt:
MSmits: bot doing reasonably good, but it's probably a lucky streak
Scarfield: why are we even here. without these emojies whats the point
Astrobytes: Life is meaningless.
MSmits: life is empty without a screaming cat
jacek: :soccer:
MSmits: :scream_cat:
Scarfield: wait there is always "xD" aahh im happy again
Astrobytes: :scream_cat:
Astrobytes: Here he comes to save the day. EeeeecksDee is on his wayyyy
Astrobytes: (Mighty Mouse reference)
MSmits: ohh yeah i remember that
MSmits: yellow cape right
Scarfield: https://youtu.be/T0HLAKm1_1Q?t=2 xD
Scarfield: no idea what this is
MSmits: ahh not his cape, the rest is yellow
PatrickMcGinnisII: :beer:
Astrobytes: It was a cartoon about a... well, like a super mouse
MSmits: it was good
MSmits: i liked it
Astrobytes: iirc it was quite funny
MSmits: bah i was doing so well. now getting wrecked by re curse NN
Astrobytes: heh heh heh
MSmits: wish i could do proper benchmarking
MSmits: but cg bench not working
jacek: it is now
MSmits: really how?
MSmits: new release?
jacek: https://github.com/s-vivien/CGBenchmark/releases/tag/1.2.0
MSmits: yay
MSmits: thanks
MSmits: my new bot has 60-70k rollouts
MSmits: much heavier though
Astrobytes: Yeah was about 8pm or thereabouts he posted the update in here
MSmits: he did that fast
Astrobytes: Don't think it was a particularly difficult thing to fix, something about the JSON response returning the result object as the root (you've seen JSON responses)
MSmits: i've seen loads of texty stuff if thats what you mean
MSmits: thats professional jargon btw
Astrobytes: lol yeah, but the hierarchy/format of the JSON response.
MSmits: not lately
MSmits: but i'm totally ok with that
Astrobytes: Before there was some other shit before you got to the actual result of your API request
MSmits: they removed the :poop: ?
MSmits: no more jobby in the json?
Astrobytes: Didn't check, probably just moved it
Astrobytes: OMG
Astrobytes: jobby in the json
Astrobytes: Actually spat out my tea
MSmits: :grin:
Astrobytes: Well played
MSmits: i have a bad memory for some things, but I can do fairly well with poopywords
Astrobytes: I think that's a male thing
Scarfield: looking forward to the chat word image from euler with a super large "jobby"
MSmits: hmm maybe
MSmits: yeah i noticed my wife is not into poop
Astrobytes: I refuse to ask why or how you found this out.
MSmits: better not
Astrobytes: Scarfield: :rofl:
MSmits: my submit: http://cgstats.magusgeek.com/app/multi-tic-tac-toe/msmits
MSmits: I think it's an improvement. I did a complete rewrite of sim, so tons of room for experimentation
MSmits: wrecked by re curse, but its hard to draw conclusions from that
Astrobytes: Definitely hitting the wall wrt reBless, tomatoes is giving you a run for your money
MSmits: 50% tomatoes
MSmits: bad run vs vingte though
MSmits: he definitely improved
Astrobytes: I'd be more concerned about fixing that 50%
reCurse: Why don't you spam submit that version then for a change :P
Astrobytes: initially
MSmits: hmm, i kinda owe it to you dont I reCurse
MSmits: one more should do it maybe?
Astrobytes: heh
MSmits: another submit incoming
reCurse: Tempting to resubmit
MSmits: dont
MSmits: I will fix it
MSmits: you'll be nr 1 for a little while
reCurse: Nah but I can get higher by resubmitting
reCurse: More games vs you :P
MSmits: more than if i spam>/
AntiSquid: submit wars?
reCurse: Yeah because then I get games vs kar too
Astrobytes: Get a room you two :P
MSmits: ohh
reCurse: Alright F it
reCurse: Let's do it live
MSmits: I think my new bot is much stronger vs lower ranked players than all the other ones i had
Astrobytes: eSports!
AntiSquid: good stream opportunity, just add some trash talking, throw some dirt at each other
reCurse: There's a million other better options for that
Wontonimo: well, whatever you are doing is causing me to go up in rank, so keep it coming
MSmits: probably random though
Astrobytes: that's usually not a good thing unless you're top 5
Scarfield: one day there will be an anime about your uttt rivalry
reCurse: CG anime
reCurse: There's enough material for a couple of seasons without fillers
MSmits: I don't feel that much rivalry with re curse atm because I'm like rank 4-5 atm
Astrobytes: More of a classic kung-fu movie I think
jacek: https://www.youtube.com/watch?v=RxCukwjjhhg
Astrobytes: You beat my beloved master and adopted his techniques!
Wontonimo: I really think there is a rock-paper-scissors with the bots going on, and trueskill doesn't capture that there are different flavours. It's a fairly basic machine learning problem that is similar to learning user preferences or a recommendation system
AntiSquid: Scarfield https://averagejoe351213450.files.wordpress.com/2018/07/inuyashiki-header.jpg
reCurse: There's also a fairly basic solution, optimize wrt pair winrates
reCurse: Trueskill is very ill suited for bot ranking
reCurse: Yet everyone does it for some reason
MSmits: Wontonimo there is a rock paper scissors going on, but it's actually much worse in other games
Scarfield: lol squid
MSmits: not doing as poorly against you this time around reCurse
Wontonimo: MSmits all the more reason to promote a solution. If we had tennis players, wouldn't we keep track of (or calculate from probabilities) left and right handed?
reCurse: It's ok it's my best chance at 2 minutes of glory
MSmits: Wontonimo this has been discussed soooo often. there's no point to it now
reCurse: Fighting this system is too frustrating
reCurse: If this community wasn't dying I'd make alternative leaderboards
Astrobytes: There are ways of calculating more reasonably
Astrobytes: Ah yeah, what reCurse said
reCurse: Parse new matches every hour and compute new leaderboard every day
Wontonimo: haha MSmits , just cuz i'm late to the game, doesn't mean I get to miss out on giving my 0.02 . That said, i get it, we got we got
MSmits: i accept that the rankings are flawed and that rank is onyl accurate between two bots with obviously different strengths
MSmits: Wontonimo sure, didn't mean to silence you on this
MSmits: just trying not to give you false hope this will be fixed
Wontonimo: all good dude
Astrobytes: reCurse: The community isn't dying it's just cg seems hell bent on killing it
reCurse: No it's dying let's be honest
Wontonimo: which community?
reCurse: Not nearly as much activity from the 5% as there used to be
Astrobytes: Yes but the reason is
AntiSquid: he's talking about us Wontonimo
Wontonimo: :(
AntiSquid: i consider myself still young though
Astrobytes: euler's inactivity made me quite sad
Scarfield: ^
reCurse: I was surprised he sticked around so long after
Wontonimo: i feel like i missed out by not joining 2 years ago
reCurse: I don't nearly put as much in as he does
reCurse: I'd be frankly disgusted
MSmits: Wontonimo try not to get too discouraged. There is no better platform for botting out there still... afaik
Astrobytes: Yeah you're not wrong there reCurse
AntiSquid: what happened to euler? wasn't around much lately
Astrobytes: It's beyond a slap in the face/kick in the teeth
MSmits: he designed a contest
reCurse: Think he got fed up being played like a violin for months trying to get his community contest in
MSmits: but never happened
reCurse: Only to get the abysmal response CG gave
AntiSquid: aaaah that, i was surprised they wouldn't accept a ready made game to fill in an empty downtime
reCurse: Gotta get those juicy numbers twice a year
Astrobytes: I mean that was ready since last end of summer/beginning of autumn (fall you heathens) right?
Astrobytes: TryAngle Catch I mean
AntiSquid: don't think it helps retention, but i don't have all the stats
reCurse: I think the clashers are the audience they're going for
reCurse: Well soon they'll get what they want
Astrobytes: Yep.
Wontonimo: honestly, i thought this was mostly a job recruitment site that just tolerated the top % players who are just doing it for fun and not recruitment. So personally, I'm super impressed that they pay for all this cloud compute that doesn't really feed their core business
reCurse: I am very thankful for the years of fun
reCurse: I am very frustrated with how it ends
reCurse: End of story
Astrobytes: It's kinda built on the community, or has been these days, and the reason we've supported them is because they've done all this.
Andriamanitra: i'm sure they would like this to be a recruitment site because that would make money
Andriamanitra: but i'm not convinced anyone is getting recruited here :D
Astrobytes: Now, it's a different feeling.
AntiSquid: are you looking for a job here Wontonimo ?
Wontonimo: hey hey ... reCurse did you change your bot or just resubmit? if change, mind sharing?
reCurse: They have a very different side of business we don't see
reCurse: Same bot
reCurse: When I complain about the broken rankings
Astrobytes: CG for Work is separate
reCurse: I'm not fucking kidding
Wontonimo: AntiSquid no, i'm happy with my job
reCurse: They are broken as hell
Astrobytes: that's the same one reCurse?
reCurse: Yes
MSmits: well... not everyone feels the same way about this. I do think it is sad that they removed community contests, but it wasnt the main fun for me
reCurse: Exact code
Astrobytes: ffs lol
MSmits: well
MSmits: karlisos bot does very well against some subtop players
MSmits: better than recurse
MSmits: so when they submit...
reCurse: There is no excuse
AntiSquid: contests are the most active time
Wontonimo: HAHA
reCurse: I have 65-70% vs him, I don't care if he 100% the rest
MSmits: i had a bot for a long time that had 25% vs karliso and 50% vs you
Astrobytes: Perhaps they never really factored in the fact that people would write such competitive bots - or perhaps they just never bothered updating it accordingly :thinking:
MSmits: reCurse thats not fair though, you're saying you want RPS to make you win, but not when ti makes you lose
reCurse: Ok let me rephrase
AntiSquid: even OOC was fun, didn't like the game itself, but it's the event and how it plays out and the chatting and ranting and debugging and complaints and the memes ....
reCurse: There is no world in which you have >65% against all top bots
reCurse: And you're not rank 1
Astrobytes: No he's saying it's nonsense
reCurse: There is no justification
Scarfield: and ceg asking about the rules
MSmits: i can give 1, sort of
MSmits: what if there's only 1 clear rival
reCurse: I can fit an elo curve on winrates if you give me 24 hours to batch it
MSmits: and having >50 % against everyone else is trivial
reCurse: It's a joke
MSmits: you migth RPS that rival
AntiSquid: ceg "speed runs", save time on reading statement Scarfield
MSmits: does that mean your bot is better?
reCurse: If you have a significant positive winrate against every bot
reCurse: You are number 1
reCurse: End of story
reCurse: I can document all the ways the system is broken
AntiSquid: :thumbsup:
MSmits: allright, so what if they did a round of games where all the top 10 playeed 10 games against the other 9
MSmits: do you think you would win that?
MSmits: if all games counted equal
MSmits: because I doubt that
reCurse: Low sampling means there is a probability distribution
Astrobytes: There was a reason for pb's alternative scoring in BR (and in subsequent contests)
MSmits: ok 1000 games then
reCurse: So let's say I have 70% vs the rest
reCurse: And he 100% the rest
Astrobytes: *(and grpahs in subsequent contests)
reCurse: But I 70% him
reCurse: You're fine saying I'm number 2
MSmits: yeah
struct: yes
MSmits: because RPS
MSmits: what if he changes 1 parameter
reCurse: I don't know what to say. It's pure nonsense.
MSmits: and your 70% becomes 40%
MSmits: maybe thats the case
MSmits: RPs is fickle
reCurse: What kind of hypothetical scenario is that
MSmits: seen it happen a lot
reCurse: It's not about what could be
reCurse: It's about what the winrates *are*
struct: But what would your win rate vs the rest be?
reCurse: I just explained
MSmits: yes but you are making your own definition
reCurse: I just exaggerated in his favor the current scenario happening
reCurse: And you still think it's fine
MSmits: I know
reCurse: So it's kind of mindboggling to me
MSmits: you are choosing a particular definition of *best bot*
MSmits: i am not defending the trueskill system btw
reCurse: Someone saying a bot that wins significantly above everyone else doesn't get rank 1
MSmits: i am just not certain its fair to say karlisos bot is weaker overall
reCurse: To me it's like 1+1=3
reCurse: I don't know how and why I should debate that
struct: Which rating system would not place you second in the example you provided?
reCurse: An elo curve fitted against pairwise winrates
Astrobytes: (which is also, I may add, asking the impossible on CG)
MSmits: but you yourself said it's important that a bot is generalized well.
Wontonimo: trueskill has both win chance and fickleness of win built in. With these two factors all sorts of weirdness come out.
MSmits: isn't karlisos bot generalized better if he does well against everione except your deterministic nn
reCurse: You think a better generalization is better winrates against bots you dominate while losing significantly to a rival bot
MSmits: no, i think it needs to all be taken into account
MSmits: and i think RPS is a bad judge of bot strength
reCurse: It's not RPS
MSmits: why not, i tried like 5 completely different bots against karliso
MSmits: i even have had 55% winrate vs you
AntiSquid: well continuous games would fix this even at large intervals
MSmits: with one of them
reCurse: You don't need continuous games
MSmits: none of them got over 30% vs karliso
reCurse: You just need to fit historical data
Wontonimo: once i found the score was trueskill, i change my strategy to try and target more consistent results and really make sure I don't lose against much lower (and also it can hurt to win against much higher)
Wontonimo: i don't think that's the case with your bot reCurse cuz it just wins as far as i can tell ... so your uncertainly should be near zero
MSmits: http://cgstats.magusgeek.com/app/multi-tic-tac-toe/reCurse
MSmits: looks very much rps to me
MSmits: look at that insane winrate vs karliso
reCurse: Can we please not look at submit results
reCurse: And instead rely on big batches
jacek: and little vs jacek
reCurse: I've had 75 games 55% winrate batches turn into 68% at 250 games
MSmits: it's small sure... but still... 30-6
Astrobytes: Wontonimo: It's a bit sad you feel you have to actually change your approach purely based on the scoring algorithm
MSmits: yeah i would not do that
reCurse: I don't have 83% on a large batch
reCurse: So I won't debate that
AntiSquid: i'd expect to be 1st with win rates like that tbh
MSmits: no i know, but you do not have much difference between winrates vs karliso and other players
MSmits: right/
MSmits: ?
MSmits: he;s sooo much stronger than everyone else. Only your bot has this experience
MSmits: which is quite the achievement to be sure
MSmits: not debating that
reCurse: You're assuming RPS when it could be unexplored transitivity
MSmits: what that?
Astrobytes: Transitivity that has not currently been explored.
reCurse: The opposite of RPS is perfect transitivity right?
AntiSquid: no i mean recurz got better win rates on that magugeek site, or am i missing something?
MSmits: sure ok
reCurse: But all you have is few data points at the top with a very large distance
MSmits: yeah thats the main problem, the two guys at the top are far ahead
reCurse: With more diversified top bots you could still find it's not RPS at all
MSmits: exactly
MSmits: if we had more bots of that strength, we could know for sure
reCurse: So I don't know why you assume RPS when there's no strong evidence for it
Astrobytes: Do some alt submits
MSmits: I don';t assume, i think it;s possible, even likely because of my experience with various bots submitting. But I cant assume
AntiSquid: oh ya flood UTTT with clones ! :D
Astrobytes: No, not clones. I mean meaningfully different but strong bots.
MSmits: various bots benchmarking i mean
Astrobytes: "Multiplayer is a zoo" right
AntiSquid: right, but some animals live in large colonies
Astrobytes: don't get OT
Astrobytes: Nah I can't make head nor tail of that UTTT leaderboard really.
Just_Spark: I can't understand clashincode game xD
AntiSquid: well then ... that's end of the road
AntiSquid: omega_xD ?
PatrickMcGinnisII: lol pushed papyjo out of wood2 connect4 with a depth 1
Astrobytes: You're supposed to push yourself out of wood 2
PatrickMcGinnisII: its deoth 1
Wontonimo: i think i may just start a metagame of trying to make a bot good enough to push players out of wood league in games but purposefully lose agains the boss
PatrickMcGinnisII: rofl
PatrickMcGinnisII: i win against boss, but not players
PatrickMcGinnisII: boss is pretty dumb
struct: If you do it in CSB, its similar to DDOS
Wontonimo: eventually they'll get auto-promoted and only leave you and the bot
MSmits: PatrickMcGinnisII the boss is supposed to be a weak bar of entry
Astrobytes: lol struct
Astrobytes: anyway, goodnight/goodbye all, laters
struct: gn
MSmits: gn Astrobytes
reCurse: :wave:
Wontonimo: laters Astrobytes
AntiSquid: gn8
Wontonimo: it's only 6pm dude (timezone lol)
MSmits: midnight here :)
MSmits: In Dutchieland
Scarfield: gn
Wontonimo: so, 90% away approximately
AntiSquid: are you leaving too Scarfield ?
MSmits: 90% ?
struct: Just make unbeatable bot, problem solved
Scarfield: nono, just a little late to say gn to sleepstrobytes
Wontonimo: if you tell me how high the moon rises off the horizon then I can get a rough estimate of long and lat from my position
Wontonimo: actually, no, i'm too lazy the to the math
Wontonimo: 90 degrees
Wontonimo: not %
Wontonimo: there really should be a degree symbol on the keyboard
MSmits: yeah i dont have it either
AntiSquid: there is, you just have wrong keyboard / layout
AntiSquid: not on mine either btw
Wontonimo: (waves fist)
milad-mehri: °
milad-mehri: there is no key
milad-mehri: use alt codes
Wontonimo: or spoon apparently
AntiSquid: you have it on german keyboards https://en.wikipedia.org/wiki/German_keyboard_layout
AntiSquid: top left
Wontonimo: hmm .. how much damage has all these resubmits caused me on uttt ...
MSmits: your 31 rank would be worth top 20 a year ago and top 10 the year before that
MSmits: uttt has evolved a lot
Wontonimo: that's what i gets for being late to the party
MSmits: nagrarok was nr 1 for over a year i think
MSmits: now 16
MSmits: same bot
reCurse: Booking doesn't age well
Wontonimo: i also get some lessons learnt from trail blazers and community support
MSmits: I do you really think he booked so much? I never noticed it
reCurse: Not nearly as much as those coming after
MSmits: maybe because he didnt counterbook me personally
reCurse: I meant booking, counterbooking is more specific and harder
reCurse: To detect
MSmits: ah
MSmits: well booking doesnt even help that much
MSmits: regular books i mean in uttt
reCurse: I'm speculating btw, haven't had official confirmation/deny
MSmits: ahh ok
MSmits: i noticed the counterbooks when i pitted you, karliso and miarem against eachother in my IDE and saw the same games for 15-20 plies. Identical every time :P
MSmits: of course could be other reasons... but not likely :0
reCurse: True
Wontonimo: that just seems kinda cheap ... and tempting
MSmits: tell me about it
Wontonimo: not my speed though
MSmits: they created a monster by doing that :P
Wontonimo: i'd get a lot more satisfaction from having an algo that wins vs it than just hardcode something
MSmits: its no longer done much in uttt now though
MSmits: dont worry about it
MSmits: the thing is, counterbooking makes you vulnerable to counterbooking
MSmits: by counterbooking, you're making yourself deterministic
MSmits: it's selfdefeating, unless it's perfect play
Wontonimo: if i was gonna attack such a thing, my kneejerk would be to deep train offline for that config, and apply it as a soft heuristic like i do with teccles heuristic. It isn't deterministic in my algo.
Wontonimo: but ... now thinking of it, i can't see how it wouldn't be deterministic and open to counter attack
Wontonimo: okay, point taken
MSmits: if you apply teccles only in the early game it's pretty much perfect play
MSmits: i cant prove that, but i've seen enough to think thats the case
Scarfield: just dont google it, apparently :p
MSmits: lol
Wontonimo: at least from my basement bathroom
reCurse: Counterbooking doesn't necessarily imply deterministic
reCurse: You could make it next level with generic counterbooking
MSmits: that's true i guess
Wontonimo: hmm ... this refactoring to not have the whole board copied in each node is proving time consuing
MSmits: worth it though
MSmits: you will have more space
MSmits: can even do tree reuse then
MSmits: i am assuming you currently dont reuse
MSmits: with those huge nodes
Wontonimo: no reduce / reuse / recycle here
struct: Also storing the current game state every time can be expensive in the long run
struct: Especially when you dont need it
MSmits: what most of us do is have 20-30 million node pool, when you apply a move, you select the child as the new root, then when the opponent plays a move, you search that new root for the move and make the new child the root. then you continue searching
MSmits: the new turn will have a large tree as a base already
MSmits: it's probably about 10% of the original tree. Still helps
MSmits: with a small exploration, it could be much more or less, more if your opponent plays what you expect
Wontonimo: oh, i do reuse the tree from the last turn and search for the opponent move
MSmits: ohh ok
MSmits: but then you probably run out of nodes?
MSmits: or maybe not if you dont have that many rollouts
Wontonimo: i don't pre-allocate nodes. I tried that yesterday after talking and got the segmentation violation blues ... with small nodes it'll be okay
Valdez001: hi
MSmits: yeah it will be
MSmits: preallocating nodes will help a ton
struct: you create nodes every expansion?
MSmits: also, why not reset your tree when you run out?
PatrickMcGinnisII: wood1, puke\
MSmits: just check at the beginning of your turn if you're near the end of your pool and reset it
Wontonimo: also, i tried to allocate on the stack. ... which is the wrong place. so, that's an easy fix
MSmits: you'll lose the tree, but thats fine
Wontonimo: struct yes i create nodes as i need them
struct: I wonder how big the performance penalty is
Wontonimo: dinner calls! later
MSmits: gtg as well. GN!
struct: gn
Wontonimo: oh, and my nodes contain a whole board each
Wontonimo: atm
Scarfield: gn gn
Wendys: yes
Wendys: hello mr guillermo
Guillermo.: who are you mr wendys
Wendys: yes
Just_Spark: I'M DONE WITH THIS GAME 'Power of Thor'
Wontonimo: and by done, do you mean you KILLED IT?
middleverse: am i doing something wrong, it's printing nothing
middleverse: am i suppose to return
middleverse: or am i suppose to print()
Wontonimo: pint
Wontonimo: print
Wontonimo: (after the pint)
Wontonimo: what lang are you using middleverse ?
Wontonimo: you can also print to standard error using print("debug",file=sys.stderr,flush=True)
Wontonimo: in python. The standard error will be displayed in the output but won't be considered part of your solution. it's really handy for debugging
YodaLightsabr: why not do something more challenging
YodaLightsabr: like JavaScript?
Notgoodatcoding5: Hi
YodaLightsabr: python has a lot of built in functions
Andriamanitra: as if javascript didn't have a lot of built-in functions :p
likki1811: https://www.codingame.com/clashofcode/clash/16084212331a309ab8be4b56f79753b1893c01d
IaMiNteLLIGENT: I dont think shortest mode is practical becuz 2 launguages could be as efficient as possible and one of them would be shorter than th other, making it unfair... here is a link to an example: https://www.codingame.com/clashofcode/clash/report/160846793ad19de9f55d2d44d3ed52777aef602