Chat:World/2020-03-14

MII_Man: Anyone know if it's possible to change the length of the output text

Uljahn: there is a built-in limit

tutubalin: you can use __int128_t

DenShlk: __int128_t cant be printed

Danyosate: strange.. I passed test cases in ide, but when submit, the same cases raises a failure

Z_Doctor: "The tests provided are similar to the validation tests used to compute the final score but remain different."

Danyosate: welll

Danyosate: thx

Z_Doctor: Could be a potential flaw that isn't triggered in the test data

Z_Doctor: One time for me it was because I was creating a bunch of lists

Z_Doctor: and that took too long

Anton4ik200: доброе утро

Z_Doctor: and caused the test to time out

Danyosate: :sweat_smile:

Hobogodnayamish: от нас все в мир чат сбежали.....

Z_Doctor: @Danyosate What test is it anyway?

Danyosate: Disordered First Contact

Danyosate: it's something wrong in my code, fixed it now~

Z_Doctor: was it related to lists? :smirk:

Danyosate: nope, it's something wrong with my algorithm

Z_Doctor: :thumbsup_tone2:

Danyosate: btw I don't know

Danyosate: the maximum recursion litmit

ZAKVAZ: pleas answer chuk norris pleas\

Z_Doctor: I'd say use them sparingly

Danyosate: are they same when compiling in IDS and after submition?

Danyosate: submission

Z_Doctor: No idea

Danyosate: hope the test in submission could give some feedback

Danyosate: hey! will anyone see if I upvote my own code lol:joy:

Z_Doctor: No one will know it was you

Z_Doctor: Except whoever reads this

Danyosate: from now my code will get at least 1 upvote

Danyosate: c'mon, i bet most of the codes with 1 vote is self voting :sunglasses:

Anton4ik200: What?

Uljahn: Danyosate: each upvote is 5 ex worth :smirk:

Danyosate: including self upvote?

Hobogodnayamish: #ОРДЕНМЯЧИКА

Uljahn: Danyosate yep

Danyosate: now i'm going to upvote all my codes

Danyosate: well and urs. sometimes I saw ur code

Uljahn: also you can deactivate your upvote afterwards and exp would still stay

Danyosate: :sweat_smile:

MSmits: morning

MrTermik33: Hi

dbdr: hi!

eulerscheZahl: hi

eulerscheZahl: i think i'm missing something fundamental at bulls and cows

eulerscheZahl: 100 runs offline, best was 307 :/

eulerscheZahl: averages

eulerscheZahl: 321.13 2: 5.06 3: 5.038 4: 5.298 5: 5.738 6: 6.42 7: 7.174 8: 8.254 9: 9.552 10: 10.692

MSmits: hey Euler. I took some cp from Royale while keeping your the same :P

eulerscheZahl: which game?

MSmits: in langtons ant

eulerscheZahl: cool

MSmits: my bot started working

dbdr: eulerscheZahl: how do you select your guess? random?

MSmits: but redstrike is insane... what does he do ??

eulerscheZahl: oh, you are 2nd now :o

eulerscheZahl: no idea

eulerscheZahl: depends on the amount of candidates

dbdr: for MSmits, "started working" means #2 :D

eulerscheZahl: very few: all remaining + testing random excluded candidates

MSmits: doesn usually happen this fast. I dont think people spent many days on langtons ant

eulerscheZahl: for more candidates remaining i test all of i can

eulerscheZahl: or some random candidates among the possible ones

dbdr: what do you mean test all?

eulerscheZahl: trying to minimize the largest group of possible new candidates

dbdr: ah ok

eulerscheZahl: i have a list of possible numbers, which shrinks

eulerscheZahl: i try to test all of these at least

eulerscheZahl: possibly some more, that i already excluded

dbdr: sounds to me like you are trying to optimize the worst case

dbdr: maybe try to optimize the average case instead

eulerscheZahl: easy to change...

MSmits: or optimize the yogi-case

dbdr: yogi-case?

dbdr: what ref am I missing? :D

MSmits: he's better than the average bear

MSmits: since you're going to be submit spamming :)

dbdr: is there a reference?

eulerscheZahl: batch running, let's see if that's better

eulerscheZahl: oh, a 302 already

MSmits: https://idioms.thefreedictionary.com/than+your+average+bear

eulerscheZahl: best before was 307

dbdr: :)

dbdr: ah, that's yogi bear

eulerscheZahl: btw by secrets are fully random

dbdr: didn't know that name, although the pic looks familiar

MSmits: ahh ok

eulerscheZahl: no fixed number of bulls/cows like for the online version

MSmits: does that matter euler?

dbdr: it probably affects the average

eulerscheZahl: i think that shifts the expected score a bit

dbdr: I also don't use them

dbdr: you use 5 for N=1 right?

eulerscheZahl: yes

eulerscheZahl: int score = 5; for (int digits = 2; digits <= 10; digits++) { for (int run = 0; run < 5; run++) {

dbdr: yeah

MSmits: why only 5 runs?

eulerscheZahl: for (int run = 1; run <= 100; run++) { int turn = batchRun();

MSmits: just a quick test?

dbdr: there are 1 + 9 * 5 validators

eulerscheZahl: batchRun() doing the 5 times for each length

MSmits: oh right

MSmits: like the validators

MSmits: you guys are going all NS on this cattle game

dbdr: BNC is the new NS :D

MSmits: apparently

dbdr: guys, I've been thinking about a place for the community to share and store information

eulerscheZahl: forum.codingame.com?

eulerscheZahl: or a wiki?

dbdr: played with a few tools yesterday, and I like mediawiki (same software as wikipedia)

dbdr: yeah, it's complementary. forum is more of a feed, wiki can be organized

eulerscheZahl: i don't know that one docuwiki is easy to set up i think

MSmits: like working tricks for various multis?

dbdr: https://cg.spdns.eu/wiki/Main_Page

dbdr: yes, for instance

dbdr: can link to the forum thread of course

jacek: BNC?

eulerscheZahl: bulls and cows

dbdr: or whatever people find useful

jacek: also, happy caturday

MSmits: would this replace the tech.io stuff people sometimes write?

dbdr: no

dbdr: I think tech.io is when you spend more time to write one article, no?

dbdr: on a wiki people can make a quick edit

MSmits: some people do that for wiki

dbdr: sure

dbdr: but you can't edit tech.io if it's not yours, right? probably you can comment

MSmits: thats true

MSmits: i think it's a good idea

dbdr: MSmits you said something about listing puzzles that help learning, right?

dbdr: bot writing I mean

MSmits: but these things need to achieve critical mass, before they start working, I hope that happens

MSmits: yes

dbdr: where would that go?

eulerscheZahl: average of 100 runs: from 321.13 to 316.2 with average instead of worst case

dbdr: nice eulerscheZahl

MSmits: if the wiki didnt exist, it would go on tech.io, but as you said, then 1 person would be responsible

eulerscheZahl: could help newbies but sounds like a lot of work :scream:

dbdr: yeah, tech.io is definitely great for some stuff, I don't intend to replace it, it has unique features

dbdr: well, the idea if there is something that comes up, just put it in the wiki

MSmits: having just a progression tree for puzzles toward bot games is something that would fit great on a wiki

dbdr: no need to spend hours on it

dbdr: for wikis, the philosophy is "don't hesitate to edit, don't sweat it, it will be improved later if needed"

MSmits: before i did my first multi's and contest I did a lot of puzzles, looking for ones that practices graphs and pathfinding. But there were only like 50 community puzzles then I think

MSmits: so it was not too hard

eulerscheZahl: now we have a tagging system on CG

dbf: you can also keep it on github in markdown - no hosting/etc is required in this case, but updates are a bit slower (when PR is merged)

eulerscheZahl: i know that not all puzzles use it correctly, but could be edited here

MSmits: sure, but newbies dont know the tags they need I think

dbf: we have one such article: https://github.com/core2duo/RHC-AI/

dbdr: dbf: interesting idea. but then you need to give commit access to it, right?

eulerscheZahl: oh dear, it's all Russian

dbf: eulerscheZahl, I think it should be fine for you :)

MSmits: you probably read some, after all that RAIC :P

dbdr: 7 место (Magus) :D

eulerscheZahl: and the RAIC contests i really tried are missing

dbf: dbdr - you can give commit permissions to several trusted (top 10 in CG rating, for example :)) persons and others will just submit pull-requests

eulerscheZahl: mini raic 4 and the last big one

dbf: eulerscheZahl, they are in my PR, but it is not merged yet, lol: https://github.com/core2duo/RHC-AI/pull/14

dbdr: dbf: that sounds fine for very curated content. but it does not encourage everyone to contribute. so it's a different approach

eulerscheZahl: you forgot https://github.com/eulerscheZahl/RussianAICup

eulerscheZahl: and i agree, a wiki is more dynamic

eulerscheZahl: there are 2 or 3 contests a year

eulerscheZahl: while dbdr wants more changes

dbf: exactly, github is slower, as I mentioned but the good thing is that you don't need to host it and a lot of pepole are already registered there

dbdr: sure, different use case

dbdr: you don't even need to register on a wiki if you don't want to

dbf: eulerscheZahl, it is a list of artlicels mostly (not links to repos), so I will be able to add a link only after you write an article (in russian, if possible)

dbdr: step 1: learn russian

eulerscheZahl: meh, not that relevant

eulerscheZahl: cyka blat

eulerscheZahl: where is my medal?

dbdr: 🥇

dbf: there was some hosting for wikis - there was a nice one about chess programming, but I think it is closed already :/

dbdr: I think there's a backup of it now

dbf: yes

dbf: https://www.chessprogramming.org/Main_Page

dbdr: anyways, I made the install on my host, so we can play with it already :)

dbdr: initially I wanted just to have chat logs searchable, and I realized if I put it in a wiki, I get search for free :)

eulerscheZahl: :D

dbdr: strangely I did not find a simpler solution

eulerscheZahl: i lost the chat of last night, MK spamming with Automaton

dbdr: damn

dbdr: I'll try to restore it for you ;)

dbdr: would be sad to lose that

eulerscheZahl: i don't know if it would

eulerscheZahl: as i don't know what i missed

dbdr: but seriously, we share some pretty useful information on the chat quite often

dbdr: so it could be quite useful to make it searchable

dbdr: among a lot of noise, obviously :D

AntiSquid: chat log with filters :thinking:

dbdr: what kind of filters?

AntiSquid: filter by content

AntiSquid: and maybe users

AntiSquid: use sentiment analysis :D

dbdr: filter by content is text search, no?

dbdr: by user would be useful, yes. should be doable

dbdr: sentiment analysis is future work. anything is possible :)

AntiSquid: yes you'd look for keywords

MSmits: mmh I would have to be more careful what I say here if my students can read it back :P

dbdr: well, they can already :)

MSmits: well, up to a point

MSmits: it disappears after a spam session doesnt it

AntiSquid: you mean like that night out you had which you wouldn't tell the students about? or is it the morning after?

MSmits: lol

dbdr: anyone can use pidgin and log everything

MSmits: sure, but requires planning right

MSmits: and effort

eulerscheZahl: and staying connected, i turn off my PC at night

dbdr: noob

eulerscheZahl: i sleep in that room

dbdr: get a server :)

AntiSquid: carcassonne got declined?

eulerscheZahl: auto-declined i guess

eulerscheZahl: after 30 days

dbdr: https://cg.spdns.eu/wiki/Chat:Fr/2020-03-13

dbdr: might need to improve the formatting :D

eulerscheZahl: you don't like linebreaks, do you?

darkhorse64: Inputs and graphics IP were unclear

dbdr: I put \n, but that's not enough for wiki

Uljahn: AntiSquid: wlesavo is thinking of tiles remake and providing some starter bots

kovi: so many bugs

struct: Hello

dbdr: what bugs kovi?

kovi: in my bandc speedup

kovi: it basicly became full random (and sometime crash)

eulerscheZahl: can we agree on B&C?

dbdr: B'n'C

kovi: mine is a single token ;)

darkhorse64: Bacteriological Nuclear Chemical bot ?

eulerscheZahl: bogus and non-deterministic challenge

kovi: :)

dbdr: yay, avatars

dbdr: https://cg.spdns.eu/wiki/Chat:Fr/2020-03-13

dbdr: who needs line breaks when you got avatars?

dbdr: kovi 304, nice!

kovi: thx

darkhorse64: Still unreadable but beautiful. The drawback is now that every insanity, every typo will get recorded forever

eulerscheZahl: i see your history goes back a few days

dbdr: yeah, wrote my xmpp bot last week

dbdr: and of course first time I mass import I mess up :/

eulerscheZahl: when i change my profile pic and CG cleans up their servers, your logs will break to some degree :thinking:

AntiSquid: it's not christmas anymore

eulerscheZahl: it's not?

dbdr: hum, the avatar links to a specific image?

dbdr: the XMPP contains this:

dbdr: <vCard xmlns='vcard-temp'><X-USERID>1500515</X-USERID><X-PUBLICHANDLE>8374201b6f1d19eb99d61c80351465b65150051</X-PUBLICHANDLE><NICKNAME>eulerscheZahl</NICKNAME><X-AVATAR>37132928861372</X-AVATAR></vCard>

eulerscheZahl: your representation i think disappeared now, so i can't analyze it in detail

eulerscheZahl: you deleted it?

dbdr: I'm deleting because I imported wrong

dbdr: I'll put it back

dbdr: but with the info above from XMPP

dbdr: does it become stale if you change your avatar?

eulerscheZahl: no

eulerscheZahl: 37132928861372 is to that very specific xmas toad

struct: What if you delete the account?

eulerscheZahl: i think i already broke the default vindinium battle, not finding my old avatar anymore

dbdr: right

eulerscheZahl: what if you try that for us struct?

dbdr: I can always find the latest avatar for the given userid

dbdr: I store the XMPP as ground truth, the rest I can always regenerate

eulerscheZahl: true

struct: Maybe I can

dbdr: or would there be a way to "the current avatar for user X" ?

struct: I'm doing stuff outside CG atm

struct: Decided to learn asm

eulerscheZahl: not in HTML i think

eulerscheZahl: you will need an API call to get user id -> avatar

dbdr: and that would be slow as hell

eulerscheZahl: and spamming servers possibly

dbdr: unless cache

dbdr: but not worth it I think

dbdr: I could cache the avatars locally

dbdr: and symchronize them when I add a log

dbdr: ok, chat history is back

dbdr: and search works: https://cg.spdns.eu/index.php?title=Special:Search&profile=all&search=bitboard&fulltext=1

struct: nice

dbdr: these avatar links do not look nice in the search results, but oh well...

eulerscheZahl: still highly appreciated

struct: It's fine like it is

struct: Imo

jang888: hi

jang888: i am jang888

dbdr: every DOS attack on 2.15.156.27 ;)

YurkovAS: dbdr chat history - is great!

dbdr: cool, thanks!

dbdr: I just made it read-only, so people cannot change history anymore :)

dbdr: a contributor, nice :)

eulerscheZahl: french IP

JFB: dbdr - language option for automatic translation ft -> en ?

JFB: *fr -> en

dbdr: https://translate.google.com/translate?sl=fr&tl=en&u=https%3A%2F%2Fcg.spdns.eu%2Fwiki%2FChat%3AFr%2F2020-03-11

eulerscheZahl: i think the suggestion was to set the language of the page itself in the metadata so the browser suggests to translate the page

dbdr: the google translation is definitely entertaining

eulerscheZahl: chat = cat you are real pet lovers over there

JFB: eulerscheZahl - exacly - but dbdr solution is nice

dbdr: it's chaturday

dbdr: ah

dbdr: how do I do that?

JFB: using backspace ?

eulerscheZahl: for wiki: no idea, as you don't edit plain HTML

dbdr: well, I have a Chat:Fr/ namespace, so it might be possible

dbdr: backspace??

JFB: chat + backspace = cat

Uljahn: dbdr: could we make another page (or template) to represent the same history in different ways like with no avatars (or smaller ones) or with a bold font for nicks? just having no idea of wiki engine abilities

dbdr: definitely the display is not ideal

dbdr: it's a quick POC

dbdr: I like how discord and webchat do it, not repeating the nick for consecutive lines

JFB: chat (fr) -> cat (en) -

AntiSquid: chat + backspace = cha

JFB: true

JFB: simply 'cat' mean 'cat' in france co translator converted enlish 'chat' word to 'cat' thinking that it is france

JFB: :-)

JFB: I chust make 'ch <backspace> at' mistake :-)

JFB: *just

JFB: dbdr - any way it is great idea - I can not simultanously program and read this chat and it is a lot of great ideas on the chat now I can do both (program and from time to time looks for new ideas in the chat history)

Astrobytes: Wow, great work dbdr, thanks for that

dbdr: :)

MSmits: you even made the automatons make more snese

MSmits: sense

MSmits: does maddy speak French !>

MSmits: ?

dbdr: he does!

MSmits: damn

dbdr: pretty good too

eulerscheZahl: self-taught afaik

MSmits: so basically, we can now spy on the French during contests

MSmits: get some more non-Frenchies in the top 10 :)

MSmits: yeah i remember, maddy has a thing for languages

dbdr: google mistranlation -> spend hours debugging your bot :D

MadKnight: yea MSmits u can always google translate it

MSmits: yes but dbdr made it so easy now

dbdr: best is google translation of the french Automatons

MSmits: are they better trained?

dbdr: I think they are similar, could be wrong

MSmits: or do the French make more sense to machine when they speak

dbdr: the translation adds a little spice :)

MSmits: yeah

MSmits: WhatTrickeryIsThis: and the groundhog puts the chocolate in the aluminum

darkhorse64: The last one who said it got a wind on #world

MSmits: I think it's not just French. They speak in code to confuse us !

eulerscheZahl: do you speak any languages apart from dutch and english?

MSmits: I can probably read German reasonably well

MSmits: and some easy sentences in French

MSmits: but not speak

Astrobytes: I can understand Rammstein songs

eulerscheZahl: i'm not even sure if i can

Astrobytes: :D

MSmits: should those really be called songs?

eulerscheZahl: i always struggle to understand lyrics, no matter if german or english

Astrobytes: MSmits Rammstein are great

Astrobytes: yes but you're averse to music euler

MSmits: mmh I prefer other music

struct: I struggle so much to learn by reading books

MadKnight: what kind of music MSmits ?

struct: I looose focus quickly

MadKnight: hey struct

struct: Hi MadKnight

MadKnight: have u found any NN tech io tuto ?

MSmits: I'll pm you MadKnight

struct: There isn't

Astrobytes: lol MSmits

MSmits: better not to share it publicly, or it triggers Astrobytes :P

Astrobytes: Absolutely

Astrobytes: Away and rub your magic lamp

MSmits: lol

struct: I'm waiting for you tuto MadKnight

darkhorse64: I love french -> english translation

JBM: what's your rate?

darkhorse64: I mean the one from google

JBM: cheap

Astrobytes: Did I tell you about the time I was searching on Bing for petit farci niçois, and it translated as "The Stuffed Babies Of Nice"

JBM: ever seen that viral image of petite bites?

Astrobytes: Not sure I want to lol

JBM: oh the image is as sfw as it gets

JBM: googling for it seems to be tough though

JBM: in .fr at least

JBM: well anyway this ought to do the job: https://www.reddit.com/r/CrappyDesign/comments/4jwfrl/petite_bite_means_little_dick_in_french/

Astrobytes: We used to serve Bitburger beer at a pub I worked in, and our Parisian bartender used to crack up any time anybody used the phrase "bitte ein bit"

Astrobytes: Hah! Classic.

JBM: clbuttic

Astrobytes: lol, indeed

darkhorse64: just read the news about UK health policy. I read it twice but I still cannot believe it ...

JFB: What is it?

JFB: This UK jealth policy?

darkhorse64: Collective immunity. Get as many people sick as possible.

struct: lol

struct: logic

MSmits: link

MSmits: if you have it

MSmits: could be fake news

JFB: probably

darkhorse64: It's in French from one of our respected media and it's not fake

darkhorse64: *most respected*

MSmits: I cant find that anywhere

eulerscheZahl: Astrobytes mentioned something like that yesterday too

eulerscheZahl: thank god Britains are the only Europeans who can still travel to the USA

darkhorse64: and I did not understand because I could not imagine such a policy

struct: It's called herd immunity

JFB: Witch such politicy is only problem of time

eulerscheZahl: witch :D

JFB: and number of hospital

JFB: olicy - get as many people sick as possible

Astrobytes: They appear to be backtracking on some elements such as cancelling major events etc but they still plan to kill a load of us

JFB: *policy

MSmits: Astrobytes where are the official UK guidelines?

MSmits: it really doesnt matter if boris johnson says stupid :poop:

MSmits: trump does too.. no use taking it seriously

Astrobytes: Official guidlines for what, precisely?

MSmits: how to handle the crisis

Astrobytes: *guidelines

Astrobytes: They don't tend to publish those

MSmits: does it say anywhere official that it is their goal to infect people

Astrobytes: As such. But they outlined the policy in a speech the other day

MSmits: a speech by Boris Johnson?

Astrobytes: Yes, they explained that they want to let the virus pass through society

MSmits: thought so

MSmits: Trump contradicts his own specialists daily

MSmits: so I am assume the British version does also

MSmits: am assuming

darkhorse64: 60 % of the population infected, 1% deaths leads to 400.000 people

MSmits: keeping the schools open is not necessarily a measure to "get more people infected". There are good arguments for and against it

eulerscheZahl: johnson seems like trump on a first look, but i think he's smarter and more self reflected, sometimes pretending to be stupid

JFB: Passing throught society may be OK - if yoi have enough resources to care for very serious cases

Astrobytes: No, MSmits, Boris, plus the chief scientific officer, Sir Patrick Vallance and Chief Medical Officer Chris Whitty

eulerscheZahl: trump vs johnson in more detail https://www.youtube.com/watch?v=dXyO_MC9g3k

MSmits: eulerscheZahl true, he is, but even if you are 10x more smart and self reflective than trump, you're comparing to a very low bar

Astrobytes: TRump != Boris

Uljahn: JFB good point

MSmits: I know that

MSmits: i saw that show :)

Pegasus7: hi to everybody

MSmits: always watch John Oliver

eulerscheZahl: same

MSmits: he's the best Brit

JFB: But I'm not sure do GB have resource to care for all serious cases.

Astrobytes: No. We've had 10 years of austerity. Cuts, running down funding for the health service

MadKnight: > TRump != Boris but does Trump == Boris ?

darkhorse64: 5000 slots to reanimate people

JFB: In Poland we have only 10 000 ventilator (for 38 milions people). Not all working

MSmits: i think you mean respirator?

JFB: Yes - respirators

MSmits: 10000 is probably not as bad as it sounds

Astrobytes: Ventilator is correct

Astrobytes: *also

darkhorse64: In france, the figures are not that much higher but growing. Therefore, the government policy is to restrain virus propagation as much as possible in the hope that our resources do not get saturated

MSmits: less than 20% of the people catching the virus actually needs a respirator and many people will not catch it and also they will not need respirators at the same time

Astrobytes: Well, in the UK we have a serious healthcare crisis even without this pandemic so...

MSmits: yeah that makes it alot worse =/

MSmits: hopefully this crisis will change politics to focus more on health care in the future

Astrobytes: Don't be silly.

Astrobytes: :D

MSmits: I'm optimistic :P

Astrobytes: No one ever learns.

MSmits: though, a lot of healthcare problems can be prevented. So much diabetes out there for example

eulerscheZahl: summary of my school time

MSmits: diabetics are vulnerably to corona also

MSmits: vulnerable

Astrobytes: Our healthcare crisis is that the health service lacks funding. They plan to privatise it and sell it off, so they've been effectively asset-stripping it for a long time

MSmits: health care privatisation is bad

FuSheng: hello

FuSheng: oh hey guys youre here!!!

MSmits: where else woulld we be

JFB: In my country is officially to work remotly for all who has such possibility

FuSheng: sorry i left that day, was in a bad mood honestly

MSmits: JFB here too

MadKnight: welcome back FuSheng

FuSheng: hey

Astrobytes: How is your Python FuSheng?

FuSheng: what are you guys up to?

MadKnight: what are u gonna do today ?

FuSheng: oh hey astro

FuSheng: i dont know, ive wasted 10hours since i woken up lol

FuSheng: just mayybe read up on tutorials from ww3 schools

JFB: But our government was similar to Trump in previous weeks - and now try to "start fight" So make stupid moves too. Cancelled all flight - immediatelly.

FuSheng: oh yeah crap my project starts on monday... officially

JFB: So if yo are abroad - you can not return to country

Astrobytes: Most of the European response has been shockingly slow JFB

FuSheng: well im so screwed if i cant finish learning python before next week

MSmits: They should have at least cancelled all events as soon as the first case appeared

MSmits: you can't contain this

MSmits: but slow it down helps a ton

Astrobytes: Trying to balance economic crisis and health crisis... priorities

MSmits: right

Astrobytes: Markets > People

MSmits: for some yeah

FuSheng: oh yeah i wondered why python in codingame is different from the compilers online

FuSheng: like i wanna manually input stuff lol

struct: CG is not for that

FuSheng: http://chat.codingame.com/pastebin/101354e1-138f-4817-a766-7b72c6528256

FuSheng: lol nevermind im so gonna fail this

MSmits: if you fail at it, make sure you know if this because of lack of knowledge of python or because you dont know what the program is supposed to do

FuSheng: uhhhh

FuSheng: first i needa create a function right?

FuSheng: now i know how to get n via n = int(input())

MSmits: no, first you need to know how the program is supposed to solve htis

MSmits: you shouldnt even really be coding yet

MSmits: take a small number, say below 100

FuSheng: so when u input the value u wanna see the prime factors of the value

MSmits: see if you can solve it by hand

MSmits: without any code

MSmits: then think about how you did that

MSmits: do it a couple of times if necessary

MSmits: then turn it into code

FuSheng: solve it by hand?

MSmits: yes, piece of paper, pen

FuSheng: like writing a pseudo code?

MSmits: if you want, but pseudo code is already one step further

MSmits: first you need to mess with it

MSmits: learn about the problem

MSmits: see what works

MSmits: then write it down as pseudo, or directly into python, whatever you want

FuSheng: cant use a while loop for this project..

Astrobytes: Yes but take each step you made in solving by hand, then translate that. The steps you take is your algorithm.

FuSheng: uhhhhhh

MSmits: you're already thinking about loops FuSheng... thats a step too far

MSmits: dont think about code yet

MSmits: just solve it on paper

FuSheng: how do i solve it on paper

MSmits: if you can;t, you cant solve it with code either

Astrobytes: ^

FuSheng: isnt it the same as typing on here

MSmits: no

MSmits: on paper you can draw

struct: By using pen and paper you will find how to solve the problem

MSmits: scribble stuff.. cross it out, chane it

struct: What steps you need to do

MSmits: change

FuSheng: uh

MSmits: the thing is, if a solution on paper works for small numbers (which is all you can do on paper), it will work for larger numbers when you turn it into code

FuSheng: # get n #display prime factors of n # find total sum of prime factors ?

MSmits: allright, say the number 20

MSmits: how would you factorize it with prime numbers?

FuSheng: #1 x 20 # 2 x 10 # 2 x 2 x 5 ?

MSmits: right

MSmits: so what you did was

MSmits: you took the smallest prime number you can divide the number by

MSmits: and used it until you could no longer

MSmits: then you tried the next prime number

MSmits: and so on

MSmits: one division by 2 works

MSmits: another works too (then you get 5)

MSmits: 5 is a prime number, so you're done

MSmits: you need several things

FuSheng: oh

MSmits: you need a function to get the next prime number starting from some number

FuSheng: crap someone with javascript just did it in 321 code size

MSmits: how do you know the code sizE?

FuSheng: oh its shown in his tab

MSmits: oh ok

FuSheng: under console output lol

MSmits: is this a clash of code?

FuSheng: yeah its clash code

MSmits: ohh

struct: If you want to learn python or anything

struct: stay away from CoC

Uljahn: +

FuSheng: i figured out its sort of similar to ww3 projects

MSmits: if you're still struggling with coding, it might be more relaxed to just try easy puzzles

MSmits: take your time

FuSheng: like on ww3projects theres lots of project and example solutions

FuSheng: what puzzles can i work on for basics to practice like functions ?

FuSheng: strings list tuples sets

Uljahn: classic easy puzzles with corresponding tags

FuSheng: ok cool let me see

FuSheng: oh yeah @smith, i never thought of getting the next prime number in a function

FuSheng: i think i lack in logical thinking xDDD

MSmits: it's something you can improve a lot with practice

FuSheng: really ?

MSmits: yeah, logic follows patterns

MSmits: you are already very good at pattern recognition. You just did not learn to do it consciously

FuSheng: oh

FuSheng: wait how do you know that

MSmits: cuz every animal is

FuSheng: you should test me on one and see if i can do it xDD

MSmits: especially humans

PatrickMcGinnisII: hey peepz

MSmits: except PatrickMcGinnisII

MSmits: :P

MSmits: hi

FuSheng: oh yeah is it me or i find a 14inch screen kinda small to code on codingame

eulerscheZahl: it is small

MSmits: bigger is better, but i code on a small laptop sometimes too

eulerscheZahl: barbaric

FuSheng: ok so crap my weekends were supposed to be for study

FuSheng: or not actually

MSmits: well i can't take my desktop with me in the train :P

Uljahn: rip eyes

AntiSquid: when on holiday i sometimes use my phone and USB keyboard

FuSheng: ok im on a puzzle

eulerscheZahl: if it's chuck norris, try another one

FuSheng: its called ANEO sponsored Puzzple

eulerscheZahl: isn't that in medium?

MSmits: that sounds hard

MSmits: do something easy

Uljahn: you can't have community solutions for it

FuSheng: sad i dont see any functions catagory in the puzzles

eulerscheZahl: https://www.codingame.com/training/easy go there

Uljahn: functions can be used everywhere

eulerscheZahl: try the puzzles without "by <username>"

FuSheng: i see the loops puzzple

FuSheng: conditions , arrays, and more loops

eulerscheZahl: a good place to start for you

FuSheng: wheres the puzzle that you mentioned @euler

Uljahn: try thor, temperatures, ascii art, mime type if you haven't solved them already

eulerscheZahl: and descent, another really simple one

FuSheng: ok uijahn

FuSheng: how long on average for starters to finish testcase 1?

FuSheng: 15?

Uljahn: wut

FuSheng: 15minutes?

MSmits: FuSheng really hard to say

MSmits: I got stuck a long time on that puzzle somehow. It isn't hard, but i still got stuck

MSmits: you need a loop, get the tallest mountain, output it

FuSheng: oh im on the ascii art puzzle

dbdr: wth MSmits?

MSmits: ah ok

FuSheng: i remember the horror of being stuck on the mountain

FuSheng: remember the other day

MSmits: dbdr... this was 2 years ago when i couldnt do much :P

dbdr: long time to do a very easy puzzle, but top 1-3 on many multis :D

MSmits: i got confused by the double loop

MSmits: the first while loop being the input loop

dbdr: well, if that trend continues, we can all go home ;)

eulerscheZahl: i didn't enjoy ascii art and chuck norris

MSmits: :)

eulerscheZahl: not really stuck but unmotivated to do string manipulation

dbdr: yeah that's something else

struct: I hate string puzzles

FuSheng: wait

AntiSquid: i did bruce lee before i did chuck norris and for some reason chuck norris one felt harder :p

FuSheng: ascii art is hard

FuSheng: :(

FuSheng: uhh can i just use this lol

FuSheng: import datetime

FuSheng: print(datetime.datetime.now())

FuSheng: wew dark mode feels better

MadKnight: it sure does

FuSheng: damnit ima go back to descent mountains

BioLampshadeFromTheSwamps_299c: hi everyone

MadKnight: hi bio lampshade

BioLampshadeFromTheSwamps_299c: i'm looking for information on a design pattern

BioLampshadeFromTheSwamps_299c: I saw the implementation on a test but couldn't figure out which one it was

BioLampshadeFromTheSwamps_299c: i remember in the code they called the class driver and they tested if a file existed then it was loaded, if not an exception was thrown

reCurse: MSmits have you been able to identify the features leading to RPS in oware?

MSmits: reCurse not really, I didnt watch many games. It's just that whenever I found a good value for any parameter, if i micro-fit it (changes < 1%) I get wildly differing results, especially in regards to fitness vs Robo

MSmits: it is true for almost any parameter

FuSheng: zzzzzzzzzzzxxxzxzxzxzxzxzxzx

FuSheng: why does codingame look so hard

MSmits: it doesnt really happen when a parameter has a bad value. But as soon as it gets closer to the best fit for that particular local maximum, it starts to really vary in winrate

FuSheng: is there any other website that has stuff like this :c

reCurse: Ok I was hoping you had a better idea what difference in play leads to such unstability

MSmits: no. I usually start to understand the game better when i start running a meta mcts, but you dont want me to :P

reCurse: I ran a whole arena over all my ML checkpoints and while it's really stable, batches vs you and robo vary wildly

reCurse: My experiment with trying to diminish exploitation results in weaker play too

MSmits: weird

reCurse: Also learns a lot slower

MSmits: is that similar to using SA with a slower temperature change?

reCurse: I found a checkpoint that's slightly better in arena

reCurse: It's 90% vs you and robo

reCurse: So far anyway

MadKnight: anyone getting random timeouts ?

reCurse: Yeah

MSmits: What you really need is to submit that and have me overfit it again. If I dont succeed that time, then your good

MSmits: you're

FuSheng: OH MY GOD I FINISHED 1 TESTCASE LOL

MSmits: grats :P

FuSheng: ok crap now im failing the second one

reCurse: I'm more curious at what's going on tbh

MSmits: me too... we discussed it a few times here. Me and Astrobytes mostly. There is something weird about this game

AntiSquid: i test cased the finish

reCurse: I'm not interested enough to study the game too

reCurse: If it was something visual like a physics game then sure

reCurse: But abstract games like that, ugh

MSmits: I think it may be related to the fact that it is so simple as to be almost solvable

FuSheng: wait how does

reCurse: At least it's not uttt

FuSheng: thor is so weird

MSmits: uttt doesnt have this problem though :)

MSmits: the only overfitting you can do is a counter opening book

FuSheng: lol im just printing "E to make it go straight

reCurse: Maybe I should create a NN for the first 20 moves and another for the rest

MSmits: That could be a good idea in general

reCurse: Lots of things to try but not enough compute power

MSmits: if you are only using, say 30% of the code size, make one for the early, mid and late game

Astrobytes: endgames are weird

Astrobytes: Well, different

reCurse: The problem I have is it's too specific

reCurse: I'd like to find solutions that are not too arbitrary

Astrobytes: "make one for the early, mid and late game" - sounds like the best thing to try

reCurse: But right now I think it's wasting too much time with the end game

AntiSquid: what's the full secret passcode for c++ bots? i have

1. pragma GCC optimize "O3,mavx2,omit-frame-pointer,inline,unroll-loops"

MSmits: wasting too much time training?

AntiSquid: just realized i left it @_@

reCurse: I'm sure 80%+ of my resources are dedicated to endgame

MSmits: or searching?

reCurse: Everything

FuSheng: nooooooooo i dont understand thor game :(

MSmits: use a solver

reCurse: Playing them out, learning them, etc

MSmits: i think it might actually help here

reCurse: There are endgames which can't be solved

FuSheng: OH MY GOD THERES CARD GAMES ON CODINGAME

reCurse: Not without spending too much anyway

reCurse: Can you stop spamming FuSheng

MSmits: i mean solve them near the turn limit, many games go to 200

FuSheng: ok sorry

MSmits: my bot solves them 25-30 turns before the end

reCurse: Sure when there's a solution

MSmits: should say plies

reCurse: There's a lot of time spent on unsolvable games

MSmits: there always is, with a turn limit

reCurse: If the turn limit was pushed to 1k it would be even more waste

MSmits: yes it would

reCurse: Getting rid of that condition would speed up training by at least 10x I think

MSmits: could be

reCurse: Either that or like I said I just train on first 20 moves

MSmits: you might be teaching it the best way to keep into an 48 seed loop

MSmits: and not know what to do with less than 48 seeds

reCurse: For all I know it's spending all its time learning the fine difference between having 3 seeds left and 4 seeds left

MSmits: and your opponent will make a mistake

reCurse: When it doesn't matter

MSmits: thats possible

MSmits: you could do the following

reCurse: I need my own server farm

MSmits: instead of first 20 moves

MSmits: stop at a seed limit

MSmits: say, stop at 20 seeds

MSmits: left

reCurse: But how do I determine winner

MSmits: from the score ?

MSmits: you might end up with a bad board layout ofc...

reCurse: There's tons of games where the winner is behind the whole game

MSmits: but just doing the first 20 turns is just as bad

reCurse: Even if I train on 20 I still need to playout the whole game ugh

MSmits: ok, do this then

MSmits: keep track of states

MSmits: as soon as one repeats, it's a draw

MSmits: games end on loops immediately

reCurse: I'm afraid it will learn how to avoid losses

reCurse: With a mechanism that doesnt exist

dbdr: you don't control the opponent

reCurse: It would be less of a problem if at 200 turns the game would always end as a draw

reCurse: Maybe just use current score then

reCurse: shru

reCurse: g

dbdr: if the opponent chooses to get back to the same state

dbdr: it does not look like he is winning

reCurse: Ohh

MSmits: oh right, that's what I should have said

reCurse: Alphago used self resignation

MSmits: on repeated state, higher score wins

dbdr: that might be the reverse problem

dbdr: the guy ahead choses the loop then :)

MSmits: the thing is, if both sides decide to loop, then they dont expect to gain any more score out of the rest of the game

dbdr: so intuitively I would count a loop as a draw

MSmits: no, as a win for the one with most score

reCurse: Is that detectable from a state?

reCurse: Instead of a sequence

dbdr: disagreed based on what we just said

MSmits: it's not detectable from a state

dbdr: you can include the set of previous state hashes into the state

dbdr: then it is :)

MSmits: but from any state, the future score change does not depend on the past states, only the current state

MSmits: except when the turn limit comes into play

reCurse: Maybe I should just trust my AI when it says it's over

reCurse: And allow it to kneel

MSmits: what do you mean says it's over?

dbdr: but what if it hasn't learned yet?

reCurse: If it estimates chances of winning is at 5%

MSmits: oh ok

dbdr: the estimate is random at first

MSmits: it may learn to lie about that :P

reCurse: Sure, just let it kick in after a while

reCurse: Might not even matter if it's random too

reCurse: Results are random at first anyway

dbdr: sure

reCurse: Lie about that?

reCurse: No it just confirms its beliefs

reCurse: It won't get surprised

reCurse: If the theory that the game is decided on the first half holdds

reCurse: Then it's a net benefit

dbdr: you can constrain the evaluation to be symmetrical no?

reCurse: Already did

MSmits: hmm, well I am not sure how your process works, but if you're going to give it the win if it thinks it has 95% chance to win, it might learn to tell you that when it isnt true

dbdr: then it cannot lie

reCurse: No it's the other way around

reCurse: If it doesn't see a chance of winning it resigns

MSmits: oh ok

MSmits: btw I am currently doing cg bench to finish up langtons ant. Only 3 params. Want to get them to a good value before I quit

MSmits: dont think it'll beat Redstrike, but it'll be a very strong second

MSmits: then i'll get back to experimenting a bit with oware. I won't keep overfitting though. I just want to leave it where it's strong before i move on from that as well

reCurse: I thought I already did that...

MSmits: did what?

reCurse: Leave it where it's strong before moving on

MSmits: well your bot really wasn't strong than Robo's. It was just stronger vs Robo imho... they were almost on par otherwise. I'm guessing it is now though

reCurse: This is entirely distracting me from FB too

reCurse: You keep saying my bots aren't as strong as whoever else is competing :P

MSmits: i mean what i did was, i tried A LOT of parameter combinations

MSmits: robo was harder to beat most of the time

MSmits: and you already noticed it is easy to overfit vs someone

MSmits: dont you think your previous version was a bit overfitted vs Robo?

reCurse: How could I tell

MSmits: you can't, but all i am saying is, your 60 something % winrate vs robo doesnt say much

reCurse: Also it can't overfit vs robo because it doesn't learn vs him

MSmits: it can still end up with a framework that works better vs him than the other way around without being objectively stronger

reCurse: You're tuning your params according to benches

reCurse: My AI is tuning its params according to itself

reCurse: Huge difference

MSmits: I know

MSmits: but it still ends up with a structure that may be weaker to some and stronger vs others

MSmits: even if it is not intentional

MSmits: rps happens whether you want it or not :)

MSmits: I actually have had versions that lost 95% of games vs Robo and won 55% vs you

MSmits: and not just with 1 value either... in a range of param values

MSmits: but it's moot now, you've improved your NN

reCurse: Maybe your params were naturally more predisposed to overfitting my bot

reCurse: Who knows

MSmits: completely possible

reCurse: That's why I wanted more game insight

MSmits: honestly I dont think this game is much easier to understand than uttt :P

reCurse: Well it has score for one

PatrickMcGinnisII: just fyi MSmits running the minmax on UTTT small boards took 3s on my machine, and would never complete in under 1s on CG

MSmits: sucks PatrickMcGinnisII :(

PatrickMcGinnisII: stupid php

MSmits: true reCurse, but as you noticed, you can have 5 points vs 20 and still win because you're hogging all the seeds

PatrickMcGinnisII: so I then tried to put the results in a constant and kept getting codesize too big

reCurse: The simple fact AI can learn something vs not at all for uttt

reCurse: Makes me think there's something readable

MSmits: it's also a simpler game. If you had more resources, you could probably do the same for uttt

MSmits: i mean both locally and on cg servers

PatrickMcGinnisII: So i shrunk it to 44k, and then ran into line break issues in the ide

MSmits: awari is actually solved, the only thing our oware has thats different is the handling of loops and the turn limit

PatrickMcGinnisII: so now I can autohandle small board decisions

MSmits: i think uttt is slightly more complex than checkers

dbdr: PatrickMcGinnisII don't you want to use a different, faster language?

PatrickMcGinnisII: so until i evolve my big board ai, I'm stuck at 2nd place in wood

MSmits: i told him many times dbdr. He likes php and the challenge

PatrickMcGinnisII: yes, is a challenge

dbdr: you have the spirit to do golf :)

MSmits: dbdr are you good at regular multis? you seem to really excel at golf, optims and such

PatrickMcGinnisII: So a few days of fumbling around and I finally am getting from cp from uttt

PatrickMcGinnisII: cp from

dbdr: well, you can check my results, MSmits. I have a few good ranks in multi, less than optim and golf indeed

PatrickMcGinnisII: i thought my simple approach would get me silver....stupid boss is blocking me

PatrickMcGinnisII: golf even sucks for php

MSmits: I play those board games specifically, i dont see you there too often. I am guessing you did a lot of those older multis and did well

dbdr: I had a lot to learn, and using Rust is obviously better than Java to get that last bit of performance

PatrickMcGinnisII: gold is the goal, but the game has been out so long the competition is stiff

MSmits: your bandas and breakthrough are good

dbdr: there are still many multis I haven't done seriously

MSmits: I'll try to beat you at breakthrough in the near future... it seems like a real challenge. No easy way to make a good bot

dbdr: even more true for you MSmits ;)

PatrickMcGinnisII: how do yall get large datasets (<100k) into your CG bots?

MSmits: well for me it's hard to do them half seriously

MSmits: so it's not at all, or all the way

MSmits: most multis are completely untouched

dbdr: kind of same, but it also depends on the game

dbdr: I'm gold in codebusters with a simple heuristic

MSmits: PatrickMcGinnisII you can do max 200k if you encode it with unicode characters

dbdr: CSB too of course :)

MSmits: I should do one of those heuristics for an old multi and get Dutchie nr 1 :)

MSmits: counterBalance still has it

PatrickMcGinnisII: unicode, hmmm...ahh 2 bytes per char....I just used the 88 web safe chars

MSmits: it gets messy pretty quickly if you try to use the full unicode range

PatrickMcGinnisII: i should say, php safe

MSmits: some of that stuff doesnt copy-paste properly

MSmits: invalid characters

dbdr: worse than that, it's not legal characters

MSmits: yeah

MSmits: you can send it still

PatrickMcGinnisII: i stuck with $symbols="0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!#%()*+,-.:;<=>?@[]^_`{|}~";

PatrickMcGinnisII: for encode/decode

MSmits: not sure how, re curse did it, others too

reCurse: Copy/paste works on invalid chars

MSmits: didnt work last time i tried...

dbdr: any value from 0 to 65536 as binary? I doubt it

reCurse: Then you did it wrong

MSmits: that's possible

MSmits: might be my operating system?

MSmits: or something?

MSmits: browser?

MSmits: not sure what controls the clipboard tbh

reCurse: No good software would risk corrupting your data because it thinks it's invalid

PatrickMcGinnisII: hmmm, extended codes would have worked too i guess

MSmits: so you're using bad software reCurse?

reCurse: No?

MSmits: :confused:

Uljahn: can you submit via handcrafted request to api?

reCurse: Why would a good software change data on your behalf because it thinks it's invalid

reCurse: It will just treat the char as invalid and keep it

dbdr: because it's the correct thing to do?

reCurse: ...

dbdr: as per the standard

MSmits: I have no opinion on this as I have no ideas of the pros and cons. Pasting unicode stuff into the CG IDE is not exactly what systems are designed around

PatrickMcGinnisII: i still think only 215 of the standard codes are gonna interpret correctly

reCurse: Anyway, all of the software I use correctly keeps invalid characters through copy/paste and whatnot

MSmits: ah ok

dbdr: you don't control what softare CG uses

reCurse: I don't even

reCurse: What

dbdr: the data goes through many layers

dbdr: maybe we don't speak about the same thing

PatrickMcGinnisII: CG chat breaks on alot of weird combos

MSmits: but apparently the CG controlled layers dont corrupt the data dbdr

MSmits: otherwise reCurse wouldnt be able to do it

MSmits: so it must be on my end

reCurse: If it sees an invalid codepoint why would it delete the character

reCurse: That makes no sense

reCurse: Nothing does that

PatrickMcGinnisII: but the ide accepts alot more

dbdr: https://www.fileformat.info/info/unicode/char/fffd/index.htm

reCurse: That's for display purposes

dbdr: there's a unicode character designed for that especially

reCurse: Not data

dbdr: no

MSmits: it doesnt delete the character, it creates a ? with a box around it or something

MSmits: so it changes it

dbdr: "used to replace an incoming character whose value is unknown or unrepresentable in Unicode compare the use of U+001A <https://www.fileformat.info/info/unicode/char/001a/index.htm> as a control character to indicate the substitute function"

dbdr: *replace*

MSmits: that sounds like what happens in my case

reCurse: For display

dbdr: reCurse: are you checksumming the data? if you use it for weights, it's possible you get garbage in those rare cases :)

dbdr: and don't know about it

PatrickMcGinnisII: �

MSmits: that would be an interesting bug lol

reCurse: Yes it's 100% valid

reCurse: Because it would make no sense for a transport layer to change its data because it thinks a codepoint is invalid, when it could be valid in the future

reCurse: Makes 0.0000% sense

PatrickMcGinnisII: i can represent a 26 bit number with 4 characters and 88 symbols

MSmits: I should try it again at some point. Last time i tried it was with C#, that probably doesnt matter though

reCurse: That's the very definition of unicode to be extensible

reCurse: Deprecating data would be senseless

dbdr: some codepoints are guaranteed to be invalid

dbdr: that won't change

dbdr: why do you think they defined those?

reCurse: Doesn't mean it's safe to replace them?

PatrickMcGinnisII: so utf16 works in CG?

MSmits: you know... it seems easy to prove doesnt it? reCurse says he proved it works. dbdr now you try it. Let me know if you succeed :)

reCurse: Not utf16

reCurse: utf8

dbdr: utf8, sure

dbdr: but that can take many bytes per character

reCurse: That's the whole point

reCurse: CG counts characters not bytes

struct: yah

dbdr: utf8 is binary, not characters

dbdr: we are mixing everything :D

PatrickMcGinnisII: reCurse, hell to the yea, i get it now

dbdr: CG counts UCS2 units, I think.

PatrickMcGinnisII: i still think there are weird combos to avoid, care to clue us in reCurse?

reCurse: No

reCurse: It counts UTF8 characters

dbdr: given to us for legacy reason from the time where there were less than 2^16 unicode codepoints

dbdr: absolutely not

darkhorse64: Using PHP on UTT is like bashing your head with a hammer and pretending it's good because when you stop you feel better

dbdr: "UTF8 character" does not make sense

struct: Using PHP is like ...*

struct: Fixed

PatrickMcGinnisII: tx darkhorse

reCurse: Number of codepoints if you prefer

reCurse: Come on

PatrickMcGinnisII: I'm finally getting some CP in php in uttt

dbdr: but there are more than 2^16 codepoints

PatrickMcGinnisII: :p

MSmits: how do you get cp in wood PatrickMcGinnisII?

reCurse: So?

darkhorse64: PHP uses $ to pretend it's a rich language

dbdr: can you pack more then 100K * 2 bytes?

PatrickMcGinnisII: MSmits just got to be in top 1000 for cp

MSmits: ah

dbdr: it's not codepoints either

dbdr: my guess is their backend is all Java, so they just count length of java String objects

dbdr: which is UCS2 encoded

PatrickMcGinnisII: still .11 points from silver...stupid boss

reCurse: Pretty much

reCurse: And it won't replace invalid characters

reCurse: To come back to the original point

darkhorse64: There is an achievement to code horse duals in bash, there should be onz for TTT in PHP

MSmits: well i'm encouraged by your conviction. I will give it another try if I run out of codespace next time I try to make an opening book for a multi. thanks reCurse :)

MSmits: though running out of codespace doing that seems very hard

reCurse: Haven't had that problem

PatrickMcGinnisII: hmm, how many 26bit data points can you pack in CG with unicode?

dbdr: reCurse: I think we are speaking too generally to be clear. do you want to clarify what technique you use?

dbdr: or is it a trade secret? :)

reCurse: Well it used to be, but not anymore because I keep getting involved in discussions

dbdr: your choice

PatrickMcGinnisII: hahahaha

PatrickMcGinnisII: i would get there eventually...next year

reCurse: I'll just say put codepoints to get a lot more than 8 bits per character

dbdr: given almost all characters is valid, it does not matter much if it's 99% or 100%, does not give a measurable advantage

dbdr: i'm rather curious from the correctness point of view

reCurse: All I can say is I'm using invalid codepoints and it has never mattered

dbdr: sure

reCurse: And upon further thinking it would make no sense for it to matter

dbdr: are you saying you get exactly 16 bits per character?

MSmits: or more than that?

reCurse: You can't get more than 16 bits because of what dbdr said

reCurse: But in theory you could

reCurse: If java ever supports more for instance

dbdr: it all depends how CG counts, doesn't it

PatrickMcGinnisII: it reminds me of md5

reCurse: 99% sure CG only does string.length

dbdr: yeah, I think the same

MSmits: allright, getting some food, ttyl

PatrickMcGinnisII: still have to put in linebreaks, ide hates long data

reCurse: Hehe yeah

dbdr: the thing, I don't think you can paste byte 0 in your IDE, can you? or send it through the API

PatrickMcGinnisII: crap, my gf is off work... gotta make the bed. ;)

PatrickMcGinnisII: laterz, good stuff reCurse

reCurse: You want to avoid control characters sure

dbdr: I though you said all characters get through

dbdr: except those that don't? ;)

PatrickMcGinnisII: maybe i will leave chat open, recurse might spill some more...gj dbdr

reCurse: Well even control characters shouldn't matter tbh

reCurse: It's just software tend to attach meaning to them

reCurse: But it's not an encoding issue

reCurse: Completely different

reCurse: It's not because of invalidity but because of meaning attached to it

JBM: awwwww missed an encoding debate again :(

JBM: so what's the consensus? CG just counts codepoints?

reCurse: Yes

JBM: with no normalization?

dbdr: UCS2 units, not Unicode codepoints

dbdr: welcome to the chat JBM :D

JBM: i'm like feeling 10 years younger

dbdr: just added level of complexity, we were missing that ;)

JBM: debating that on the early days of the golfing stackexchange

dbdr: doesn't golf usually count bytes?

dbdr: outside of CG?

JBM: multiple sites, multiple rulesets

JBM: and bytes still need an encoding

dbdr: yes. you could restrict to ASCII, honestly. or UTF8 bytes

JBM: ucs-2 doesn't seem to tolerate emoji, does it?

dbdr: you use two of them for one unicode codepoint

JBM: ugh

JBM: wasn't that encoding supposed to be a mistake that should have died years ago?

dbdr: actually that's UTF-16

dbdr: https://en.wikipedia.org/wiki/UTF-16#Code_points_from_U+010000_to_U+10FFFF

dbdr: mistakes never die

JBM: well if ucs-2 does the same, it's no better

dbdr: UCS-2 just does not support them, it seems

dbdr: otherwise they're essentially the same

JBM: so ucs-2 is just a list of unicode codepoints, laid out with whatever byte order, excluding anything out of the BMP?

dbdr: I think the word pretty much converged on utf-8 where space matters, and utf-32 when you want constant time indexing

JBM: yes, that's where i was

dbdr: but you still need to support all the choices made earlier

JBM: with ucs-2 supposed to be dead “doesn"t handle unicode” and utf-16 “useless compromise”

struct: I just use the chars available on my keyboard

struct: and I think it will remain this way

JBM: that's not much

JBM: unless you've got a 16-bit keyboard

dbdr: 1 bit keyboard with two keys

JBM: ctrl and alt?

dbdr: you can type anything

dbdr: 0 and 1

JBM: lemme summon chuckie

PatrickMcGinnisII: yall ain't right in the head

PatrickMcGinnisII: \u0079\u0061\u006c\u006c \u0061\u0069\u006e\u0027\u0074 \u0072\u0069\u0067\u0068\u0074 \u0069\u006e \u0074\u0068\u0065 \u0068\u0065\u0061\u0064

AntiSquid: MadKnight ban this spammer ^

dbdr: MadKnight ban this banner ^

struct: ban me instead

ThePythonian: Why not ban everyone?

dbdr: /ban *

ThePythonian: del server

AntiSquid: no need to del server

AntiSquid: they are burning when new league opens

Astrobytes: Or someone promotes in CSB

dbdr: Astrobytes: thanks for mentioning Peep Show

dbdr: I like it

dbdr: similar humour, but more weird :D

Astrobytes: dbdr it really is their finest work imo

dbdr: reminds me a bit of the UK Office

Astrobytes: Style-wise, for sure yeah

Astrobytes: And equally cringe-based humour

dbdr: exactly

dbdr: some internal monologues are just incredibly funny

Astrobytes: Ohhh yeah :D

AntiSquid: MadKnight are you there?

Uljahn: has it something common with Monty Python?

Astrobytes: Quite different humour Uljahn, though I love Monty Python

dbdr: I'm sure MP had some influence on them still

dbdr: but they have their own style, definitely

Astrobytes: Pythons influenced almost every UK comedy, and beyond

dbdr: yeah :)

dbdr: also Fry and Laurie

Astrobytes: Definitely

Astrobytes: You've also seen the Blackadder series'?

dbdr: yeah, but I didn't get that much into it

Astrobytes: Some series are better than others, but it can be wickedly funny

dbdr: I tried a few episodes in the 17th century or sth, and a few 1st word war

Astrobytes: I guess it's quite British humour, even though it's ripping the p*ss out of the British establishment through the ages

dbdr: I love british humour

dbdr: but yeah, maybe this one has more references to specifically british things

Astrobytes: It's certainly quite unique :)

Astrobytes: True

Astrobytes: Like Prime Minister William Pitt the Younger is actually portrayed as a young boy in Blackadder, just based on "the Younger"

Astrobytes: Lots of little things that are easy to miss

Astrobytes: Twists on historical facts, characters etc

MSmits: is the humor in last week tonight by John Oliver typical British humor?

MSmits: if so, then I really like it too

Astrobytes: Yeah, mostly. Tweaked a bit for American/worldwide audiences but at heart yeah

MSmits: cool

Astrobytes: He is good

MSmits: very, I remember when i he did the daily show for a few months

MSmits: he was almost as good as jon stewart and better than trevor noah

Astrobytes: Yep me too. Jon Stewart was a hard act to follow

Astrobytes: I'd almost forgotten about Trevor, lol. HE just wasn't all that

MSmits: um he is still doing it :P

Astrobytes: I know, I stopped

Astrobytes: Watching, that is

MSmits: ahh ok, I still watch it, but only because there isnt a better alternative on the air here. I have to see other shows on YT

MSmits: like colbert, seth meyers, jimmy kimmel. I prefer those

MSmits: conan is awesome too

Astrobytes: I very rarely watch anything on the TV

MSmits: same

Astrobytes: Well, I hook it up to the PC and watch stuff but you know what I mean

MSmits: yea

MSmits: trevor noah's stand up is actually quite good

MSmits: I just dont think he's that great of a late show host

Astrobytes: Haven't checked his standup, but he's a pretty wooden host

MSmits: yeah

MSmits: try some netflix specials of his, you wont regret it

MSmits: bill burr cracks me up more though

MSmits: and ricky gervais

MSmits: ricky's stand up is the best I've ever seen

Astrobytes: Cool. Gervais is good, haven't seen any of Bill Burr's

Astrobytes: I grew up watching Billy Connolly standup

Astrobytes: *standup shows before I get a smart comment :P

MSmits: Bill Burr is like Seinfeld commenting on stupid daily stuff only it is less being annoyed and more raging about it

MSmits: very politically incorrect

MSmits: also

MSmits: dont know bill Connoly

Astrobytes: Famous Scottish comedian, look up some of his old classic performances. Very un-PC but extremely funny

Astrobytes: https://www.youtube.com/watch?v=30IOKP2GcjY

MSmits: can't really understand what he's saying

MSmits: automatic subtitles gets all the words wrong also

Astrobytes: Seriously? He's quite polite

MSmits: it's not about politeness

Astrobytes: Accent-wise

MSmits: well it's too much for me =/ I love the accent, I just dont know what he's sayin

Astrobytes: (easier to understand) but he does get into some Weegie language

Astrobytes: (Glaswegian)

MSmits: ahh

Astrobytes: Maybe look up his international shows, he's less colloquial-scots in those

MSmits: i was thinking Wookiiee

Astrobytes: lol

struct: A bit hard to understand

struct: But I can manage

dbdr: This guy really knows how to tell a joke: https://www.youtube.com/watch?v=Oseqh7SMIvo

Astrobytes: Billy's a wonderful storyteller, very Celtic way of being funny

MSmits: I must have watched every talk show appearance by Norm several times now :P

MSmits: he's most funny on talk shows

MSmits: outside of that, he's ok, but brilliant on those late night talk shows, mostly conna

MSmits: conan

Astrobytes: ffs, punchline was lame but how he got there was great :D

Astrobytes: Nicely done

MSmits: all his jokes are like that

MSmits: the moth joke is famous

Astrobytes: Will check him out

MSmits: https://www.youtube.com/watch?v=JUjIM-GFWhk

Astrobytes: FOR F*CKS SAKE

dbdr: yeah, the moth one is pretty good! didn't know it

Astrobytes: Dear oh dear. Love it

Astrobytes: Like his style

dbdr: I get the feeling he might be slightly high, no?

MSmits: he speaks that way

MSmits: someone animated the moth joke also

Astrobytes: Nah he ain't high

dbdr: maybe he speaks like this when he's high ;)

MSmits: anyway, watching norm appearances is one of the things i do when i am too tired from work to do any coding

Astrobytes: Nice. I pet my cats and contemplate my existential dread :P

MSmits: sounds like a reasonable alternative

Astrobytes: :D

AntiSquid: hey for cgbench : 20:05:14,355 FATAL Error while retrieving account cookie and session java.lang.IllegalStateException: Login failed, please check login/pwd in configuration

AntiSquid: i don't get it @_@

AntiSquid: the login is correct

MSmits: we can check it for you

MSmits: :P

Astrobytes: lul

AntiSquid: sure i give you an alt

MSmits: no no

MSmits: i have had this sometimes

Astrobytes: it's a session issue no

Astrobytes: ?

AntiSquid: so what should i do ?

MSmits: i think maybe it is more likely to happen when you login to the website and back out, or maybe when you run other games during the batch

MSmits: not sure

MSmits: does it happen every time:

MSmits: ?

AntiSquid: yes

Astrobytes: Yeah, I've always cleaned cache/cookies and stayed logged out when I've run bench, not sure if that is necessary

AntiSquid: does it need to be specific email address ?

MSmits: your e-mail adress is the login isnt it

AntiSquid: i mean from certain providers

MSmits: oh, nah if you can login here, you can login through cgbench

Astrobytes: Yeah as long as it's a valid CG account it's cool

dbdr: well, do you ever need to relogin?

dbdr: I never seem to get logged out

AntiSquid: which browser does it try to access though ?

AntiSquid: or it's not via browser ?

MSmits: probably not

AntiSquid: i mean why else would i have to log out

MSmits: no i meant, clicking logout may do something server-side that invalidates your cg bench cookie

MSmits: if you dont do that, then i have no idea what happens

MSmits: Astrobytes I am fitting langtons ant parameters now. It's such a relief to have a coherent result for once

MSmits: https://imgur.com/a/Bb3efvT

MSmits: my next dot falls right on that line also

Astrobytes: yay, no wildly varying periodic madness

MSmits: yeah

MSmits: can't get close to Redstrike with it I think though, but that's ok. Ill fit what I have and then it's time for Onitama I think

Astrobytes: Don't imagine there would be for Langton's. I haven't looked at Langton's lb, how far ahead is he? I'm going back to Oware when I'm better.

MSmits: he's extremely far ahead. Even more so than the NN's in oware if that's possible

Astrobytes: wow, what's he using?

MSmits: no idea, but he uses java, so I am pretty sure it is just a good algorithm or a magic heuristic somewhere

MSmits: I think he spent a lot of time experimenting

Astrobytes: Interesting. I'd guess it involves heavy pruning

MSmits: I'm also heavily pruning, in fact that parameter I just shared is the pruning parameter

MSmits: the number is how many moves I consider

Astrobytes: It's necessary in Langton's for sure, but he must be doing some very clever pruning

MSmits: maybe, I thought mine was pretty much common sense

MSmits: it's just the available cells in the ants path

Astrobytes: That's the most logical yeah

MSmits: He might be doing something in the last few plies

MSmits: langtons ant really hinges on the last few cells colored

MSmits: because often the ant goes off into an empty area and make a whole lot of cells one color at the end

Astrobytes: Right, yeah that's a good point

MSmits: But for me, not worth spending two weeks experimenting just to try and achieve 2 -> 1

Astrobytes: I've not really gone deep enough into it to comment much further but that I have definitely notices

Astrobytes: *noticed

MSmits: onitama caught my eye because it is very bitboardable. I did not know that

Astrobytes: That's one of euler's multis right?

MSmits: Yeah

Astrobytes: Are you done with checkers now then?

MSmits: Oh yeah, I could have done 2 more things. Write a better eval and switch to a different search

MSmits: different search would be too much work

MSmits: and better eval... well... I didnt feel like doing another oware

Astrobytes: You went MCTS EPT on checkers?

MSmits: it also feels like less of an achievement with a game thats less popular in CG and done to death outside CG

MSmits: yes I did

Astrobytes: Interesting.

MSmits: very raw implementation. I think I could easily get 4 with it. But the top 3 are very far ahead

Astrobytes: Well, not really. CG versions of games are often very different because of the time constraints

MSmits: true

MSmits: I was speaking more of the game itself, the bots would be different of course

MSmits: also, people overestimate how easy it is to find good free bots by google, for checkers

MSmits: you can find opening books though

MSmits: but then you need to make a selection, which is so much work, you may as well write your own meta mcts fo rit

Astrobytes: Where's the fun in finding a bot to copy?

MSmits: i wouldnt copy a bot, but i do like finding ideas for a good eval

Astrobytes: Oh for sure yeah

Astrobytes: brb

struct: I enjoy seeing others code

struct: Seeing how the implemented stuff and such

MSmits: ah, not me, i have trouble reading it somehow

struct: When I'm done with CSB I'll send you a copy

MSmits: maybe because I have a totally different style than most? I dont code collaboratively

struct: Me neither

MSmits: struct ohhh, that is one thing I would really like. A working AVX sim for csb

MSmits: I'd spend the time trying to understand that

MSmits: would not need to see the whole bot, just the avx calcs would be fine :)

struct: The only hard part is the collisions/bounces tbh

MSmits: maybe once i see it work, I could figure out how to use it in other places

struct: rest is preety straight forward

MSmits: yeah I just need a more complicated example than those squareroots by Marchete. Mind you, they were useful, but its not the same

struct: Well it does have if elses

MSmits: cant really avoid that

MSmits: i did some coding for compute shaders. If elses hurt a lot there too, but as long as there is a net-gain it is ok

struct: From now I just need to focus on making it work

struct: and ditch peformance

Astrobytes: Yeah I don't mind seeing people's code, always learn something

struct: I mean, stop pre opptimizing

Astrobytes: oh chat scroll

Astrobytes: Yes struct, do that

struct: I was trying AVX to gather stuff from array[]

struct: but AVX performance is similar to for loop 8 times

Astrobytes: yeah, multiple simulataneous operations. Just the loading/unloading overhead you need t obe concerned about iiuc

Astrobytes: *simultaneous

struct: Is just that it does this

struct: http://chat.codingame.com/pastebin/dfe5a9cb-4a8c-4f67-9773-71fd195a45ee

AntiSquid: http://chat.codingame.com/pastebin/555b9b98-ab9a-43d7-8fe0-f2918a605c24

AntiSquid: i don't get it

Astrobytes: No idea what I'm looking at struct

AntiSquid: turned off AV to see it maybe blocks

MSmits: ok, i have maybe a long shot

MSmits: AntiSquid

MSmits: one time I had problems with the encoding of the text file the program is in

MSmits: if I ran it straight from the file VS uses

struct: https://software.intel.com/en-us/cpp-compiler-developer-guide-and-reference-mm-i32gather-ps-mm256-i32gather-ps

MSmits: so I copied it in a text file with a different encoding and the error stopped

struct: Basicly you pass indexes

AntiSquid: the config file?

struct: and array

MSmits: the code file

struct: then you get data from that array on those indexes

AntiSquid: ahm but it bugs at login

AntiSquid: it doesn't even reach that point you mentioned

MSmits: oh, that last thing you pasted seemed different

MSmits: i thought you had a different error now

AntiSquid: nah

AntiSquid: struggling with login lol

AntiSquid: tried diff mail providers

MSmits: are you trying to use multiple alts to run more cg batches?

MSmits: you mentioned alts earlier

Astrobytes: OK struct, yep, makes sense.

Astrobytes: The fact it's slower/as slow as a loop must be the loading/unloading. MSmits help

struct: its a bit faster

MSmits: eh, i am no expert, but i know that if you try to load it from a source with data that is spread out, it's going to be slow

struct: but only like 20-30%

MSmits: you can load sequential data very quickly

Astrobytes: 20-30% is not significant?

MSmits: normally you'd get 200-300%

MSmits: if your avx is doing well

struct: Even more

struct: Can even go above 500%

Astrobytes: OK

Astrobytes: But yeah, given how it works loading fragmented data isn't gonna work well

MSmits: thats why my AVX-ed ucb1 code didnt work

MSmits: I was gathering visit counts from separate nodes

Astrobytes: Yep, that makes sense

MSmits: even though the nodes were sequential (being siblings), they were also larger objects with other stuff in between

MSmits: like wincounts, indices etc.

struct: was performance worse?

MSmits: almost the same. slightly worse than the fastest way I could do it

MSmits: the fastest way being a lookup

MSmits: but i dont use lookups now

Astrobytes: Absolutely, yeah. Is there any way to make that contiguous? I don't think there is

MSmits: so it's about the same

MSmits: you can make it contiguous by ruining other parts of the code and making those slower

struct: why no more lookups?

MSmits: i hated having an artificial limit on visit counts

Astrobytes: lol, doesn't sound like the best tradeoff

struct: I thought cache misses

MSmits: cache misses hurt, but apparently not enough to make lookups slower

MSmits: in this case

dbdr: struct, compilers also know how to use AVX automatically sometimes

dbdr: in that case you will get no boost from doing it by hand

MSmits: the thing is, even though you may have 500k visits, you only use the first part of the array a lot, so cache misses may be less painful

dbdr: because it's the same code

struct: oh

struct: is this why?

dbdr: it could be

dbdr: depends on the case

MSmits: not in my case

dbdr: that's what you can tell if you look at the asm

MSmits: I was doing a special inverse squareroot for avx that was faster because it has less accuracy

struct: special?

struct: I will look at it dbd r

MSmits: yeah, it uses half sized floats

MSmits: 16 bit floats

MSmits: dont remember the details

dbdr: isn't it the low precision inverse squareroot?

MSmits: one of them

struct: the clock cycles

struct: of the instruct of gather

struct: are quite high

Astrobytes: Just read that as "the dock cycles of the instinct of father"... Probably time to leave CG

MSmits: go get some sleep :P

Astrobytes: feck that, Picard night

Astrobytes: Illness be damned

MSmits: hf with mr :poop:

Astrobytes: lol, not going just yet

Astrobytes: I'll just read the chat very carefully for now

MSmits: gewd eye dea

Astrobytes: 'lul'

Astrobytes: :P

MSmits: :open_mouth:

Astrobytes: You magic carpet has broken down you say? ;)

Astrobytes: *Your

MSmits: wrong song, thats a whole new world. That one doesnt impress me much

Astrobytes: Jeez man

Astrobytes: I feel there is a Shania Twain meme appropriate here but... no

MSmits: Yeah that one didnt impress me much either

Astrobytes: :smile:

MSmits: https://www.youtube.com/watch?v=pDNRjL7cY9w this one is better

AntiSquid: ok now i can finally bench vs Automaton2020

AntiSquid: Automaton2000

Astrobytes: MSmits https://www.youtube.com/watch?v=_sk0W1_1sJ8

Automaton2000: i think it would work

MSmits: what was the bug AntiSquid

AntiSquid: multiple ones

MSmits: and why is everyone saying 2020 lately when pinging automaton

AntiSquid: asked neu man for help

AntiSquid: i use tab for autocompleting user names

AntiSquid: that thing keeps poping up

AntiSquid: someone trolling @_@

MSmits: argh my ears.... Astrobytes

AntiSquid: /ban Automaton2020

Astrobytes: MSmits as much as I would have given Shania one her voice is a different story :P

Astrobytes: and yes, that cover is atrocious

MSmits: yea

MSmits: btw Astrobytes, you asked how far Redstrike was ahead

MSmits: http://chat.codingame.com/pastebin/62cedbef-a67e-4c3d-8e72-deb5334854ed

MSmits: thats a cg bench. See how I am slaughtering the guys below me and how Redstrike is slaughtering me :P

Astrobytes: Ooft

MSmits: i cant even to a balanced test

MSmits: cuz noone is around my bot's strength

MSmits: do

AntiSquid: sharing a random song too https://www.youtube.com/watch?v=s9XDGD-HA1Q

MSmits: did maddy pm you that AntiSquid

AntiSquid: no lol

AntiSquid: MadKnight has no taste

Astrobytes: Nice. They still around are they?

MSmits: maybe, but the stuff he pms me sounds the same :P

MSmits: it's loud

AntiSquid: it needs to be noisy, it's music .

AntiSquid: /s

MSmits: meh, there's levels of noisy

MSmits: I like it less

Astrobytes: MSmits https://www.youtube.com/watch?v=axxJ5dmgIi0

MSmits: whyyyy

AntiSquid: MSmits i guess you prefer a soft female voice? https://youtu.be/1eZqMMQBXp8?t=130

Astrobytes: There would be no Rammstein if not for Ministry!

MSmits: AntiSquid that girl needs help

Astrobytes: lol AntiSquid

Astrobytes: MSmits https://www.youtube.com/watch?v=U-dc1Z4Z-8I

MSmits: :crying_cat_face:

AntiSquid: ok no bs this time, i actually do like this song https://www.youtube.com/watch?v=R0PMXNUY9mg

AntiSquid: but it's not what i usually listen to

MSmits: allright I am going to stop clicking on your links :P

AntiSquid: i usually just use shoutcast to listen to rnadom radio stations

Astrobytes: I like all the songs I posted bar the Still The One cover

AntiSquid: MSmits last link is legit

AntiSquid: if you don't like it, then you're like madao

AntiSquid: don't be madao

MSmits: well I like it in so far as I like screaming maniacs I encounter on the street

Astrobytes: AntiSquid sounds like Killswitch Engage or earlier Chimaira or something

Astrobytes: MSmits the one true classic: https://www.youtube.com/watch?v=gNhN6lT-y5U

AntiSquid: astro i don't know any of those ... i tend to just search 90s pop in about any language

Astrobytes: Oh I like at least something from just about every 'genre' of music. Not fussed on opera or modern pop/EDM though.

AntiSquid: MSmits give me some tips for best test results with cgbench, how do you go about it?

AntiSquid: oh yeah, i do listen to future funk i guess or whatever it's called, electro jazz ? too many random names but they are all similar

Astrobytes: Too much pigeonholing

AntiSquid: idk, they completely overlap imo

AntiSquid: is it possible to get AgentID of boss for cgbench ?

Astrobytes: Exactly, people need to categorise everything these days, let music be music

Astrobytes: Don't think so AntiSquid, rather - not that I know of

AntiSquid: that makes life bit more difficult then

AntiSquid: did we get rid of smits with our links?

Neumann: Yes you can

Astrobytes: Nah, MSmits will be cleansing his ears with some Naomi

struct: 2 people that were being diagnosed for coronavirus just escaped from the hospital

Astrobytes: O cool Neumann

AntiSquid: Neumann what's the agentid for boss ? (gold)

Neumann: AgentId : -2 is the boss of your league, or the defaultAI if in max league

Neumann: If you want to fight a boss from another league, you can also grab it's id from a replay of him

AntiSquid: ah ok thanks

Neumann: Gold of XMR ?

AntiSquid: yes

AntiSquid: stuck at the gates

Astrobytes: Did not know that, thanks

Neumann: 2223822

MSmits: back

MSmits: AntiSquid, if you want to fit params, start with short batches, maybe 100-200, spread them around the value you think is good

MSmits: then you'll probably see a curve coming, try more batches close to where you think it will peak and make those batches longer if necessary

AntiSquid: lots of params :/

MSmits: dont try to get too accurate at first

MSmits: switch params often

AntiSquid: maybe should have done a NN instead

MSmits: how many params, which game?

AntiSquid: XR

MSmits: I had very few params for XR

AntiSquid: about 20

struct: 20 params?

MSmits: wow

struct: wtf

AntiSquid: ya i had fewer ...

AntiSquid: but then i couldn't figure where i went wrong with my bot

MSmits: seems like i have about 5

AntiSquid: others seem to be around #30 legend with same approach, either some bugs or hell knows what ...

struct: what search do you use?

AntiSquid: and couldn't find any bugs .

AntiSquid: just basic minimax

struct: Never tried the game

miszu: The hardest/boring part of the AI game is to code the actual game engine

miszu: which takes like 70% of coding time

struct: I find it the most enjoyable part

BenjaminUrquhart: same

miszu: when the game engine is simple, then maybe. But when you need to play with physics and etc, so many little bugs you can have

struct: Does anyone have the link to the "non official" CSB referee

struct: I want to check some stuff

AntiSquid: what does "non official" mean?

JBM: any one that's not the one that runs on CG

struct: the one that was leaked

AntiSquid: https://github.com/SpiritusSancti5/codinGame/blob/master/Referees/Coders%20Strike%20Back/Referee.java

struct: thanks

struct: I guess the off by 1 errors are explained

NoiselessKnight_28b2: Is anyone coding in Java? Is this thing bugged? Im trying to copy the maze into an array but it can't even get to the end of my code before running out of time. As in, a singular nested for loop.

AntiSquid: you must have a bug somewhere

NoiselessKnight_28b2: I figured it out. My code was expecting the length of ROW to be C, but for some reason it isn't. Any idea why this is?

NoiselessKnight_28b2: ohhhhhh

NoiselessKnight_28b2: somehow R and C got flipped hahaha

NoiselessKnight_28b2: the order in which the scanner sees them. sorry for the nuisance

AntiSquid: no worries

AntiSquid: feel free to ask Automaton2000 any question

Automaton2000: but now that i think

tartiflet: hello, is there a way to get the full "expected" string when it's too long to be printed fully in the "console" frame

Z_Doctor: You could use an external ide to simulate a run and get the out from there.

MeirKlemfner: Hi

SilentKnightFromAfar_d6e: hi

tekki: hello all

tekki: wassup?

JavaMan: hard to win in shortest code using java

Z_Doctor: also could be that you are using poorly optimized java

JavaMan: you are probably right

Z_Doctor: and even then it's impossible to beat a good perl user at shortest too

JavaMan: does it also count the imported libraries?

Z_Doctor: It counts the import statements