Chat:World/2021-10-10
struct: nice rank eulerscheZahl
miszu: I have been running my code offline for 32 min and can only find a score of 11k for the horde scenario
miszu: 111k
miszu: *
struct: I think my code was taking under 2 minutes to find the 400k+
miszu: bruuh
struct: maybe increase the search time localy?
struct: instead of using 150ms use 1 sec per turn?
miszu: what you mean?
struct: idk
miszu: I simply repeat random moves
struct: from the first turn?
miszu: yeah
struct: that wont do it I think
struct: my code doesnt find the 400k in the first turn
miszu: ok so after running X seconds you take the best move, you move in then restart?
struct: i play the move and keep searching
struct: until game ends
struct: I dont know if I run it for seconds
struct: I cant remember
miszu: ok fair enough
miszu: I will do that approach
miszu: going to sleep now. Good night everyone
struct: gn
eulerscheZahl: thanks struct
eulerscheZahl: but my last improvement gives wrong answer :(
eulerscheZahl: fixed it. just not sure if the extra time cost is worth it
eulerscheZahl: yes, was worth it
jacek: :tada:
jacek: good morning
aCat: hi
Aliloulaye: What is the purpose of this chat ?
eulerscheZahl: chatting
AntiSquid: random rush league promotion in crystal rush \o/
Cyber_Loki: Hi dudes
Falkern: whats upp
Cyber_Loki: can you help me
Cyber_Loki: anyone
jacek: oO
jacek: stupid tanh. target = 1, prediction = -1. and i was wondering why no learning occurs
dreadylein: :D
eulerscheZahl: stupid codeforces. gives WRONG_ANSWER but no clue what's wrong
alexey4eg: user power of the brain
dreadylein: i guess the answer is wrong :P
MSmits: I think dreadylein is on to something
barbora: you can get WRONG_ANSWER if you output 0 paths, though not sure if that is your case
Ran_42: Why is there no way to add new code golf?
JFB: jacek - what game?
MSmits: don't know actually, but my thought after doing code golf was never: "I wish there was more of this". Could just be me though :)
Ran_42: xd
Ran_42: Actually i would like if there would be more :p
MSmits: yeah tastes differ. At least the is clash shortest.
MSmits: there
Ran_42: ye well I hope they would add more code golfs outside of clashes
MSmits: i guess the system only allows optims and multis right now
MSmits: as contributions
Ran_42: ye I guess only cg can add golfs
MSmits: you could ask for it on the forum, if a lot of people would request it, they might consider
MSmits: but i doubt there is enough support
MSmits: they're very busy with their escape thingy
MSmits: some optims are code golf-like
MSmits: as in the output needs to be golfed
Ran_42: which ones are those?
MSmits: I can think of 3
MSmits: sec
Ran_42: sure
MSmits: CFunge, Bender 4 and code of the rings
MSmits: the code itself is not golfed though, just the output. But the output is like a program
Ran_42: I see I will check those out
MSmits: not sure if it appeals to you, but it's something
Ran_42: thanks :)
MSmits: any time
_NikJ: Has anyone tried using the parameter fiddler with UTTT? It keeps throwing errors for me after 6th game
Cyber_Loki: https://escape.codingame.com/game-session/HCm-wdb-rrP-v0B
struct: maybe the fiddler doesnt support the uttt referee
struct: it was made when brutaltester was in version 0.2(?)
_NikJ: well it does say that it's a supported referee
eulerscheZahl: did you use the official UTTT referee or a modified version that can be run from the command line?
eulerscheZahl: official won't work. never did for any game
jacek: whats parameter fiddler
_NikJ: I used the one linked on the github of the fiddler
_NikJ: Under "List of compatible referees "
MSmits: for UTTT I recommend you write your own tool. It's such a simple game that it's better to do it yourself.
MSmits: it's good practice.
struct: yeah but they are compatible with brutaltester 1.0
struct: not compatible with brutaltester 0.2
_NikJ: I have the latest 1.0 version struct
struct: yeah the fiddler wont work
_NikJ: So I might as well write my own? :D Or is there a way to make this one work
struct: Just write your own I guess
struct: Usually I just send the move the other player made
eulerscheZahl: write a vanilla MCTS, then you don't have to tune parameters
struct: and if the player that is gonna move has found a win i send WIN
struct: and the referee will know who won
eulerscheZahl: also UTTT might be a really bad game for self-play parameter tweaking, as there are just 2 different initial boards
struct: yeap
struct: I agree with euler
MSmits: why is that a problem?
jacek: huh
MSmits: the game diverges a little later, but it does diverge
_NikJ: I wanted to try and modify my C constant in UCB1
_NikJ: Just for fun, see if it works any better
MSmits: when you backpropagate, is -1 a loss, 0 a draw and 1 a win?
MSmits: or do you use 0, 0.5 and 1?
_NikJ: 0-1
_NikJ: range
MSmits: then your exploration parameter should be around 0.5
_NikJ: 0.45 at the moment
MSmits: then it won't improve much
_NikJ: The "sweetest' spot I've found so far
MSmits: yeah that corresponds with 0.9 in my case and between 0.9 and 1.1 it plateaus
_NikJ: ah, you use -1,1 range?
MSmits: yes, i don't like fractions
MSmits: and it's symmetric
_NikJ: True, I like the symmetry part
MSmits: in practice it should hardly matter
TheBlueBias: Plateaus can go up according to french politicians...
MSmits: I don't listen to any politicians
MSmits: they only repeat what people with brains say or tell you what you want to hear. I listen to the brains :)
TheBlueBias: Well, not sure this one had any brains available...
MSmits: not necessary for his line of work
MSmits: not these days anyways
Fedor_Emelianenko: sup guys
MSmits: :point_up:
MSmits: that way
Fedor_Emelianenko: yes
Fedor_Emelianenko: )
MSmits: There is also :point_down: which is down
TheBlueBias: :thinking:
MSmits: I am filling in for jacek btw
TheBlueBias: Is it really required? :/
MSmits: probably not
TheBlueBias: :3
jacek: :upside_down:
TheBlueBias: (╯°□°)╯︵ :)
mask-007: Which interface do you play games on?
MSmits: i don't understand the question
MSmits: my bots play games and use input and output. I don't play games
MSmits: my motto is: Why a play a game when you can code a bot to do it.
_NikJ: Aww, it's a coding escape? Now I'm sad
TheBlueBias: Welcome in our reality _NikJ... :(
Einwickler: Wow, I was gone for over a year and yet it's the same usual suspects in here. Nice to see! :grin:
TheBlueBias: People are strange...
Astrobytes: hello Jim
Einwickler: Haha reference to my avatar intended?
TheBlueBias: :door:
TheBlueBias: +s
Einwickler: :blush::ok_hand:
TheBlueBias: Hello Franck o/
MaxTi: o/
TheBlueBias: \o
jacek: smits, have you checked out nine men's morris
TheBlueBias: eulerscheZahl are you here?:persevere:
TheBlueBias: On ICPC do you consider "temporality"? I mean when a flow is sent on the network, is it on an edge, then on the next one, etc. Or Everywhere on its path ?
gavi: Hi! I cannot find anymore a competition in which there are 3 monsters in a square grid. each has to reach one border of the grid. They can move or place an obstacle. What puzzle is that, please?
TheBlueBias: The Great Escape.
gavi: yes, that one. Thank you, @TheBlueBias
TheBlueBias: :thumbsup:
jacek: :notebook: :soccer:
TKZI_Knowledge_Commander: I'm keep craching my PC just by beefup the speed to max speed
jacek: hm?
MSmits: struct here?
struct: yes
MSmits: this works right:
MSmits: http://chat.codingame.com/pastebin/cc2f3f68-e767-4a35-957f-5cb27f96d6b4
MSmits: i got this from you a while ago
MSmits: adapted it a bit, but still your function
MSmits: should be called random move btw
struct: Should work
struct: its similar
MSmits: kk, i am changing my bot not to save the state on the node
MSmits: also implementing TT
struct: they should help
struct: I just dont know how to implement it with MCTS
MSmits: you store the childindex and childcount in your hashtable
MSmits: so when you expand, you assign the index and count you already have
MSmits: parents will share the same set of children
eulerscheZahl: TheBlueBias 05:00PM eulerscheZahl are you here?:persevere:
now I am. But not sure about codeforces rules and if I'm allowed to answer anything beyond understanding the statement
MSmits: it messes with mcts statistics, so it's not perfect
struct: I think I understand the rules
eulerscheZahl: i understand them
struct: its always a bit cryptic
eulerscheZahl: at the last Huawei marathon it took me more than 1 hour to understand the task
eulerscheZahl: this one was a bit easier. the only tricky thing was:
eulerscheZahl: It's guaranteed that only multiple edges may share the same GroupID.
MSmits: was it in Chinese euler?
eulerscheZahl: no, English only. Not even a Russian version available
struct: i always find it easier to read the inputs
MSmits: ah ok
struct: and understand the problem from there
eulerscheZahl: "multiple edges" is multi-edges. 1 reserved expression
eulerscheZahl: https://en.wikipedia.org/wiki/Multiple_edges
eulerscheZahl: but now I'm confident that I understood everything. Last time I even went back to reading the statement on the last day
MSmits: that happens in dots and boxes
MSmits: if you make a graph
eulerscheZahl: 1st Smits axiom: no matter what you are talking about, it can detail to dots & boxes
eulerscheZahl: oh nice, i'm 3rd \o/
struct: wow
eulerscheZahl: still more than 50% to go
jacek: this is how you prove NP-complete problems: to make them into other NP-complete problems
eulerscheZahl: yes
jacek: in this case its smits-complete, when it is convertible to d&b
eulerscheZahl: :D
MSmits: you guys are well versed in smits-theory
TheBlueBias: So it's just about comprehension, everything's in the statement,
TheBlueBias: ?
struct: if i want to more test cases for cf, do I need to make my own?
eulerscheZahl: yes :(
eulerscheZahl: i still only have that 1 alibi testcase
TheBlueBias: Meh. Don't see anything about that in the statement...
TheBlueBias: So I'll just consider omnipresent flows I guess...
eulerscheZahl: then i didn't understand your question it seems. each flow is blocking the whole path
eulerscheZahl: i thought you were asking about my internal representation
jrke: eulerscheZalh nice rank in ICPC!
TheBlueBias: No of course. I was asking if the flows "moves" on the network.
TheBlueBias: So no. Thanks :)
eulerscheZahl: thanks jrke. but still way too early. tourist is also lurking. last time he was way behind and then submitted straight to 2nd place
jrke: i am also looking in contest currently just trying for few flows to reach target
MACKEYTH: ICPC?
TheBlueBias: https://codeforces.com/icpcmoscow
jacek: pspspsps
_NikJ: Ah, Gennady is insane
jacek: hm?
TimothyAlexisVass: Can I do this using map instead with Python? list=[int(i,16) for i in list] list=[*map(int...?
TheBlueBias: Did tou test it?
TimothyAlexisVass: map only take 2 arguments...
secantsquared: You can do that, just make sure to pass a lambda or reference to a defined function first, then the iterable that is being mapped over... and the * will then unpack the map object into a list
jacek: :snake:
TimothyAlexisVass: OK
TheBlueBias: :eagle:
TimothyAlexisVass: That's what I thought. Thank you :)
Uljahn: guess you can create a partial function
Hyperplane: http://chat.codingame.com/pastebin/0136d782-195e-4974-8bd5-a5c7c9737e9e
Hyperplane: https://youtu.be/dQw4w9WgXcQ
Manjius: :angry:
VeljkoZ: Hi guys
VeljkoZ: What is u exp of finding job here
AguliRojo: I don't expect to find job using this or any other site.
AguliRojo: Experience matters more, and that can be achieved by projects.
Manjius: u need both
Manjius: or only this
Manjius: tbh projects is the least important
Dren: that depends a lot on the job
Manjius: yeah true
Manjius: im here because i failed technicals after my projects got me interviews :joy:
Aderner: im preparing to technicals too :D
Manjius: good luck
Thorcode: hi
Hyperplane: pls subacribe my yt i make code
Wontonimo: http://chat.codingame.com/pastebin/f22943d8-59ff-4d9f-bc28-5ed4e48fc142
Wontonimo: ^^ for Hyperplane
Hyperplane: :joy:
Thorcode: nice ascill
Wontonimo: hey Thorcode :wave:
Thorcode: hi
Thorcode: could you help me to understand this DEFIBRILLATORS easy ?
Wontonimo: sure
Wontonimo: what do you want help with?
Thorcode: the distance what does it do?
Thorcode: sometimes some puzzle make my brain boom
Wontonimo: you need to calculate the distance between the user location and the defibrillators
Wontonimo: then return the closest defibrillator
Wontonimo: here is my defib class in python for it
Wontonimo: class Defib: http://chat.codingame.com/pastebin/c72f7629-46c1-419f-a2a0-5ec961556f1f
Wontonimo: the equation for the calculation is in the problem description
Wontonimo: it is just a literal implementation of it
Wontonimo: the 'toRadians' function reads in the French formatted degrees,
Wontonimo: replaces the "," with a "."
Wontonimo: converts that into a float
Wontonimo: then converts that into radians
Thorcode: oh ok
Wontonimo: http://chat.codingame.com/pastebin/a1c21989-c996-4468-b95b-4c1b9f04c237
Wontonimo: ^^ pseudo code for the problem
Thorcode: thanks Wontonimo
zakacat: How do I find the problem of the week>
zakacat: I saw it before I left, and now I wanna try it
5922: You can probably find it in your notifications in the top right
Wontonimo: 5922 check your notifications
Wontonimo: oh, sorry, i was multitasking. didn't realize zakacat asked the question and 5922 answered it already lol
Wontonimo: a huge hint is the name of the puzzle
Wontonimo: good night all
Wontonimo: :wave:
Thorcode: Good night to you
Thorcode: hi