Halite Home

Published Bot Source Code


In order of current non-final scoreboard ranking

- #1 mzotkiew C# Writeup here
- #2 shummie Python Writeup here
- #3 Erdman Python Writeup here
- #5 cdurbin Clojure Writeup here
- #7 DexGroves Python Writeup here, with gifs
- #8 Nmalaguti Kotlin
- #9 moonbirth c++
- #10 Ewirkerman Python
- #11 Veeden Java
- #12 KalraA Machine Learning
- #14 jstaker7 Python Writeup here
- #15 fohristiwhirl Go
- #19 djma Java. #1 internal competition. Writeup here
- #21 Maximophone Machine Learning
- #23 En3rG Python
- #24 yangle C++

- #36 breeko Python
- #38 DanielVF Ruby Tutorial/Writeup
- #52 Sydriax Python
- #55 0x0L Python
- #82 jheilema-nerdery Ruby Writeup here



Mine is also here. Looking forward to learning everyone's secrets!

Edit: Writeup, general rambling and animated pictures here: http://dexgroves.github.io/halite/2017/02/15/halite.html


My Bot is up as well.


My C++ bot is here.


You can add mine.

Code here (Warning. It's ugly):

Writeup here:


I wanted to publish this tool for anyone continuing to develop a bot or interested in analyzing finals replays. You might find it useful to pinpoint problem areas. Will be releasing my bot source and writeup as soon as possible.



Awesome tool though. I thought I had no bad merges, but I saw a few with this that I didn't realize I had. Kudos!

I'm working on a pull request, but a zero strength move is just a standard investment move if you moving to a zero strength neutral with no enemies nearby.



My C++ Bot :slight_smile:


Writeup and bot code:



When @erdman said his bot was 100 loc i had a sneaking suspicion the catch was each line would be 5000 characters long.

But it really is approx 100 loc excl logging and comments. Hats off to you my good sir :smiley:


My bot, basically a big convolution of a potential function plus some micro management:

Many thanks to the organizers, the game was real fun and you ran the competition like pros ! \o/


Congrats to the winners. Hopefully, some of the ML bots will publish too. I would be very interested to study the approach of @KalraA. Also, anyone that got reinforcement learning to work well (maybe @hmate9)?


originations lol


@arjunvis: I believe you had written a faster stand alone version of halite for Reinforcement Learning, do you mind sharing it now that it is over?


Curious to understand just how good the top players are, I ran erdman's latest version against my humble 51 final rank (albeit first ever python program). Of over 200 games I beat him 4 times and of these, 3 are mostly luck due to small maps. Only 1 was a genuine win on a 30x30 map. So there you have it these guys at the top really are on another planet compared to us lesser mortals.

p.s. Many thanks to organizers, it was fun to be part of this all the way from Cape Town, one of the best places to live in the world.


Here's mine. Writeup forthcoming.



@Maximophone sure- i did 2 things

a) just recompiled halite for my cpu with gcc -o3 -march=native . this made halite.exe run 3x (ish) faster than the stock halite.exe at the time (early jan)

b) rewrote the simulation (process_frame) function only , but in matlab , so not sure how useful it will be to you. Happy to send. For halite 2.0 I'll step aside for the expert c programmers, probably makes more sense for someone like that to optimise the simulation engine (and of course the organisers choose rules that can be simulated fast)


My bot is here. Will try to do a write up when i get a chance.

Edit: write up is here.


Got a change to write something and publish my not-so-nice codeMyBot and simple summary.


Here is my ML C# bot and tools: https://github.com/cbovar/Halite
I used this competition as an opportunity to use and improve my library ConvNetSharp

It only ranked #118 but might interest some of you :slight_smile: