The open source your ant bot topic

[Memetix]

I'm a psychotherapist, not a programmer, and I'm not sure how best to share my code ... so I have attached it.
I have shared most of the concepts on the forums already, but I have written an overview on my website and I do divulge how I pulled the concepts together and a few of the tweaks I made along the way.

My bot peaked at skill level 93.76 (2nd place at the time).

[lama3oid]

http://lama.od.ua/ex/lama3oid_v18.zip
profile: http://aichallenge.org/profile.php?user=453
was maximum on #34 a couple of weeks ago. current version is much smarter, but my oponents also became much smarter, so hard to predict which place I'm going to take. I think it's gonna be somewhere around #50, but won't be surprised if I take higher or much lower position
The code is a little bit messy since it's my first C# program, but not too much.
p.s. some comments are in russian -- I wasn't going to post the code

[NmdSnprEnigma]

My bot (of the same name) ended the initial contest in the low 200s:


I took a diffusive approach and quickly learned that positive pressures (pushing) was a bad idea. Enemies (most of the time), enemy hills, and food exert negative pressures. I'd had explicit exploration forces at one point, but it kept getting me stuck in dead ends. Instead, I did bfs expansion with Nodes with my hills as seeds. Each node had a value of one less (attractive negative pressures) than its parent's. As a node had no more possible expansions, it would be pruned until it hit a node with children. These nodes persisted over turns. Each turn these nodes would expand outward with positive pressure, so highways would find the best path away from my hills and pressures would push towards the highways. Enemy hills could never be pruned and nodes could kidnap other nodes if it was a shorter path, so eventually I'd be guaranteed a shortest past from one of my hills, to every enemy hill. Ants would choose randomly between two equal adjacent values, so every branching path would distribute and even share to each path.

Combat was also done using a test pressure. For each enemy ant with my ants near it, remove the positive pressure and see which directions my adjacent ants would move (typically inward to attack). I permuted the orders in which my ants would move where their first choice of destination conflicted. I checked the target ants 5 possible ants and then aborted if my ants would get caught alone. This left me with the problem that my bot had no good way to determine when it was safe to a attack a single ant (because it considered enemy ants one at a time), so I hacked a heuristic together.

[Darhuuk]

[Manwe56]

Alea jacta es
You will find in the attached zip the code of my bot (src/main) the tests of my bot (src/test using gtest) plus an already built binary for windows (check binary)

I have been ranked 25th on top, so I hope been around the top 50. If you run my tests, you will notice 5 will test over 110...I changed my mind yesterday very late, and did a little correction I was not thinking it might have a great impact. It is not really the case, and I have now a more aggressiv bot, with less expansion...It was unwanted, but does not seems too bad after all looking at the first matchs

I will detail later a few interesting thing I coded for combat and food gathering, with illustrations etc... But for now I just release the code

Good luck everybody for the finals

[KonaEarth]

[cheeser]

Hello!

My rank was 9 before the start of the final. I hope to finish in top 30.
May be my bot will be one of the best written in C#

My bot has a bulky code with a lot of Russian comments, im sorry for this, I have translated only the key comments

From Russia with love
P.S. I'll drink vodka if I'm in the top 10! Good luck and have fun!

[BenJackson]

[xoposhiy]

[Nionita]

Hi,

If someone cares to see a Haskell entry, here is mine:

git@github.com:nionita/MyAnts.git

It is good, but not the best (position 105 or so in the end unofficial top). Not very well commented, but anyway...

Nicu