[rabbah]

I watched the youtube videos of all the battles so far and I was really surprised to see that a very simple strategy (in retrospect) was sufficient to win the competitions.

I've been waiting to see if someone implements an AI bot that's trained on all the bots in the competition but maybe you just don't need to.

[tout]

That's very much something that happens in Battlesnake[1]! There's a fun Go Time episode about it all[2].

[1]: https://play.battlesnake.com/

[2]: https://changelog.com/gotime/182

[ska]

In most of these sorts of competition environments, it's a lot less work (and cheaper) to make a competitive hand-tuned algorithm that a really competitive AI version. If you are already working on a similar training environment with a bunch of free/cheap compute jumping in for a weekend or whatever could give decent results, but otherwise you're likely going to get destroyed by bespoke tactics.

[judohacker]

Years ago, I played around with Robocode and genetic programming to "evolve" battle strategies. I was always surprised by how simple the generated winning strategies were. They were always counter intuitive.

[msciabarra]

Actually the discussion of using an AI trained robot has been frequently discussed on the various forums. Nimbella has even an "python ai" runtime that would support it. However does not seem the winning robots actually uses any A.I.