Starcraft has always been more like chess than a video game, on the multiplayer scene it’s all about learning strategies and unit build orders. So much like it happened to chess one enterprising coder took to the task of writing up a little program that using evolutionary algorithms (wrongly reported as genetic in the source article) to find the optimal building strategies on a trial and error basis. Much like in real evolution the program will generate offspring from the previous successful matches and keep branching and refining in it’s quest for the best strategies.

Although the program is still quite new, it has already come up with strategies much similar to those used by high level players, now it would be very interesting to see it come up with some original approaches to victory. On the other hand this only highlights what many find off-putting about competitive Starcraft, it has become a very mechanical endeavor where a single wrong move early on can make the outcome pretty obvious. How long until we see the top tier competitive players go against a highly developed fan-made AI? It’s Deep Blue all over again!

[lbrandy]