Apologies for the lack of posts lately. You see, I didn’t properly plan my life around the release of Starcraft 2, so instead of being able to spend the weeks after the release playing the game and blogging here, I had to go get married and go on my honeymoon. I know, I know, I need to get my priorities straight.
Anyway, before I launch back into the Science of Starcraft, I thought I would share this post about the Computer Science of Starcraft over at Twenty Sided. You may be good enough at Starcraft to regularly beat the AI, but I’m not. I’m just too slow. So I was interested to hear the results of this experiment, where Shamus set up a map with seven AI players and a human observer, and ran it repeatedly, tabulating the results to learn a little more about how the AI ticks. He found (somewhat unexpectedly) that the computer was by far most effective with Protoss and worst with Terrans. He also has some interesting observations about how the AI uses special abilities like the Templar’s psi-storm agains human players and computer players:
I’ve come to suspect that the AI cheats a bit and detects clusters of units which have been grouped by hotkey by human players. This is very naughty if it’s true.
Anyway, it’s an interesting look at the AI of the original Starcraft. The custom map is even available to download if you want to try the experiment yourself!
Tell him to make the same test in StarCraft II!
Well, the Terran AI was horrible. Any normal player would use bunkers and siege-mode tanks for base defense.
I’ve tried to change the AI from Insane to Custom and what I saw is more battles for expansion points, and the Terran AI wasn’t so bad as in the insane AI script.
There were many ways to improve AI even on bnet.