Starcraft AI blog

OK, this is officially Too Much. Steamhammer hit a string of good luck and its elo went over 2300—it should be about as high as this version gets. After its first loss to Steamhammer, Krasi0’s opening learning is still struggling to find the answer (which is: go back to the default build and win almost all games). After many losses against Iron, Steamhammer scored back-to-back wins, which can only be called luck. The wins came with opposite opening styles. One game was 9 pool speed. Steamhammer transitioned to mutalisks, which fought well enough to stop serious drone loss from the vultures. Iron built mass turrets but was unable to cope on the open map, and ultralisks eventually put in the final word. The other was 3 hatch muta. In the second game, the mutalisks were early and numerous and cut down Iron where it stood. A bunker and a couple rings of turrets are not enough; this mainstream opening is a flat-out counter to Iron’s build. If zerg took a little more care in defense, Iron would stand no chance against the 3 hatchery mutalisks.

Steamhammer also scored two fluke wins over LetaBot versions when LetaBot crashed while far ahead. Skynet by Andrew Smith finally put an end to the incredible winning streak—with a dark templar rush! Ha ha! Skynet played it well, killing spotting overlords with a dragoon and taking full advantage of Steamhammer’s blindness to its build.