Starcraft AI blog

I’ve been exploring build orders in which zerg uses the extractor trick repeatedly to make its first two buildings at 10 supply before the overlord. I was looking for a build as finely tuned as the Styx opening, but for 3 hatcheries. Here is the best one I found. I call it the Phlegethon build, parallel to the Styx, and I compare it to a more standard 3 hatch ling build.

I inserted blank steps on the left and right to show how parallel the builds are. Phlegethon’s overlord before spawning pool gets the needed extra drones with minimal loss of time and pulls ahead in minerals compared to overlord after pool. It’s possible to make either 3 or 4 drones after the spawning pool; there is little difference in timings, but the 4 drone version is slightly superior in keeping down its larva count in the runout. The third hatchery starts slightly before the spawning pool finishes, and taking gas delays the first pair of zerglings by around 2 seconds. If for some reason you want zerglings right now instead of zergling speed later, I think it’s reasonable to insert as many as 5 pairs of zerglings before the extractor (but it delays speed a lot in return for getting zerglings a little earlier).

You could say that 4 pool is a one hatchery member of the Styx family. It ends up with 3 drones and produces zerglings constantly from 1 hatchery. Styx ends up with 7 drones and produces speed-upgraded zerglings constantly from 2 hatcheries starting from the 7th pair of zerglings. Phlegethon and 3 hatch ling both end up with 11 drones and produce constantly from 3 hatcheries. Each member of the family has one more hatchery than the last, so that it starts zergling production later but produces at a higher rate. All have a good drone count for constant production.

Now to compare them. Was my clever trickery worth it? The Styx timings are from the data for my post on timing the Styx build. The Phlegethon and 3 hatch ling timings are from single test runs with variants optimized the same way, no scouting and all hatcheries in the main.

The standard 3 hatch ling build is superior in every way that matters. Phlegethon gets its second hatchery faster, but is slower in every other respect; the margins are not large, but they are definite. I draw lessons: 1. Standard builds are standard for a reason. Somebody considered all these possibilities years ago. 2. If you’re aiming for a given drone count, get as many as possible up front. The early drone catches the minerals.

I’ll graph the production curves another time. The 3 hatchery builds produce zerglings 1.5 times as fast as Styx, so they will surpass it after a while despite starting later. Whether that is better or worse in the game will depend on what the opponent is doing. We can expect that later members of the family will have decreasing results as they take ever longer to catch up. Also an over-large zergling army gains in strength less than it gains in numbers, because zerglings don’t have range and the ones in back can’t fight.

So—can anyone find a better build that has the same goal of constant speedling production from 3 hatcheries, or is 3 hatch ling optimal? There is reason to suspect that the Styx build is optimal for its goal. I don’t know any better way to prove it than exhaustive search.