archive by month
Skip to content

PurpleSwarm vs human show match

Two short posts today.

The human-machine show match from AIST S1 was streamed today, showing PurpleSwarm against a protoss.

On Sparkle, PurpleSwarm was able to take the crystal gas, unlike Steamhammer. Except for that, the two bots seemed to have similar skills and play styles on the map. One extra gas is a big difference, though.

2 games were on Third World. PurpleSwarm showed no sign that it knew how to mineral walk through the gates on the map. Steamhammer doesn’t know either. From past comments, it sounds like McRave is ignorant too, but Tscmoo has the skill. Does anyone know about MadMix? Just curious.

Trackbacks

No Trackbacks

Comments

Dan on :

Funnily enough, PW has automatically mineral walked between bases for a year now, but because it's relying on the BWTA region graph it doesn't think there's a path connecting the two halves of Third World so it doesn't try. Sadly that would've required replacing BWTA entirely and the time wasn't there -- most of the preparation time went towards just trying to beat (an old version of) McRave, which I only achieved a few hours before the deadline. He's a tough, tough opponent.

I brought three features to Sparkle to try gaining an edge:

First is the highly tuned 13 Pool build which hits with 6 Mutalisks at a very fast timing. Being able to do it with no defense means having the third Hatchery and gas (with Psi Disruptor removed) at around 6:45.

Second is moving shot for most eligible units, including Mutalisks. In particular, this is great at killing Corsairs which get out of position. See 27:50 in the VOD.

Third is proactively spreading units against splash damage, including Corsairs. See 30:45 where the Mutalisks crash through the Corsairs and spread out while shooting. They're trying to stay about one Corsair-splash-radius apart.

This build, naturally, is to accommodate Swarm's inability to expand on islands. Adding it was feasible in the AIST timeline, but without opponents to test and tune a macro build I thought my odds were better cheesing (of course, I love cheese and never need an excuse to employ it). I thought this the best solution for maximizing Mutalisks(time) for most values of time :).

ate850, of course, smartly avoided all of these with a proper build order (the cannons are a must to survive the Mutalisk timing, which I suspect most AIST participants omitted) and exercising saintly patience with his Corsair fleet.

This experience reinforced for me a big idea across BWAI history, which is that you can't beat what you can't test against. The main reason bots are stronger today is because their testing opponents are stronger too. Swarm had a Hive/Devourer/Guardian transition designed to handle large Corsair fleets and Photon Cannon turtling, but without any bot testing opponents who could produce a realistic Sparkle build, the transition was poorly tuned and arrived about 2 minutes late.

AIST1 and the showmatch were a lot of fun and good practice! Thanks again to Nathan for hosting.

Jay Scott on :

Oh yes, the moving shot was effective. Steamhammer will also get guardians and devourers, but with only 2 gas it doesn’t have a chance to keep up. Also great against corsairs on Sparkle is queen ensnare—that is the Sparkle skill I would have added next if I had solved the 3rd gas.

Jay Scott on :

“You can’t beat what you don’t play against” (to change the wording slightly) is true for humans too, across games. A strong community raises strong players. It’s true for Brood War in Korea. For me it’s even clearer in the somewhat separate go playing communities in China, Korea, and Japan. I think it’s the same: Bot authors learn from each other, directly and indirectly, in the same way that players learn from each other, and both rise in the same way to the challenge they face.

Gathers on :

MadMix can mineral walk now. It was actually easier to add than I had suspected. It probably helped that I never try to determine if there is a valid path to the destination or not, I just send a worker and hope for the best.

I actually started on 3 different ways of picking the correct mineral-patch to right-click, but couldn't make it work in all situations. So now it just randomly uses one of the 3 ways to mineral walk when it detects stuck workers. It might still fail in some situations but there wasn't time to find an always correct solution.

Add Comment

E-Mail addresses will not be displayed and will only be used for E-Mail notifications.

To prevent automated Bots from commentspamming, please enter the string you see in the image below in the appropriate input box. Your comment will only be submitted if the strings match. Please ensure that your browser supports and accepts cookies, or your comment cannot be verified correctly.
CAPTCHA

Form options

Submitted comments will be subject to moderation before being displayed.