I'm hoping this attracts more attention to building Starcraft AIs to solve games. (http://www.starcraftai.com/wiki/Main_Page as a springboard) It's a unique problem with respect to not having a fixed number of unique game states; which grow exponentially over time. From anecdotal experience in watching technologies evolve in my lifetime, some of the greatest leaps in computing research (hw+sw) have been covariated with solving algorithmic problems, with the latest being the use of neural nets to beat the very best humans at Go. Excited to see if an official API would be written to allow greater programmatic control for hobbyists!
For those interested, here's how to get a simple bot running in 30 minutes or less. Click the Twitch link to see the end goal and see how good/not-so-good various bots are today.
Watch StarCraft Broodwar bots live on Twitch now (runs 24x7 showing newly uploaded bots, commentary on Sunday afternoons): https://www.twitch.tv/sscait
I dabbled with SC bots back in late 90's and early 2000's. The game isn't very open-ended. The "board", so-to-speak, is constrained by trade-offs such as resource gathering, "unit tiers", "unit counters", "current meta" and map designs that reduce the number of things you can do at any given minute.
So much so that I think the 300-400 APM we see consistently being pumped out by the top players isn't the human limit, it's the actual game limit. In which case I would consider the game solved.
