I just thought of an interesting (and somewhat more "real-world") task for AI research: Can an AI outperform a human at play-calling in [American] football?
Yahoo's fantasy football software creates millions of machine-written summaries of their fantasy football matches after each week of play. They're several paragraphs long and point out interesting things about the particulars of the teams' performance.
I would imagine that given the "All-22" camera footage from a football game, it would be possible to construct a piece of software to monitor what was going on (visually identify players by name/number, determine who has the ball, etc.) and then convert that model to a stream of English text.
Making it nuanced and entertaining, on the other hand, would be a challenge. :)
Some of the coaches already do this. Their "card" includes plays that say "In situation X, the probabilities say to call play Y."
The problem is small sample size. You only have the plays as called in the games to rely on. You can't "replay" the game with a different set of plays and see what the result would be.
This was a problem even in Moneyball. It's just that On-Base Percentage has a really good correlation to runs and therefore wins.
Considering that plays are often called or changed by quarterbacks on the field right before the snap based on what the QB sees from the opposing team, I think it would be difficult to provide all that input to a computer AI.
A neural net would be able to read the opposing defense more quickly and accurately than any human, but it would not be a fair fight (the human doesn't get a bird's eye view of things!).
It just doesn't have as big of an impact though. Each team has a very limited set of plays it's actually able to call and execute. Maybe 300 max (as a guess). Many times much less than that. So there's really not much room for an AI to see huge improvements via play calling.
It's really all about execution on the field and the "feel" of the game.
