I'm probably biased by my high school experiences with science fairs, and getting knocked out in the regionals by someone who just wrote software to make a picture move around on a computer screen. (which by the posing of their problem, should've been categorized as 'biology').
... but I'm of the opinion that a good science fair project uses the scientific method, and isn't just the 'can I build something?' type project. By that definition, consider what the question is that you're trying to answer, and what your hypothesis and null hypothesis might be:
- Can a computer beat a teenage human player in (some game?)
- Will (x strategy) beat (y strategy) in (some game?)
- Can a computer distinguish between four authors (of some genre/time period/etc) using text analysis?
And don't forget that you'll need to consider not just the time to write the program, but to conduct the experiment as well ... hopefully, at least 30 times per group, which might make my first phrasing of the game idea problematic.
update: fixed typos.