Time to talk about games again. Here’s a simple one: a computerised version of Twenty Questions. It claims to update itself and learn from what people tell it. I reckon it’s a third year AI programming problem, but a good one.
At home, I have been playing quite a bit of Age of Mythology. After reckoning it not much chop a while back, I have since definitely come 'round to it. There are a couple of reasons for this. First, the computer opponent was good enough to keep me coming back for a while, and then I tried online games against human opposition. This is particularly good because of the rating system that the online service (ESO) uses. This means that I play against people who are roughly my level, and keeps things interesting.
The rating system used at ESO is the Elo system, used extensively in the chess world. It’s zero-sum. I don’t know if this is a good idea or not. Researching the issue on the Web, I found a description of the more recent Glicko system, which isn’t zero-sum. There are lots of hairy formulas lower down, but the initial prose explains the motivation well.