Level of Difficulty: Advanced Undergraduate
You’re observing a parking lot of length L, initially empty (this is a one-dimensional parking lot, mind you). Cars start coming in sequentially, one after the other. Each car has width W, and parks randomly with equal probability somewhere in the space that’s available. Eventually the parking lot will fill up. The question is, when space runs out and the parking lot gets filled up, what percentage of the parking lot will be filled on average? I’ll call that the filling factor.
Note that the parking lot is “continuous”. There are no marked parking spaces - every car that parks does so with equal probability at each available point. Also, this is not a problem in psychology. Forget about the way humans behave when they park. The probabilities here are totally random (a.k.a. equiprobable). Here’s a little illustration to make things clearer:

Now, a word of warning. This puzzle is very difficult to solve exactly, so I’m not going to ask you to do so, and you shouldn’t be hard on yourself for not solving it, regardless of your skill level. Instead, I’m going to ask you to estimate the filling factor. For example, to get you started: can you bound the filling factor from above? From below? Be creative! I’m more interested in your reasoning than the actual answer (I already know it).
Good luck and have fun!
EDITED, 21/June/2009: the comments now contain the actual answer, links to a detailed solution (from a paper, in pdf) and further discussions in other sites. Still, the problem of estimating the result correctly remains open, so if you happen to stumble by this page do take a stab at it. Trying to approximate the solution is, I think, more interesting than deriving the actual solution itself!