[GeeTwo]

Whatever seems most likely to (dis)prove a concept the most quickly and within budget.
Last year (Aerial Assist), our prototype "Woody" (and also, by Hobson's choice, the practice 'bot) had a 2"x4" lumber chassis and kicker and 2"x2" ball pickup arms; the pickup shaft bearings were an aluminum rod rotating in holes drilled through 2x2s (OK, we used a circular rasp to smooth it). The biggest issue came in the build of the competition robot. The group that worked the pickup arms worked from the floor, and the group that built the kicker worked from the top of the frame. Since the aluminum frame was only 1-1/4" tall (vice 3-1/2"), the pickup didn't put the ball where it needed to be on "Buzz", the aluminum competition model.
This year (Recycle Rush), the only wooden prototype was a lift frame that we set up just to find out how fast our motors went when under various loads. They agreed with the number-crunching that we needed to gear our chain down by a factor of about 2 (we finally geared down 32:15). Other than that, we went straight from the drawings or cardboard cutouts to aluminum this year; we never had a whole prototype robot, though we built two metal 'bots, Atlas and Peabody.