I kind of favor a setup involving surface convergence that is distinct and identifiable enough to forecast with some confidence at least several hours out, but not so sharp as to force the linear mode. Often, in the potential chase area I keep an eye on (eg. eastern and central KS and NE), this will take the form of a pre-frontal trough. Occassionally, the dry line will punch this far, but that is almost invariably accompanied by some other synoptic process such as a strong surface low.
Since visibility is very important to my enjoyment of a chase, the haze and general cloud cover close to a warm front aren't my ideal. (This isn't always the case, though: as I recall the Audubon, IA chase was along a warm front, but underneath a fairly strong cap, so the sky was nice and blue and the supercell top distinctly visible from a great distance.)
Outflow boundaries are cool, but - at least for my level of experience - kind of a wild card, not a feature I would plan a whole chase day around. Likewise, gravity waves fascinate me, and I hope to learn more about theoretical implications of gravity waves for tornadogenesis. (The Jarrell, TX case is an intriguing example.)
Aside from the surface setup, I found the following paper a very good primer for keeping the severe parameters (CAPE, etc.) in perspective:
A Baseline Climatology of Sounding-Derived Supercell and Tornado Forecast Parameters by Erik Rasmussen and David Blanchard
http://www.wdtb.noaa.gov/courses/dloc/svrp...on%20parameter'
It's a little dated, but the population size in their study I think is large enough to be relevant. In particular, the importance of low LCL's for tornadic storms seems to be greater than I had assumed. Also, the false alarm rate for even the best predictive parameters is very high, which reminds once again how much we don't know.