[thelyingthief]

there are formulae that determine, using your win rate and roi, the confidence which you can assume in placing bets. of course, each increase in confidence (90,95,99) requires a correspondingly larger sample size. these equations will tell you what that number of trials must be. i have those formulae somewhere, but i'm too lazy to dig them out. let me know if you can't obtain them elsewhere, and i'll look them up.

ask over in the probability link for the info, first; major contributors there usually have these things memorized.