It's worth reading the Wikipedia entry on the following report and Feynman's findings. Anonymous polling of engineers showed they estimated a general probability of catastrophic disaster in a shuttle launch between 1% and 2%.
Both of those predictions turned out to be extraordinarily accurate.