Paul Graham's formula for estimating a startup's chances of success involves idea, product, execution, team, and a random number for luck. His company sold for 30 million, but it is considered a failure in many ways.