Dicing on the Streett

Streett/Rabin games are an adequate model of strong fairness in reactive systems. We show here some results about their stochastic version. We extend the known lower bound in memory for the pure winning strategies of the Streett player to randomized strategies. We also propose algorithms computing the almost sure winning regions of both players in stochastic Streett/Rabin games. The Rabin algorithm also yields directly a pure memoryless almost-sure winning strategy.