The complexity of stochastic games
Information and Computation
Model Checking Probabilistic Pushdown Automata
LICS '04 Proceedings of the 19th Annual IEEE Symposium on Logic in Computer Science
Reachability in recursive Markov decision processes
Information and Computation
Solving simple stochastic tail games
SODA '10 Proceedings of the twenty-first annual ACM-SIAM symposium on Discrete Algorithms
One-counter Markov decision processes
SODA '10 Proceedings of the twenty-first annual ACM-SIAM symposium on Discrete Algorithms
Qualitative reachability in stochastic BPA games
Information and Computation
Approximating the termination value of one-counter MDPS and stochastic games
ICALP'11 Proceedings of the 38th international conference on Automata, languages and programming - Volume Part II
Recursive markov chains, stochastic grammars, and monotone systems of nonlinear equations
STACS'05 Proceedings of the 22nd annual conference on Theoretical Aspects of Computer Science
Hi-index | 5.23 |
We consider perfect-information reachability stochastic games for 2 players on countable graphs. Such a game is strongly determined if, whenever we fix an inequality ~@?{,=} and a threshold p, either Player Max has a strategy which forces the value of the game to satisfy ~p against any strategy of Player Min, or Min has a strategy which forces the opposite against any strategy of Max. One of our results shows that whenever one of the players has an optimal strategy in every state of a game, then this game is strongly determined. This significantly generalises, e.g., recent results on finitely-branching reachability games. For strong determinacy, our methods are substantially different, based on which player has the optimal strategy, because the roles of the players are not symmetric. We also do not restrict the branching of the games, and where we provide an extension of results for finitely-branching games, we had to overcome significant complications and employ new methods as well. The other result is finding a subclass of stochastic games where Player Max has an optimal strategy in each state. The subclass is defined by the property that if v is an accumulation point of the set of all values of a game then v=0. These results complement recent results classifying the existence of an optimal strategy for Player Min, and our general strong-determinacy theorem applies here as well. We also apply our results for Max in the context of recently studied One-Counter stochastic games. This work extends a workshop version of this paper which appeared in GandALF 2011, in particular, we prove a conjecture raised in that paper for the class of all reachability games.