I propose a formal model of regime transition in which the leader’s incentives depend on threats of international conflict. External threats are represented as a war game. In the first period, the leader decides on whether to change regime or not. In the second, he decides how to act in the international conflict. In the third period, the selectorate decides on whether or not re-select the leader. For each opponent regime type, I produce a Markov matrix of regime transition probabilities, that depend on exogenous institutional settings and endogenous probability of aggression. I use these matrices to study democratization and find that (i) limited democracies are more likely than dictatorships to democratize, (ii) the likelihood of democratization is higher when an autocratic regime faces a democracy, rather than any other regime type.