While game theory is useful when developing nuclear strategy I don't think anyone takes MAD seriously: It isn't subgame perfect. Once the rockets are up, retaliation is no longer credible... because the purpose of the threat of retaliation is to discourage an attack in the first place.