This essay explores the deterrence‐versus‐restraint dilemma in extended deterrence in the context of the Tripartite Crisis Game under incomplete information. This model was developed specifically to capture the mixed motives and contradictory impulses that oftentimes frame extended deterrence encounters. To focus the analysis and to gain tractability, we make specific assumptions about the utilities of the players: Challenger, Defender, and Protégé. Our most significant simplification concerns Defender's type. In particular, we assume that Defender, although not heavily invested in the issues in dispute, is known to prefer conflict to the breakup of its strategic relationship with Protégé. One important result concerns unequivocal commitments. We find that such commitments are efficacious but only when Protégé's threat to sever its relationship with Defender is highly credible. In the absence of this condition, a straddle (or mixed) strategy is optimal for Defender. A straddle strategy, which involves probabilistic support of Protégé, is the mechanism by which Defender attempts to resolve the deterrence‐versus‐restraint dilemma. Sometimes, the stratagem works and Challenger is deterred and Protégé is restrained. But a straddle strategy will always fail to deter determined Challengers, such as Germany in 1914, that prefer to fight rather than back down during a confrontation. It may even fail to deter hesitant Challengers with an aversion to conflict. We use these insights to explain and evaluate British policy in the runup to World War I.