This study examines methods of determining blast shelter deployments and of assessing their performance for a varity of nuclear attacks. The goal is not to seek a single optimal deployment, which generally requires making arabitrary assumptions on the nature and size of attack, thus overlooking the attacker’s freedom of choice after a blast shelter program has been deployed. Rather, the study seeks "stabilized" deployments which protect population almost as well as an optimal deployment, even though it is not truly optimal for any specified attack. The study examines the attackers freedom to vary force level, time of attack, attack objective, height of burst and targeting. A quite general and flexible computer model BLAST, based on generalized Lagrange multipliers, generates shelter deployments for the U.S and computers their effectiveness against attacks in which these factors are varied. In BLAST the nation is considered as a collection of cells two nautical miles square, providing a detailed analysis of the offense/defense interaction.