Many Algorithms have been developed recently to fire an airbag in a crash situation. These include algorithms based on velocity, energy, power, etc. All of these algorithms are based on physical principles but manipulated to perform the task of prediction. They have one or more of the following problems: 1) fire on low MPH crashes, 2) Do Not fire on time for pole, offset, etc. crashes, 3) Do Not fire on time for high MPH crashes, 4) Fire on various types of rough road scenarios, 5) Do Not reduce the effects of noise or EMI, 6) Do Not operate as effectively on deployment scenarios outside of the initial training set, 7) Are Not robust to the concatenation of the events listed above. Therefore, an algorithm is needed that can satisfy the above list. Since prediction is necessary to achieve these goals, an algorithm based on prediction should provide better performance than those currently in use. This paper presents the outline of such an algorithm. The paper first begins with a discussion of crashes. Second, a solution based on detection theory is given. Following this is a discussion of the principles behind the prediction-based algorithm. Finally, a discussion of results and conclusions is presented.