Abstract: Five problems existing in the torpedo test and evaluation are discussed, and corresponding improvement methods are suggested as follows: 1) Using the statistical method of repeated failures for reference, the repeated unqualified items can be changed to a single statistics to eliminate the impact of the difference of test plans on the pass rate; 2) Based on the existing definite number test, the requirement and principle of the “adding test” are analyzed and determined to solve the inadequate and incomprehensive problems of a test; 3) According to the sufficiency requirement of sample size and the change condition of technical status, whether interrupt or retry the torpedo test after the technical status change is analyzed and determined to clarify the test procedure and avoid mistake; 4) The availability dependability capability(ADC) effectiveness evaluation model is employed to define the work reliability and the ability in sea trial, thus the problem that the work reliability and the ability are repeatedly judged to be unqualified can be solved; and 5) According to the definition of load reliability, the requirement of consistency before and after the torpedo load test, and the actual usage analysis of a torpedo, we clear the detecting depth and breadth after the test are clarified. The methods mentioned above may be applicable to compiling torpedo test outline and modifying the evaluation criteria.