For USBL (Ultra short Baseline), the positioning precision is a significant technical index. How to use the data gained in field trail to evaluate the precision has become an important problem that the USBL users pay more attention to. This paper used error ellipses theory to evaluate the USBL positioning precision, and after analyzing and computing, verified the conclusion that the error ellipse is equivalent to the statistical results corresponding to the 39.35% of the positioning points and then established a connection between the theoretical error model and the actual statistical results. Therefore, with the correct prior errors information, the USBL positioning precision evaluation can be performed with one of the USBL error ellipses. When the actual data is polluted by system errors, USBL theoretical error ellipse is not consistent with the statistical results of the scatter points. On this occasion, the error ellipse can be a criterion to evaluate system errors. In the last, according to the analysis for the results, this paper made a summary about the method used to evaluate USBL positioning precision combining error ellipse theory and actual positioning scatter points. That is, using the statistical value of the 39.95% of the positioning scatter points as the system precision and the theoretical error ellipse to evaluate the system error. Application has been achieved in the field trail to process data.