Knowledge of diet compositions is important in ecological research. There are many methods available and numerous aspects of diet composition. Here we used modelling to evaluate how well different diet analysis methods describe the true diet of fish, expressed in mass percentages. The methods studied were both basic methods (frequency of occurrence, dominance, numeric, mass, points) and composite indices (Index of Relative Importance, Comparative Feeding Index). Analyses were based on both averaged stomach content of individual fish and on pooled content from several fish. Prey preference, prey size, and evacuation rate influenced the performance of the diet analysis methods. The basic methods performed better than composite indices. Mass and points methods produced diet compositions most similar to the true diet and were also most robust, indicating that these methods should be used to describe energetic-nutritional sources of fish.