Figure 7From: Win percentage: a novel measure for assessing the suitability of machine classifiers for biological problemsRoot mean square error between exact and sampled win percentage. Box plots show the distribution of root mean square error as a function of the ratio between M and N. When the total sample size, M, is large relative to the number of feature sets to explore, N, the error approaches zero. The dashed line indicates the best-fit exponential function (Equation 2) that estimates the error given only M and N.Back to article page