Are the displayed images the average output accuracy or the winners that happen to be accurate?