We use the idea of repeat measurement to determine the mean value and error of light curve solution parameters of contact binaries. Our simulation is realized by the Monte Carlo algorithm and Wilson-Devinney code. This method can obtain systematic and random errors simultaneously. Within our 48 models, the systematic errors are smaller than the random errors in most cases. According to the numerical calculations, it is found that the relative errors of photometric mass ratios are less than 1% for totally eclipsing contact binaries, while they are generally between 10% and 20% for partly eclipsing ones. The effect of third light on the errors of the photometric solution is also investigated. With a third light, these errors are close to 10% for totally eclipsing contact binaries. Especially, it is better to set the third light to zero in flux if that light is very faint (e.g., less than 1% contribution in luminosity), because such faint third light will bring big errors to the light curve solutions.