It would be very interesting to see a computer analysis of position strength in the discarded games and seeing, where possible, if the ratios of "likely win" and "likely loss" match the games that were finished.