That women of lose more than expected when they play against men of equivalent rating is entirely consistent with regression towards the mean [1].
Because elo ratings contain error and men are on average better at chess [2], the average true value that elo ratings are trying to measure will be lower for women than men at equivalent elo ratings.
Because elo ratings contain error and men are on average better at chess [2], the average true value that elo ratings are trying to measure will be lower for women than men at equivalent elo ratings.
[1] http://en.wikipedia.org/wiki/Regression_toward_the_mean
[2] About 1 SD, http://rspb.royalsocietypublishing.org/content/277/1692/2269...