Assuming this is done via looking at the poker site's records, some people like me mess with these numbers. Right now, I'm up big at Party and Pacific, up a little at Empire, and down at Noble, Stars, and Prima (deposited but haven't played Paradise). Thus, I factor in 3/7 times as a winner, 3/7 as a loser, and 1/7 as a break even. This would add some level of percentage to the winning side because there are a smaller number of winners, and I am adding an equal number of players to both sides. I would guess that most winning players (at least significantly winning ones) play at multiple sites, while most losing (i.e. casual) players play at only one site. Thus, the percentage of winners at any given site is much greater than the actual overall percentage of winners.

This could be corrected by studying this based upon polling data, but then we'd show 70% winners and 30% losers.

Will