Clearly you do not understand what Pyth. record means then. I will say that it's not meant to be predictive of future performance though. In a small data set (i.e., beginning of the season) it may be a very inaccurate method to determine "over" or "under" performing. However, over the course of 162 games Pyth record correlates very well to actual record most of the time. So much so that people can reasonably conclude that if a team wins more games or loses more games than their Pyth. record suggests, they "over" or "under" performed. Here's a study if you're interested or just bored. http://arxiv.org/PS_cache/math/pdf/0509/0509698v4.pdf