To be fair, they admit that their models don't work very well for a team like the Brewers. This is what the author said about the Brewers' projection of 85 wins:
Also, their individual team projections tend to be a lot closer to 500 than what usually happens. For instance, they had Boston on top of the AL East with 90 wins, but they also said there was just a 10% chance that the AL East winner would have only 90 wins.
They also predicted the Rays would be last with 74 wins. ZiPS mostly proves that computers can guess just as bad as people.