I can understand why a team like this could have more variance, but why would it be any more likely to underperform Pythag than overpreform? They go hand in hand. Pythagorean assumes that variance is noise and eventually levels out, if it's a feature/bug of your team design, then you're more likely to underperform it. To put a name to the variance, a team that sacrifices outfield defense for some extra offense will be more likely to see variance in low scoring games where everyone's hitting poorly, or maybe when the wind is blowing in and the offensive advantage is neutralized and the defensive disadvantage magnified. We're talking in generalities here, but that's how it can manifest. I don’t think that’s right. Variance goes both ways. If it trends toward underperformance, that isn’t variance. Thus, the modeling... Edit: Are you saying this kind of line up would be more likely to win big and lose small, and thus pythag is a bad way to measure win prob? I half agree if that is what you are saying. I’m not convinced that kind of team would, in fact, be more likely to have that happen than the exact opposite asymmetry.