Metrics and models are meant to be predictive. The degree to which they are predictive is important when using them, and their predictive power can change over time. Some models also miss important variables.

For example, I think TJ Watt’s PFF score is always lower than it should be because a lot of his stuff is amazing but, in 95% of cases, isn’t consistent and so does not get factored in - remember their scores are supposed to predict future performance, and if TJ Watt is consistent with someone no one else is, then you just ignore his inappropriately low score.

The “outgained” thing is, generally, predictive, but I think Mike Tomlin may be the sort of consistently good at things that model poorly type of guy, like TJ, where you just ignore the normal modeling.