There are a few flaws with EPA. It assumes the goal of every play is to score points, which isn’t necessarily the case. It also is based on results of past plays, and so garbage time passing plays against soft defenses increases the apparent effectiveness of passing plays. (Passing is still way more efficient, even without that effect)

That doesn’t mean EPA needs to change, you just need to know the flaws and how to interpret.

The last time the Eagles didn’t have the best record in football was Week 18 of 2021. I have no clue how they keep winning these close nailbiters, but I’m loving it.

Total penalties are 3rd and defensive penalties are 2nd/3rd of the 5 years in the study, but somehow the refs are calling too many penalties?

I suspect the same is going to be true of the AFD penalties — it’s going to be a nothing burger if you look at it over a period of time. You can find data to support a lot of conclusions.

Is the 2022 data through week 9, or full season? (It looks like full season in the initial charts). If so, you are comparing different things - as we get later the season there are fewer penalties:

2002:

Week 1: 204
Week 5: 201
Week 9: 135 (166 scaled to 16 games)
Week 13: 153 (163 scaled to 16 games)
Week 17: 158

2022 full year average was 177 / 16 games.

Do the refs suck? Yes, it’s an impossible job and the NFL refuses to do the obvious thing and have a replay official in the booth that can call down in 15 second before the next play and say ‘that wasn’t a hold’ or ‘you missed a facemask’, and that’s dumb.

But let’s be real, they aren’t doing it to screw any teams, and they aren’t worse about it than prior years. They just don’t have enough people and/or the technology made available to them to see and call everything that goes on during the modern day game.

Have you tested all the inputs for if the normal distribution is reasonable? I would think for some of them it would be, but for others you might have a longer tail on the up side (the elite QBs are better than the bad QBs are bad) which might explain why your final scores seem to be skewed.