For example, some are saying to throw out the Buffalo game. Why? That game happened. Buffalo is a top
Defense. Why throw it away?
Reasons, obvious ones at that, have been given. Any statistical analysis of points per game comparing Joe vs Lamar that doesn't acknowledge what a trainwreck Buffalo was in that game and that 20 of the Ravens points came on "drives" totally 20 yards is flawed to say the least.
Joe himself doesn't fare well at all in PPG if you compare his last 4 starts to his first 4 starts (or even adding the Cleveland game).
Way too much of the pro Joe statistical argument is based on that Buffalo game, the first game of the year, for honest and reasonable people to be comfortable with it.
That doesn't necessarily mean that the game should be completely thrown out, but there should be significant consideration to when that game happened, how / why they scored so much, and how much an outlier it is.
Speaking of throwing things out, you seem to like to throw out the best offensive plays in the Lamar games to make some point of how they do in the rest of the plays. Interesting approach.
But on the other side of the stat argument, the choice by a few to ignore the strength of opponent defenses in Lamar's starts is cringeworthy imo.
One other piece of data to consider in this whole debate is that points per game are down leaguewide in the span of Lamar's starts compared to Joe's. So if there was an ERA+ type of stat for points per game adjusted by each week's scoring level, that would favor Lamar more so than the standard approach.
Back to DVOA, I think it's a very good stat, but as I mentioned in a different thread, it's biggest limitation imo is that it doesn't account for injuries or how well a team is playing at a specific point. So you can catch a team at a very favorable time (like when they decide to start one of the worst QB's to start in modern football history) and that's not considered. Therefore you get a nice and not completely warranted 1 game's worth of DVOA boost in a season that isn't long enough and has too many key injuries for the assumption of everything evening out to exist.