Kind of cool video. At the same time, kind of dick-ish narration, lol. The guy from CHFF is right though, the passer rating disparity has like a RIDICULOUS correlation to winning. I know 75% may not seem that ridiculous but I am not sure there is any single metric anyone else has come up with (that isn't scoreboard related) that predicts wins at that high a rate. The closest I've heard of is passing yards per attempt disparity, which if you think about it is actually just really closely related to passer rating disparity.
Just goes to show at how much more is involved than a formula to success there is no formula to show a winner except maybe the score board. Wonder how much stock NFL coaches put into passer ratings if any at all.
The problem is that most of us try to correlate baseball stats, which are completely oriented to individuals with football which is almost completely dependent the team, not just the individual being compared... The passer rating, as complicated and convoluted as it is, is still better than ESPN's version of the same. The video was interesting though, and while it succinctly pointed out the disparities of the rating system, it is also the best thing we have to this point...
He's also big on defensive passer rating, which is basically just assigning the opposing QB's passer rating to the defense as a unit. He likely picked it up from Dom Capers, who uses it as a core statistic. Dick LeBeau as well. Not coincidental, IMO.
Well passer rating differential has a very high correlation to winning, but I think its important to point out that it is not necessarily predictive. It certainly is explanatory and retrodictive, but not predictive. Correlation does not equal causation. Keep in mind that passer rating is just an average of 4 statistics. Of those 4 statistics (completion %, TD rate, INT rate, Yds/Att.), yards per attempt is really the only one that is predictive. I believe that if you're looking for a statistic that is truly predictive of future wins, then net yards per attempt (include sack yardage) is going to have the highest correlation to future wins. Since yds/att is included in passer rating, it does have a predictive element to it, but I believe the noise and randomness from the other components has a detrimental effect on its predicative ability. ETA: I've also done some work with Success Rate for running plays, and that does appear to have some predictive qualities, but I haven't done enough to say with any certainty.