Last week, there was a lot of hand-wringing and hair pulling about the merits of two different Quarterback measurements: Adjusted Net Passing Yards / Attempt (via Pro-Football-Reference) and the conventional Quarterback Rating. The flaws and plusses of each were fully discussed here, and I'll give you my thoughts later. What I'll do first is look how each stat leads to wins, warts and all.
Before I get started, I need to make a correction to my original ANPY/A post. When I first worked through those numbers, I used 10 yards as the multiplier for Touchdown passes (hover over the ANPY/A column on Manning's page, and you'll see it as 10 yards). However, Chase Stuart, one of the writers at P-F-R, back in October did research on that number, and determined it should be 20 yards, not 10. So I went back and corrected the original page with all the new numbers. It bumped up the Win % just a tad, just barely behind Drive Success Rate (0.04% lower).
So which one leads to more wins? I'll tell you after the jump...
Here are the Win Percentages, based on various time frames:
|Off Win %||Def Win %||Off/Def Win %|
|ANPY/A||QB Rating||ANPY/A||QB Rating||ANPY/A||QB Rating|
- The two stats are very similar, and very good, at predicting wins. However, ANPY/A beats QB Rating in every comparison, as little as it may be. Seeing 90% always will make me happy.
- We've focused our comparison of stats on the both Offense and Defense being above average, and ANPY/A beats QB Rating by about 2% in each of the time frames. In the 8 year time frame, that is about 22 more games won (out of ~1000 games).
- I've actually had the QB Rating stats for a while now, and was purposely not going to include them in the Finding the Winning Factors series, because it was so close to the ANPY/A numbers. Pretty much if one of them was good, the other would be as well, so there was no sense including them both.
Both stats are pretty good predictors of wins, with ANPY/A coming out slightly ahead. I tend to lean more towards ANPY/A being the stat of choice because it has come about from recent research, and it can be shown mathematically why each of the numbers are there. It also produces an output that is easily understood by anyone. QB Rating gives equal weight to four stats, which should not be given equal weight, and it caps, both positively and negatively, each of the stats.
Does QB Rating do a good job of measuring a QB's performance? Sure it does. There's just better ones out there.