Fine print:
1. This includes all regular- and post-season games.
2. You might be wondering about era/league effects. It’s easier to win if your defense allows 20 points in the 1961 AFL than it is to win if you defense allows 20 points in the 1974 NFL. Though I didn’t mention it above, I actually did attempt to account for this. What I did was to compute the “average quarterback’s expected wins” for a given category by looking at all games in that category in the same league within two years. For example, if I’m examining Joe Namath’s 1966 season, the expected winning percentage for category (1) games, e.g., is computed by looking at all category (1) AFL games from 1964 to 1968.
Thoughts:
1. Don’t forget that the line labeled “Peyton Manning” is really Peyton Manning, Marvin Harrison, Edgerrin James, Jeff Saturday, Tony Dungy/Jim Mora/Tom Moore, Mike Vanderjagt, and a cast of thousands. What it isn’t, though, is the Colts’ defense. Or at least not as much as Manning’s raw record is. More on this later.
I’ve said before that quarterbacks don’t win games (teams do), and I still believe that. But if people are going to talk about QB wins — and it looks like it’s unfortunately too late to put the lid back on that can of snakes — they may as well try to put them into context. This post is is an effort to do that. To oversimplify things a little (or maybe more), Peyton Manning’s record is the product of the efforts of 22 guys. This exercise attempts to narrow that down to 11.
2. I think this exercise has provided me with a new all-time favorite example of Simpson’s Paradox. Check out Daunte Culpepper and Trent Dilfer:
+-------------------------------------------------------------------+
| Points allowed |
+------------------+-------------+-----------+------------+------------+------------+
| QB | 10 or under | 11--15 | 16--20 | 21--25 | 26 or more |
+------------------+-------------+-----------+------------+------------+------------+
| Daunte Culpepper | 5-0 1.000 | 7-4 0.636 | 17-6 0.739 | 6-7 0.462 | 8-39 0.170 |
| Trent Dilfer | 30-5 0.857 | 9-7 0.563 | 13-8 0.619 | 9-12 0.429 | 2-24 0.077 |
+------------------+-------------+-----------+------------+------------+------------+
Culpepper has him in every category, but Dilfer (63-46) has a better overall record (Culpepper’s is 43-56) because their distribution of opportunities has been so different. Matt Hasselbeck and Bob Griese are another example:
+-------------------------------------------------------------------+
| Points allowed |
+-----------------+-------------+------------+-------------+------------+------------+
| QB | 10 or under | 11--15 | 16--20 | 21--25 | 26 or more |
+-----------------+-------------+------------+-------------+------------+------------+
| Bob Griese | 53-1 0.981 | 20-7 0.741 | 14-11 0.560 | 7-11 0.389 | 4-31 0.114 |
| Matt Hasselbeck | 25-0 1.000 | 6-2 0.750 | 12-5 0.706 | 11-8 0.579 | 8-35 0.186 |
+-----------------+-------------+------------+-------------+------------+------------+
And you can pair Marc Bulger with just about anyone. Bart Starr, for instance:
+-------------------------------------------------------------------+
| Points allowed |
+-------------+-------------+------------+-------------+-------------+------------+
| QB | 10 or under | 11--15 | 16--20 | 21--25 | 26 or more |
+-------------+-------------+------------+-------------+-------------+------------+
| Marc Bulger | 8-0 1.000 | 8-1 0.889 | 8-4 0.667 | 9-6 0.600 | 8-38 0.174 |
| Bart Starr | 51-3 0.944 | 17-5 0.773 | 20-10 0.667 | 10-15 0.400 | 5-25 0.167 |
+-------------+-------------+------------+-------------+-------------+------------+
3. Starr is barely above average in this analysis. In Part II, where we make a few more adjustments, he’ll actually come out as below average. What are we to make of this? Those hoping for a Starr-is-a-fraud rant at this point will be disappointed. Maybe it’s just because of the soft spot I have for Troy Aikman, who finds himself ranked similarly. Or maybe it’s because I know that points scored and points allowed are correlated. Bart Starr wasn’t playing defense, but he and the rest of his offensive teammates could have, and probably did have, an indirect impact on the number of points the defense allowed. How much of an impact? That’s tough to say, but that’s the case you have to make if you think Starr is an all-time great, or even, frankly, an all-time good. What this does, in my mind, is eliminate the argument that Starr was good at doing just enough for his team to win. He wasn’t any better at that than Marc Bulger or Tony Eason or Randall Cunningham.
Lest you think this is a canned stat-head rant, let me also call attention to Ken Anderson. If there’s one thing that all historical football stat nerds seem to agree on, it’s that Ken Anderson is underrated. But this analysis says otherwise, ranking him dead average.