Defining “expectations” is challenging. I don’t have a perfect way, but I do have a simple one: use a linear regression based off of last year’s Pythagorean winning percentage to predict the number of games a team should be expected to win this year.2 I did just that, and the best-fit formula was:
Year N+1 Wins = 4.23 + 0.472 * Year N Wins
So a 3-win team should be expected to win 5.6 games in Year N+1, a 10-win team is projected at 9.0 wins, and a 13-win team drops down to 10.4 expected wins. If you subtract the number of expected wins from the number of actual wins by the coach in a season, you are left with his number of wins over expectation. You’ll see pretty quickly why this is called the Dungy Index: he fares very, very well in it.
There are many drawbacks to this system. I’ve listed some of them at the end of the post. But as a first pass, this feels like an improvement over looking at raw wins or winning percentage. The table below shows all head coaches who coached 50+ games in the last 80 years. It is sorted by the “Wins over Exp” column. Dungy ranks 3rd on the list, and here is how his line reads: He first was a head coach in 1996 and last in 2008. He coached the Bucs and then the Colts. Dungy coached 208 games, winning 139 of them, for a 0.668 winning percentage. Dungy’s teams won 26.4 more games than you would have expected based on the regression formula. And since someone in the comments would have asked to see a list of wins over expectation per season, I added that to the end of the table, too. The table is fully sortable and searchable, and you can read some fine print at this footnote.3
We can take a closer look at Dungy’s career. While he did well in Tampa Bay, and he receives a significant amount of credit for his work with the Buccaneers in 1997, the majority of Dungy’s value comes from keeping the Colts among the elite teams in the league for seven straight years.
As a reminder, this is just another tool to examine head coaches, not the tool to measure them. There are many flaws with this study: I’ll start with six, but I’m sure you guys can think of some more.
- I should probably use a different regression formula for different eras. Presumably, team winning percentages are “less sticky” now than they used to be, which would imply that a regression formula for the ’70s would differ from what we would use now. It’s worth keeping in mind that doing this would only serve to make Dungy look better, I imagine.
- I should use a different exponent to come up with Pythagorean records for different eras. I used 2.5 as the exponent for all years because I’m lazy.
- I excluded all seasons when a coach was taking over an expansion team. Since different expansion teams were given different benefits, I decided to punt on this issue instead of thinking about a thoughtful answer.
- I created career rankings by summing the values for each season. That’s probably not the best way to do things, although again, a method that focuses more on peak production will likely only serve to benefit Dungy.
- “Wins over projected wins” is a flawed way to measure coaches. It should also go without saying that projecting team wins based solely on Pythagorean record in Year N-1 is not a perfect way to measure projected wins, even if that wasn’t a flawed method to measure a coach.
- Playoff wins are obviously way more important than everything else, and are currently ignored in this system. But in case you forgot, I already looked at playoff wins over expectations when I used the Vegas lines (or the SRS) to come up with expectations. Dungy graded out as below average in that system.
So if I have all of these caveats, why am I posting this? One answer is that I’m a good enougher. Another answer: most of the things I mentioned are easy enough to fix the next time around and won’t have a big impact on the results (but would have taken some time to implement). So before I spend a lot of time doing fixing those problems, I’d be curious to hear your general thoughts on this system, and any other tweaks you think I should consider implementing.
- Admittedly, this looks less impressive when you consider that Jim Mora, Jim Caldwell, and John Fox have won 13+ games with Peyton Manning, too. [↩]
- All ties are counted as half-wins. [↩]
- A. All coaches were included, but only their seasons since 1933 were counted. So George Halas’ line does not represent his career numbers, but instead his production since 1933. Why 1933? Because 1932 is a somewhat useful cut-off point, which makes it the first Year N-1 in the sample.
B. I did not count any games where a coach was in charge of an expansion team. So Dom Capers’ line does not show his career line, either.
C. I counted a tie as half a win. Don Shula did not win 331 games; he won 328 games and tied in six other games, but I am using 331 as a shorthand.
D. For coaches who were fired or hired in mid-season, I used their actual number of wins but subtracted the pro-rated number of expected wins. [↩]