In the wake of Sunday’s disappointing effort against Indianapolis, several scribes were quick to the task to either (i) degrade Peyton Manning for a questionable postseason history, or (ii) defend Peyton Manning’s postseason history, likely with citations to a small sample size and a more difficult level of competition.
As in many discussions with statistics, the analyses with Manning mostly came down to simple averages or cherry-picked examples.
Ex 1: Peyton Manning career postseason QB rating 89.2 Brady? 88.0 You do the math.
Ex 2: Peyton Manning falls to 8-6 at home in postseason. No other starting QB in Super Bowl era has lost more than 3 home playoff games.
But using these types of arguments is both unimaginative and uninformative. There are better ways to consider Manning’s postseason performance level, right?
Well, one improved method for a player comparison is to use individual game statistics, as opposed to taking a simple average. So, going back to the start of their careers, I extracted Manning and Brady’s game-by-game passer rating, for both regular and postseason games.
A useful tool for the comparisons of the distributions of continuous variables is a density curve, which represents the smoothed curve line that we would fit over a histogram had we made a histogram of the distribution. But unlike histograms, which can be difficult to look at for comparisons, density curves make it nice and easy to compare centers, shapes, and spreads (if you want to read more on density curves in sports, check out War-on-Ice’s post here).
Here’s a graph of the density curves for Brady (blue) and Manning (red), split by game type (posteason, regular season).
The curves for Manning and Brady’s regular season performance are nearly identical. Manning’s curve is slightly to the right, implying that, overall, his game-by-game passer rating distribution tends to be a bit higher than Brady’s. Comparing the top graph to the bottom graph, both quarterbacks have notably higher centers of quarterback rating in the regular season. Of course, this isn’t surprising, given that the competition is more difficult in the playoffs.
In the postseason, the distribution of passer ratings for each QB is once again centered at roughly the same location (about 90), but Manning has a higher density in both tails. This implies that Manning has had more games that have been both really good (QB rating ~ 150) and really bad (QB rating ~ 25).
Perhaps such conclusion isn’t all that surprising. A few putrid games in the postseason from Manning can linger in people’s minds, particularly when his biggest adversary (Brady) has avoided them. Such a result could be a sign of several things; we’ll never know if the difference is due to sampling variability, s small sample size, or the chance that #18 is indeed more likely to put up a terrible game come January.
Hopefully this post encourages more looks at the distributions of player metrics (in fact, I’m working on a longer one as we speak). Also, I should point out that quarterback rating isn’t a great metric for evaluating performance, as, among other weaknesses, it is highly dependent on throwing touchdown passes and avoiding interceptions. However, passer rating was the only such game x game metric that I could get, dating back to the start of Manning’s career.