Lab Test: Strength of Schedule
Strength of Schedule (SOS) is a commonly used but misunderstood predictor of fantasy success. It assumes, for example, that a team will put up gaudy passing numbers against the worst passing defenses, and the run will flop against strong rushing defenses. Loosely speaking, SOS assumes:
Average Offense Production + Average Defense Production = Game Outcome
It turns out that the NFL is a bit more complicated than this.
Hit or Miss?
In the first edition of the Stat Lab last season, I hinted at the problems of SOS using
Peyton Manning as an example. You saw that even the best passer in the league can’t or doesn’t exploit the worst passing defenses all the time. Do the same analysis across the league, and you find that SOS can be little better than a coin flip.
Margin of error +/- 5%
Using a random sample of games over the last 6 years, I put SOS to a different kind of test. This chart shows the success rate of two forms of Strength of Schedule – How often was it right? The first pair is SOS using the previous season’s data. This is what we normally have to go on, and the type of projection you’ll see this time of year. You can see that passing predictions are a right a scant 58% of the time, while rushing is barely better than a random guess at 52%.
The second is more interesting. This is SOS calculated using perfect information. This is like making a bet today that the Giants will win the Superbowl last February. You can see that even if you knew exactly how teams would perform over the season, SOS still only offers a success rate of about 68%. What’s going on?
The Weakness of Strength of Schedule
The biggest problem is strategy. SOS assumes that a team will do the same thing in every game, overlooking the fact that coaches get paid to adjust to particular situations. Coaches get paid to win games for their teams, not ours.
The other problem is imperfect information. SOS depends on last year’s performance, and overlooks how teams change from season-to-season. It turns out that in any given season, in any statistical category, the average team can change roughly 18% up or down. A few of the best and worst teams will become average, and vice versa, and SOS doesn’t account for this.
So Strength of Schedule doesn’t account for situational details, and how teams tend to change from season to season. But what if it did?