• The weekly quarterback rankings utilize a technique called Bayesian updating, which most accurately combines PFF grading and advanced stats to forecast future performance.
Estimated Reading Time: 5 mins
This weekly quarterback ranking utilizes a statistical technique called Bayesian Updating to best project the level of quarterback play for each team’s Week 1 starter.
With Bayesian Updating, we can use the dynamics of the historical quarterback market and individual results to project their PFF grades and expected points added (EPA) per play. This begins with projections based on NFL draft position, then updates those predictions after every play in which the quarterback is involved. The longer history a quarterback has of strong play, the higher his expected play going forward. It’s a simple concept, but one that uses advanced statistical techniques to give the most accurate numbers.
MIXING PFF GRADES AND ADVANCED STATS
PFF has found the most success forecasting quarterback performance using an equal mix of grades and the advanced efficiency metric EPA per play. I chose to frame the estimates below in terms of standard deviations above or below the expectation of a typical NFL franchise quarterback. The NFL is going through a golden age of quarterback quality at this time, with most of the Week 1 starters having above-average forecasts for PFF grading and EPA per play.
In the past, I had heavily decayed past results so that most recent performance was more important in the estimation of future performance, but I found that accuracy increases by focusing only on the past five years. Therefore, these numbers are based on the 2017-2021 period.