Last week, we introduced Completion Probability, a new metric to contextualize the difficulty of an NFL pass based on the dynamic factors of a specific play. We learned why not all passes should be treated equally and we discussed how Next Gen Stats player tracking technology can improve our perception of quarterback performance on a per-play level. In this article, we introduce two new metrics derived from Completion Probability on an aggregate level -- Expected Completion Percentage and Completion Percentage Above Expectation.
Expected Completion Percentage (xComp) gives an indication of the level of difficulty of a quarterback's throws. A high xComp indicates a quarterback's propensity to throw higher probability passes, while a low xComp tells us a quarterback has a tendency to take more chances. As discussed in our introduction to Completion Probability, the level of difficulty of a specific throw is determined by in-play features specific to each play. xComp is simply the average Completion Probability within a given set of passes.
Completion Percentage Above Expectation (xComp +/-) measures a quarterback's performance relative to the difficulty of their throws on an aggregate level. We can improve our understanding of the traditional football box score by comparing a quarterback's Actual Completion Percentage to their Expected Completion Percentage. In other words, how did a quarterback perform relative to expectations over the course of a season, game, or situation?
The Limitations of Completion Percentage (Comp Pct)
In each of the last two seasons, a quarterback has broken the single-season Completion Percentage record. Sam Bradford set a new record in 2016 (71.6%) only to be broken by Drew Brees one season later (72.0%). Can we assume Bradford and Brees are two of the most accurate passers in NFL history? Or does their propensity to target higher probability passes inflate their Comp Pct? Bradford averaged 6.8 air yards per pass attempt in his record-breaking season, 1.7 yards below the 3-year average (8.5 air yards per attempt). Brees averaged 6.3 air yards per pass attempt, 2.2 yards below average. Adding context to Comp Pct to account for pass difficulty tells us more about a quarterback than traditional box score statistics.
It's no coincidence Bradford and Brees both broke the Completion Percentage record in two of the NFL's most recent seasons. The league average Comp Pct has increased steadily since the 1970 merger due to factors such as schematic innovations directly impacting the passing game.
Comparing Comp Pct across seasons is a frivolous exercise and even comparing within a season has its flaws given the variance in throw difficulty between players. xComp +/- normalizes these differences and adjusts the traditional box score statistic accordingly.
Completion Percentage Above Expectation as a Measure of Performance
Completion Percentage Above Expectation (xComp +/-) aims to answer the elusive question, is a quarterback performing better or worse than expectations relative to the difficulty of their throws? xComp +/- allows us to compare all styles of quarterback play, through the same lens, to differentiate between the accuracy of quarterback in a conservative or quick-passing system to the accuracy of a gunslinger who takes shots downfield.
Brees and Wilson were opposites when it came to xComp, on average Brees attempted the easiest throws in the NFL (68.9%) while Wilson attempted the hardest throws (58.5%). Diving into Brees and Wilson's tendencies, it's easy to see why these two quarterbacks have drastically different expectations. By examining the aggregated values of the factors used in the Completion Probability model, we can help illustrate why the two quarterbacks were at such different ends on the spectrum.