Completion percentage over expected (CPOE)

Completion percentage over expected measures how much better or worse a quarterback completes passes than a statistical model expects, given the difficulty of each individual throw. It is designed to credit accuracy on hard throws and avoid over-rewarding a high volume of easy, short completions.

The expected-completion model is trained on player-tracking data, incorporating factors such as throw distance and depth, the receiver's separation from the nearest defender at the time of the catch, and pressure on the quarterback, to estimate the league-average completion probability for a throw with those characteristics. CPOE is then the difference between a quarterback's actual completions and the sum of those expected probabilities across all their attempts, usually expressed in percentage points.

CPOE improves on raw completion percentage, which rewards quarterbacks who throw mostly short, low-risk passes and can undervalue one who attempts more difficult downfield throws, and it is used alongside expected points added as one of the more predictive quarterback evaluation stats available. The main pitfall is that CPOE quality depends entirely on the underlying tracking data and model; it is unavailable for leagues or eras without player-tracking systems, and different providers' models can produce somewhat different CPOE values for the same player.