CPOE is completion percentage over expected: how often a quarterback completed passes compared to how often an average quarterback would have completed the same throws. Every attempt gets an expected completion probability from its depth, direction, down, distance and pressure; CPOE is the actual result minus that expectation, averaged across all attempts. A +5 means five extra completions per hundred throws.
Updated Sept. 16, 2026.
Through Week 1, Trevor Lawrence (Jacksonville Jaguars) leads all quarterbacks with 20+ attempts at +22.3 CPOE, the best of all 602 qualifying starts through the same week since 1999. Patrick Mahomes (Kansas City Chiefs) is last at -11.9.
| # | Quarterback | Team | CPOE | Att |
|---|---|---|---|---|
| 1 | Trevor Lawrence | Jacksonville Jaguars | +22.3 | 21 |
| 2 | Caleb Williams | Chicago Bears | +16.6 | 26 |
| 3 | Baker Mayfield | Tampa Bay Buccaneers | +14.0 | 27 |
| 4 | Jaxson Dart | New York Giants | +13.4 | 29 |
| 5 | Geno Smith | New York Jets | +12.8 | 23 |
| 6 | Deshaun Watson | Cleveland Browns | +10.3 | 22 |
| 7 | Josh Allen | Buffalo Bills | +8.1 | 28 |
| 8 | Dak Prescott | Dallas Cowboys | +7.7 | 30 |
| 9 | Bryce Young | Carolina Panthers | +6.6 | 33 |
| 10 | Joe Burrow | Cincinnati Bengals | +6.4 | 34 |
| 11 | Drake Maye | New England Patriots | +6.0 | 31 |
| 12 | Lamar Jackson | Baltimore Ravens | +4.8 | 24 |
| 13 | Drew Lock | Seattle Seahawks | +4.8 | 22 |
| 14 | Jacoby Brissett | Arizona Cardinals | +4.6 | 37 |
| 15 | C.J. Stroud | Houston Texans | +3.2 | 37 |
| 16 | Aaron Rodgers | Pittsburgh Steelers | +1.2 | 37 |
| 17 | Jalen Hurts | Philadelphia Eagles | +0.9 | 22 |
| 18 | Brock Purdy | San Francisco 49ers | +0.8 | 34 |
| 19 | Kirk Cousins | Las Vegas Raiders | +0.7 | 29 |
| 20 | Cam Ward | Tennessee Titans | +0.4 | 28 |
| 21 | Tyler Shough | New Orleans Saints | -1.2 | 52 |
| 22 | Jared Goff | Detroit Lions | -2.9 | 39 |
| 23 | Jayden Daniels | Washington Commanders | -3.4 | 29 |
| 24 | Matthew Stafford | Los Angeles Rams | -4.7 | 24 |
| 25 | Daniel Jones | Indianapolis Colts | -5.1 | 29 |
| 26 | Justin Herbert | Los Angeles Chargers | -6.2 | 27 |
| 27 | Bo Nix | Denver Broncos | -7.8 | 27 |
| 28 | Malik Willis | Miami Dolphins | -8.9 | 27 |
| 29 | Jordan Love | Green Bay Packers | -9.7 | 40 |
| 30 | Patrick Mahomes | Kansas City Chiefs | -11.9 | 25 |
Raw completion percentage rewards the quarterback who throws screens and punishes the one who throws deep. CPOE fixes that by pricing each attempt. A 3-yard checkdown might be an 85 percent throw, a 30-yard sideline shot a 35 percent throw. Completing the first earns almost nothing; completing the second earns a lot. Missing the first costs a lot.
The model behind the public version (nflfastR's) uses air yards, pass location, down, distance, field position, whether the quarterback was hit, and a few more inputs. It does not know the coverage, the separation or the drop, which is why CPOE is a quarterback stat that quietly includes the receivers.
CPOE is one of the two ingredients in the composite that predicts quarterback play best. Paired with EPA per dropback, it separates the accurate from the lucky: a passer with high EPA and negative CPOE is probably riding yards after the catch that will not keep coming.
Two quarterbacks both complete 68 percent of their throws. One averages 5 air yards per attempt and the model expected him to complete 70 percent: CPOE -2. The other averages 10 air yards and was expected to complete 62 percent: CPOE +6. The second is the far more accurate passer, and the box score hides it.
CPOE = (completed passes minus the sum of each attempt's expected completion probability) / attempts, expressed in percentage points. Expected completion probability comes from a model fit on every pass since 2006 using air yards, pass location, down, distance, yard line, and quarterback hit.
Over a full season, +3 or better is excellent, 0 is average, and -3 or worse is a real accuracy problem. The single-season high in the public data is around +9. Small samples swing wildly: a +12 after two games means little.
The public version was built by Ben Baldwin and Sebastian Carl for nflfastR in 2020, following the expected-completion work Next Gen Stats publishes as completion probability. The site uses the nflfastR model.
Yes, against the quarterback. A perfectly placed pass a receiver drops counts as an incompletion on a throw the model expected to be caught, so it lowers CPOE. Over a season drops mostly wash out; over a month they can move the number a point or two.
Category: Stats & Analytics. Part of the StickToTheModel NFL Encyclopedia.