Completion Percentage Over Expected — CPOE — is one of the most trusted quarterback stats in modern football analytics, because it fixes the oldest problem with completion percentage: not all throws are equally hard. A quarterback completing 70% of screens and checkdowns isn’t doing what one completing 68% on downfield throws under pressure is doing. CPOE measures the difference between what a passer completed and what an average quarterback would have completed on the exact same throws.
The mechanism is a model built on tracking data. Every pass attempt gets an expected completion probability based on air distance, throw location, separation, pressure, down and distance, and more — a five-yard out to a wide-open receiver might be 92% likely, a contested 40-yard shot 28%. Sum the expected completions across a season, compare to actual completions, and the gap is CPOE: a +5% CPOE means the quarterback completed passes five percentage points more often than an average passer would have on his throw diet.
The chart below covers the formula, the benchmarks, and how CPOE pairs with other metrics. Take a look, then we’ll get into what it does and doesn’t tell you.
FOOTBALL ANALYTICS
Completion Percentage Over Expected (CPOE), Explained
Accuracy, adjusted for how hard the throws actually were
FORMULA
Actual – expected
In percentage points
SOURCE
Tracking data
Next Gen Stats, nflfastR
How CPOE is built
The model behind the number
| Component | Detail |
|---|
| Expected completion % | A model assigns every attempt a completion probability from tracking data: air yards, target location, receiver separation, pass rush, QB movement, down and distance |
| Actual completion % | The passer’s real completion rate on those attempts |
| CPOE | Actual minus expected, in percentage points — throw-difficulty-adjusted accuracy |
| Example | Expected 64.0% on a season’s throws; actual 68.5% = +4.5 CPOE |
| Sources | NFL Next Gen Stats (official tracking) and public play-by-play models (e.g., nflfastR); numbers differ slightly by model |
| Sample size | Stabilizes faster than most QB stats — meaningful signal within a few hundred attempts |
Reading the scale
Season benchmarks
| CPOE | What it means over a season |
|---|
| +5 and above | Elite accuracy — the league leaders live here; historic seasons push +8 |
| +2 to +5 | Clearly above average |
| -1 to +2 | Around league average |
| -3 to -1 | Below average |
| Below -3 | Accuracy problem the scheme can’t hide |
| Single games | Swing wildly (-15 to +20); read season-level only |
CPOE alongside other metrics
Where it fits
| Metric | What it measures | Best paired with CPOE because… |
|---|
| EPA per play | Value added per snap | CPOE is accuracy, EPA is production — the CPOE+EPA composite is one of the best QB summary stats public analytics has |
| Air yards / aDOT | Throw depth | Context for the difficulty diet — a +3 CPOE at 9.5 air yards beats +3 at 6.0 |
| Pressure rate | How often the QB is harassed | Separates accuracy under duress from clean-pocket accuracy |
| Passer rating / QBR | Composite outcomes | CPOE isolates the skill (ball placement) those composites blend with situation and receivers |
CPOE FACTS
It’s scheme-resistant
Because every throw carries its own expectation, a checkdown-heavy offense doesn’t inflate CPOE the way it inflates raw completion percentage.
It’s mostly the quarterback
Drops hurt CPOE (a drop is an incompletion), but across a season, ball placement dominates — CPOE is among the most stable year-to-year QB skill stats.
The composite is the headline
Analysts often cite ‘EPA + CPOE composite’ — a blend that has predicted QB success better than either stat alone.
What CPOE measures and why analysts trust it for quarterback accuracy.
What CPOE Misses
CPOE grades the throw, not the decision. A quarterback who never attempts tight-window passes can post a fine CPOE while leaving open receivers and big plays on the field — the stat has no opinion about the throws not taken. It also inherits its model’s blind spots: expected completion probabilities are estimates, drops are charged to the passer, and different models (Next Gen Stats vs. public play-by-play versions) produce different numbers for the same season. That’s why the standard practice is pairing it with EPA: CPOE says how well he throws it, EPA says how much it mattered.
The Bottom Line
CPOE compares a quarterback’s actual completion rate to what an average passer would have managed on the identical set of throws, making it the cleanest public measure of accuracy in football. Read it at the season level, treat +5 as elite and 0 as average, and pair it with EPA for the full picture. For that companion stat, see our guide to Expected Points Added, and for the down-to-down version of efficiency, success rate.
