CPOE EXPLAINED
(COMPLETION % OVER EXPECTED)
Two quarterbacks finish the season at 68% completion rate. One is being discussed as an MVP candidate. The other just got benched. How?
One of them threw 70% of his passes within 5 yards of the line of scrimmage to wide-open receivers in a dome. The other was throwing 15-yard outs in November rain with a mediocre line. Completion percentage treats them the same. CPOE doesn't.
// WHAT CPOE MEASURES
CPOE (Completion Percentage Over Expected) is the difference between a quarterback's actual completion rate and what a model predicts he should complete, given the difficulty of his throws.
CPOE = Actual Completion % − Expected Completion %
The expected completion rate for each throw is calculated using:
- Air yards — how far the ball traveled downfield (longer = harder)
- Receiver separation — how much space the target had from the nearest defender
- Pressure — whether the QB was hurried, hit, or sacked on the play
- Pass location — sideline throws are harder than middle-of-field throws
- Scramble — whether the QB had to leave the pocket
The result: a QB who consistently throws into tight windows, deep, or under pressure gets credit for that difficulty. A QB who dinks and dunks to open receivers doesn't inflate the stat.
// THE CPOE SCALE
// THE PROBLEM WITH RAW COMPLETION %
Completion percentage rewards conservative play-calling and punishes aggressive offenses. A coordinator who calls short, high-percentage routes inflates his QB's raw completion stats. A coordinator who attacks downfield — even successfully — deflates them. CPOE normalizes for this by modeling what percentage of those specific throws an average NFL QB would complete.
| SCENARIO | COMP % | AVG DEPTH | CPOE (APPROX.) |
|---|---|---|---|
| Checkdown-heavy offense, lots of screens | 72% | 4.2 yds | −1.5% |
| Balanced offense, average depth | 66% | 8.1 yds | +3.2% |
| Aggressive downfield attack | 58% | 13.4 yds | +2.8% |
The first QB looks best by raw completion %. The last two actually demonstrate superior accuracy once throw difficulty is factored in.
// CPOE LIMITATIONS
CPOE is a receiver skill problem. A QB who throws to a wide-open receiver looks worse by CPOE than one who throws to a covered receiver — even if both throws were identical. The model controls for separation at the catch point, but separation is partly a receiver skill, not just coverage.
It also doesn't capture pre-snap decisions: a QB who identifies the best matchup and throws to the open receiver isn't "easier" to complete than one who forces it to a covered target, even if the CPOE model treats it that way.
Use CPOE alongside EPA/play. A QB with elite CPOE and low EPA/play is being accurate but conservative — lots of completions, not enough value per attempt. A QB with elite EPA/play and average CPOE is making the right decisions even if some throws miss. The best QBs are elite at both.
// CPOE IN THE PLAYER CARD
The Player Card shows CPOE in the secondary metrics bar for all quarterbacks, alongside passer rating, time-to-throw, aggressiveness, and max air distance. The scatter plot positions each QB relative to the full league passing cohort for the season.
Look up any starting QB and check their CPOE against the league scatter. The secondary metrics bar shows CPOE, passer rating, time-to-throw, aggressiveness, and max air distance — five angles on the same question.
[OPEN PLAYER CARD →]