Interceptions lie and good corners get avoided. The stats that actually measure coverage — and why defense is so hard to grade.
Published June 6, 2026 · NFL Analytics
Ask a casual fan to name the best cornerback in football and they will reach for interceptions. It is the one defensive-back number that shows up on the broadcast, and it is almost useless for ranking coverage. Interceptions are rare and noisy: a corner might play twelve hundred coverage snaps in a season and finish with three picks, two of which were tipped balls or quarterback mistakes that had little to do with how well he covered. A single great season of interceptions rarely repeats, because the underlying skill - taking the ball away - is buried under luck and opportunity.
The deeper problem is the one that breaks every counting stat for cover men: the best corners get thrown at less. Quarterbacks are not foolish. They identify the corner they cannot beat and they throw somewhere else. So a shutdown corner can post a quiet stat line - few targets, few completions allowed, few interceptions - not because nothing happened, but because offenses schemed away from him entirely. Low target volume is genuinely ambiguous: it can mean elite coverage, or it can mean the defense hid a weak player on an island where nobody bothered to test him. The raw numbers cannot tell those two stories apart.
Modern coverage evaluation leans on a handful of per-target and per-snap rates that survive the small-sample problem better than interceptions do. None is perfect, but together they paint a far sharper picture than the box score.
Treat the defender as if he were the opposing quarterback: compute the standard passer rating on every throw into his coverage. A low number means quarterbacks who challenged him got punished. It is intuitive and widely charted, but it inherits passer rating's quirks and can swing wildly on a handful of targets.
The defensive mirror of CPOE. A tracking model estimates how likely each throw into the defender's coverage was to be completed; the defender's number is actual completions allowed minus that expectation. Negative is good - he made catchable-looking throws fall incomplete.
Total yards surrendered divided by snaps spent in coverage, not just by targets. This rewards the corner who was rarely thrown at, because the snaps he covered without giving up a yard still count in the denominator.
Player tracking measures how much space a receiver created at the moment the ball arrived. A defender who consistently keeps separation low is doing his job even when no pass comes - exactly the signal counting stats miss.
The defensive CPOE idea is worth a formula, because it is the cleanest example of grading per opportunity rather than per outcome:
CPOE allowed = Completions allowed % − Expected completion %
If quarterbacks completed 55% of throws into a corner's coverage but the tracking model said an average defender would have allowed 64% on those same throws, his CPOE allowed is about −9% - strong coverage. The expected-completion model is the same family of tool that powers CPOE for quarterbacks; here it simply runs from the defense's side of the ball.
The cleanest way to see why volume and per-snap context matter is to put two players side by side who look identical on the stat that fans cite.
Imagine two cornerbacks who each finish a season with 3 interceptions. On the broadcast they are interchangeable. Look one layer deeper at illustrative, round numbers:
| Metric | Corner A | Corner B |
|---|---|---|
| Interceptions | 3 | 3 |
| Targets in coverage | ~60 | ~120 |
| Completion % allowed | ~50% | ~70% |
| Passer rating allowed | ~70 (low, good) | ~110 (high, bad) |
| Yards per coverage snap | Low | High |
Corner A was thrown at half as often and surrendered catches at a far lower clip when challenged - the profile of a player offenses avoided and could not beat when they tried. Corner B's three picks came amid a flood of completions; he generated turnovers partly because he was targeted constantly and gave up a lot. Same interception line, opposite coverage value. The per-target and per-snap numbers separate them; the box score never could.
Even the best per-target metrics run into a wall that pass-rush analytics shares: coverage does not happen in isolation. The number a defender earns on a throw is tangled up with two things he only partly controls.
Tight coverage is far easier when the quarterback is hurried into a throw in under two seconds, and far harder when he has a clean pocket for four. A corner playing behind a dominant front looks better than the same corner behind a line that never gets home. Coverage and rush are two halves of one outcome - this is the defensive echo of the blame-sharing problem in pass-protection analytics.
In a zone, a completion into a defender's area may be a linebacker's blown drop or a safety's late rotation, not the nearest corner's fault. In man, the assignment is cleaner but a defender can be wronged by a missed safety help over the top. Charting models must decide who was actually responsible - and they do not always agree.
Zone versus man is the crux. A play-by-play feed sees a completion and a nearest defender; it does not always see that the defender was passing the receiver off to a teammate who never arrived. Assigning a blown coverage to the right player requires knowing the call, and the call is rarely in the public data. This is why two analysts can grade the same broken play differently and both be defensible.
To get at responsibility, the industry leans on human charting. Services such as Pro Football Focus assign a coverage grade to every defender on every snap, watching the film to credit a step-perfect rep even when no ball came, and to assign blame on busts that a stat sheet would pin on the wrong man. Done well, this captures exactly what per-target rates miss: the quality of coverage independent of whether the quarterback chose to throw it.
The cost is subjectivity. A grade is a trained analyst's judgment, not a measurement. Different services using different graders can disagree about the same player, and the assignment of fault on a zone bust is precisely the kind of call where reasonable people differ. The honest way to use coverage grades is as one strong input among several - cross-checked against passer rating allowed, CPOE allowed, and tracking-based separation - rather than as a single source of truth.
Defensive-back box scores mislead because interceptions are rare and noisy and because the best cover corners get thrown at less - so a quiet line can mean elite coverage or a player schemed away from, and the raw number cannot say which. Better metrics grade per opportunity: passer rating allowed when targeted, completion percentage over expected allowed (the defensive mirror of CPOE), yards per coverage snap, and tracking-based separation. The genuinely hard part is untangling coverage from the pass rush and from scheme - a completion in zone may be a linebacker's bust, not the corner's - which is why charting grades like PFF's exist and why their subjectivity means no one number should be trusted alone. The illustrative two-corner example above (same 3 interceptions, opposite per-target profiles) uses round, hypothetical figures purely to show the principle, not real players.
Want the code behind these metrics? Work through the 45-chapter NFL analytics tutorial.
Browse tutorials Free tools