Traditional football stats — passing yards, rushing yards, touchdowns, wins — tell you what happened. Advanced metrics tell you why it happened and whether it is likely to continue. That distinction matters enormously in betting because the sportsbook’s line reflects the market’s aggregate view of team quality, and beating the market requires a more nuanced understanding of performance than box-score numbers can provide.
The advanced metrics revolution in the NFL has produced a toolkit that sophisticated bettors use to build power ratings, project matchups, and identify value. The most impactful of these metrics — DVOA, EPA, and CPOE — each measure a different dimension of team and player performance, and together they form a framework for analysis that goes far deeper than “this team averages 28 points per game.” This guide explains what each metric measures, how it is calculated, and how to apply it to your betting process.
DVOA: Defense-Adjusted Value Statistics
DVOA, developed by Football Outsiders, stands for Defense-adjusted Value Over Average. It measures a team’s efficiency on every single play, compared to a league-average baseline, with adjustments for the quality of the opponent faced. The result is a percentage that tells you how much better or worse a team is than average on a per-play basis.
A team with an offensive DVOA of +15% is performing 15% better than the league average after adjusting for opponent quality. A defensive DVOA of -12% means the defense is 12% better than average (negative is good for defense in DVOA’s framework because it represents holding opponents below average). A team with a total DVOA of +25% is well above average on both sides of the ball and is likely one of the best teams in the league.
The opponent adjustment is what makes DVOA more valuable than raw efficiency stats. If a team posts 400 passing yards against the league’s worst pass defense, raw stats treat that the same as 400 yards against the league’s best. DVOA weights each play based on the defense it was achieved against, so the 400-yard performance against a bad defense is valued less than the same performance against a good one. This adjustment produces a cleaner signal of true team quality, which is exactly what bettors need when projecting future matchups.
For betting purposes, DVOA’s primary application is building power ratings. Rank all 32 teams by total DVOA, convert the DVOA gap between two teams into a projected point spread (the conversion is roughly 1 point per 3.5% of DVOA difference, though this varies), and compare your projected spread to the market number. If DVOA says the gap between two teams is equivalent to 6 points and the market has the spread at 3, you have identified a potential value bet on the favorite. The strength of DVOA is its long track record — Football Outsiders has published the metric since 2003 — and its demonstrated correlation with future performance, particularly from mid-season onward when sample sizes are sufficient.
EPA: Expected Points Added
EPA — Expected Points Added — is the metric that has most transformed modern football analysis. It assigns a point value to every single play based on how much it changed the team’s expected scoring output. A 15-yard completion on third-and-12 that converts a first down adds substantial expected points because it continues the drive and moves the team closer to scoring. A 5-yard run on first-and-10 from the 50-yard line adds a small amount of expected points because it improves field position marginally. A sack on third-and-8 subtracts expected points because it kills the drive.
The “expected points” framework is derived from historical data. At every down, distance, and field position combination, there is a historical average of how many points the possessing team eventually scores on that drive. A first-and-10 from the opponent’s 20-yard line has a high expected points value — roughly 4 points — because teams in that position score touchdowns frequently. A first-and-10 from your own 5-yard line has a low expected points value — close to 0 or slightly negative — because scoring from that position requires a long drive with many potential failure points.
EPA per play is the rate version of the metric and is the most useful form for betting comparisons. An offense averaging +0.15 EPA per play is adding roughly 0.15 expected points on every snap, which over 65 plays per game translates to approximately 10 more expected points than a league-average offense. A defense allowing -0.05 EPA per play is slightly better than average at preventing opponent scoring. The combination of offensive and defensive EPA per play, compared across teams, produces power ratings with strong predictive validity.
The advantage of EPA over DVOA for bettors is accessibility. EPA data is publicly available through sites like nflfastR and rbsdm.com, and it updates weekly. You do not need a subscription or proprietary model to access EPA numbers. The tradeoff is that raw EPA does not include the opponent adjustment that DVOA provides, so you need to either adjust for schedule strength yourself or use EPA in combination with strength-of-schedule data to get a complete picture.
CPOE: Completion Percentage Over Expected
CPOE isolates quarterback accuracy from the context that inflates or deflates raw completion percentage. A quarterback who completes 70% of his passes sounds elite, but if those passes average 4.2 air yards — meaning he is throwing short, high-percentage throws on virtually every attempt — the 70% is a product of scheme and play design, not arm talent. Another quarterback completing 62% at an average of 9.5 air yards is doing something far more difficult, and his raw completion percentage understates his actual accuracy.
CPOE solves this by calculating the expected completion percentage for every throw based on distance, direction, receiver separation, and other contextual factors. If a pass has a 55% expected completion rate and the quarterback completes it, he earns +0.45 CPOE on that throw. If he misses a pass with an 80% expected completion rate, he earns -0.80 CPOE. Aggregate these values across all throws and you get a quarterback’s season CPOE, which reveals whether he is completing passes at a higher or lower rate than the difficulty of those passes would predict.
For betting, CPOE is most useful in player prop analysis and in evaluating quarterbacks facing new matchups. A quarterback with a +4% season CPOE is consistently completing passes that the average quarterback would miss, which means his passing yards and completion props may be set too low if the sportsbook’s model does not fully weight accuracy above expectation. Conversely, a quarterback with a -3% CPOE is missing makeable throws, and his upcoming production is at risk of declining further if the opponents he faces are even moderately better than average.
CPOE also helps identify regression candidates. A quarterback who started the season with a +6% CPOE through four games is performing at an unsustainable level — very few quarterbacks maintain CPOE above +4% over a full season. If his passing props are set based on his early-season production, the over on those props is likely mispriced because the accuracy will regress toward a more normal level. Similarly, a quarterback with a -5% CPOE is likely performing below his true talent level and may see improvement, making the over on his props attractive.
PFF Grades: The Film-Based Metric
Pro Football Focus (PFF) grades every player on every play on a scale from 0 to 100, based on manual evaluation of game film. Unlike statistical metrics, PFF grades incorporate assessment of a player’s execution independent of the outcome. A quarterback who makes the correct read and delivers an accurate throw that the receiver drops still receives a positive grade for that play. A running back who makes the right cut but gets tackled due to a missed block is graded on his decision, not the result.
PFF grades are valuable for betting because they capture aspects of performance that statistics miss entirely. Offensive line grades, for example, provide insight into pass protection quality and run blocking effectiveness that no box-score stat can replicate. A team with a top-5 graded offensive line is giving its quarterback time and its running backs lanes, which elevates the entire offense’s ceiling. If that offensive line quality is not reflected in the team’s win-loss record — perhaps because of a few unlucky turnovers or special teams errors — the market may be underrating them.
Defensive grades at the individual level help project matchup advantages. If a team’s top cornerback has a PFF coverage grade of 85 and is expected to shadow the opposing team’s primary receiver, the receiver’s prop lines should be adjusted downward. If the sportsbook has not made that adjustment — or has not made it aggressively enough — the under on that receiver’s yardage carries value. These granular, position-level insights are where PFF grades provide the most actionable edge.
The limitation of PFF grades is subjectivity. Two different evaluators might grade the same play differently, and the methodology is not publicly transparent in full detail. PFF grades should be used as one input alongside statistical metrics, not as a standalone authority. When PFF grades and statistical metrics agree — both showing a player or unit performing at an elite level — the signal is strong. When they diverge, additional investigation is warranted.
Putting the Metrics Together
No single metric tells the complete story. DVOA captures overall team efficiency with opponent adjustment. EPA measures play-by-play value creation. CPOE isolates quarterback accuracy. PFF grades add a film-based evaluation layer. The most effective betting models blend these inputs rather than relying on any one in isolation.
A practical workflow for incorporating advanced metrics into weekly betting analysis starts with team-level assessment. Pull each team’s offensive and defensive DVOA and EPA per play rankings. Identify the strongest and weakest units. Then zoom into the matchup: which specific offensive strengths face which specific defensive weaknesses? Use PFF grades to evaluate the individual matchups — offensive line versus defensive front, cornerbacks versus wide receivers — and use CPOE to assess the quarterback’s true accuracy level heading into the game.
The output of this process is a projected point differential that is more informed than anything you could produce from traditional stats alone. Compare it to the market spread. If the gap is 2 or more points, you have a potential bet. If the metrics align with your projection but the market disagrees, investigate why — maybe the market knows something about an injury or a situational factor that the metrics do not capture. If you cannot find a reason for the disagreement, trust the data.
The Numbers Nobody Sees on the Broadcast
The scoreboard shows points. The broadcast shows highlights. Neither shows the underlying process that produced those results. Advanced metrics are the tool that exposes the process — the efficiency, the accuracy, the play-by-play value creation that determines whether a team’s results are sustainable or built on sand.
Most bettors never look at these numbers. They bet based on records, recent results, and narratives they absorbed from sports media. That is the population you are competing against. Every time you consult EPA per play instead of points per game, or check CPOE instead of raw completion percentage, or review DVOA instead of win-loss records, you are making a decision with better information than the average bettor has. The edge from advanced metrics is not dramatic on any single bet. But it is real, it is consistent, and it compounds across a season of disciplined application into results that surface-level analysis cannot match.
