NFL Advanced Metrics for Betting: EPA, DVOA and Efficiency Stats Explained
Available in US
A team finishes the season 10-7 and averages 380 yards per game. Their division rival goes 9-8 with 340 yards per game. Which team is better? If you answered the 10-7 team, you made the same mistake I made for three straight seasons. Win-loss records reflect schedule strength, turnover luck, and close-game variance as much as they reflect team quality. Yards per game ignores context entirely — 5 yards on 3rd-and-15 is a failure, while 5 yards on 3rd-and-4 is a first down. Advanced metrics exist to strip away that noise and expose the signal underneath.
Two metrics dominate the NFL analytics landscape for betting purposes: EPA (Expected Points Added) and DVOA (Defence-adjusted Value Over Average). They measure different aspects of performance, use different methodologies, and answer different questions. Together, they form the analytical layer beneath every spread and total I evaluate.
EPA Explained: The Points That Each Play Produces
Expected Points Added starts with a simple premise: every game situation has an expected point value. First-and-10 at your own 20-yard line is worth roughly 0.5 expected points. First-and-goal at the opponent’s 2-yard line is worth roughly 5.8 expected points. EPA measures how much each individual play changes that expected point value.
A quarterback throws a 15-yard completion on 2nd-and-8 from his own 30. The expected point value before the play was approximately 1.0. After the play — now 1st-and-10 at the 45 — it’s approximately 2.1. The quarterback’s EPA on that play is +1.1. If he’d thrown an incompletion, the expected point value would have dropped (now 3rd-and-8), and his EPA would be negative.
EPA per play is the gold standard for evaluating offensive and defensive quality in NFL analytics. It captures context that raw statistics miss entirely. A quarterback who throws for 280 yards with an EPA per dropback of +0.15 is performing better than one who throws for 320 yards with an EPA per dropback of +0.05 — because the first quarterback is generating more value per opportunity even with fewer raw yards.
Warren Sharp, who has maintained a documented 57-63% win rate on NFL totals with a 612-456 overall record in the 2025 season, uses EPA as a cornerstone of his totals analysis. His approach compares offensive EPA per play to defensive EPA per play for both teams in a game, adjusting for pace (plays per game). The resulting projection often diverges from the posted total by 2-4 points — and those divergences identify the highest-value totals bets.
For betting, I use EPA in three specific ways. First, EPA differential (offensive EPA minus defensive EPA) as a predictor of margin of victory — this correlates with future point differential more reliably than win-loss record. Second, EPA per dropback as a quarterback evaluation metric that strips out rushing production to isolate passing efficiency. Third, EPA in high-leverage situations (3rd downs, red zone, two-minute drills) to identify teams that outperform or underperform in situations that disproportionately determine outcomes.
DVOA and Success Rate: Context-Adjusted Team Quality
DVOA — Defence-adjusted Value Over Average — takes a different approach. Where EPA assigns point values to each play, DVOA measures each play’s success or failure relative to a baseline of league-average performance in the same situation, then adjusts for the quality of the opponent. A 5-yard gain on 3rd-and-3 is successful. A 5-yard gain on 3rd-and-8 is not. DVOA captures that distinction and scales it across the entire season.
Football Outsiders, the primary source for DVOA data, publishes weekly rankings that break down performance by offensive and defensive categories: passing DVOA, rushing DVOA, overall DVOA. The defensive adjustments are crucial for betting. A team with a 15% offensive DVOA that achieved it against a schedule of below-average defences is fundamentally different from a team with a 12% offensive DVOA that achieved it against a gauntlet of top-10 units. DVOA accounts for that difference; raw statistics don’t.
Corey Shank’s academic research, spanning 14 NFL seasons, identified market inefficiencies partially explained by the types of efficiency metrics that DVOA captures. The betting market prices games using composite power ratings that incorporate basic statistics, but those ratings lag behind efficiency-adjusted measures in recognising when a team’s record overstates or understates their true quality. A 5-3 team with a negative DVOA is a regression candidate; a 3-5 team with a positive DVOA is an undervalued asset.
Success rate — the percentage of plays that achieve “successful” outcomes (generally 50% of needed yards on first down, 70% on second, 100% on third/fourth) — complements DVOA as a stability metric. DVOA can be skewed by explosive plays; success rate measures consistency. A team with moderate DVOA but high success rate is performing sustainably. A team with high DVOA but low success rate is riding big plays that may not continue. For spread betting, I weight success rate as a “floor” indicator and DVOA as a “ceiling” indicator.
From Metrics to Spreads: Building the Bridge
Advanced metrics only matter for betting if you can convert them into a predicted spread and compare that prediction to the market’s number. Here’s my conversion process, stripped to its essentials.
I calculate EPA differential (offensive EPA per play minus opponent defensive EPA per play) for each team. The difference between the two teams’ EPA differentials correlates with expected margin of victory at roughly a 2.7-to-1 ratio — meaning a 0.10 EPA differential advantage translates to approximately a 2.7-point predicted margin. I adjust for home-field advantage (approximately 1.5 points in the current NFL) and specific situational factors (rest differential, travel, divisional matchup).
The resulting predicted spread gets compared to the market spread. If my EPA-derived spread says Team A should be a 5.5-point favourite but the market has them at -3, I have 2.5 points of potential value. I then cross-reference with DVOA to check whether the EPA signal is sustainable (high success rate, schedule-adjusted) or fragile (driven by explosive plays against weak opponents). Only games that clear both the EPA divergence threshold and the DVOA sustainability check make it to my bet slip.
This process sounds clinical, but it’s faster than it reads. By Week 4 of the season, I have enough data to run the conversion in about 20 minutes per week using pre-built spreadsheets. The machine learning model I’ve built automates most of this, but I still run the manual conversion as a cross-check because the model occasionally misweights situational factors that my experience recognises.
The single most important insight from working with advanced metrics: they predict future performance better than past results. A team that’s 3-5 with strong EPA and DVOA numbers is more likely to finish 8-9 or 9-8 than to continue losing. The market recognises this eventually — but “eventually” means there’s a window of weeks where the record-to-efficiency disconnect creates betting value. My job is to find that window before it closes.
Which single NFL advanced metric best predicts against-the-spread performance?
EPA differential — the gap between a team’s offensive EPA per play and their opponents’ defensive EPA per play — is the single strongest predictor of future margin of victory and, by extension, ATS performance. It captures both the team’s offensive efficiency and the context of their defensive matchups in each game. DVOA provides a valuable complementary signal by adjusting for opponent strength and distinguishing sustainable performance from explosive-play variance, but as a standalone predictor, EPA differential has slightly higher correlation with future point spreads.
Where can I access free EPA and DVOA data for NFL betting analysis?
EPA data is available for free through the nflfastR play-by-play data repository, which provides play-level EPA calculations for every NFL game dating back to 1999. Pro Football Reference offers processed EPA summaries. DVOA data is published weekly by Football Outsiders during the season, with free access to current-season rankings and limited historical data. For deeper historical DVOA analysis, a Football Outsiders subscription provides access to multi-season datasets. Both sources update within 48 hours of each game, which is fast enough for mid-week model runs.
Created by the ”nfl Betting Systems” editorial team.
