NFL Advanced Metrics for Betting: How to Use DVOA, EPA and PFF Grades to Find Value

Editorial Team «nfl Betting Strategies» June 18, 2026 Reading time: 18 min

Why Box Scores Mislead NFL Bettors — and What to Use Instead

Early in the 2025 season, a friend messaged me after the Jacksonville Jaguars beat the Indianapolis Colts: “Jags offence looked brilliant — 350 yards, three touchdowns. Backing them next week.” I pulled up the play-by-play data and saw something different. Jacksonville had gained 140 of those yards on two broken-coverage plays. Strip those out and their offence was averaging 3.8 yards per play — below league average. The box score told a story of dominance. The underlying data told a story of two lucky plays masking a mediocre performance.

This is why box scores are dangerous for bettors. Yards, touchdowns and final scores capture outcomes but not process. A quarterback who throws for 300 yards might have accumulated 120 of them on a single screen pass that went the distance, or he might have systematically moved the chains on twelve consecutive drives. The box score treats both scenarios identically. Advanced metrics don’t.

The four metrics I rely on most heavily are DVOA (Defence-adjusted Value Over Average), EPA per play (Expected Points Added), PFF grades, and success rate. Each one measures a different dimension of team and player performance, and together they form a picture that’s far more predictive of future results than anything you’ll find in a standard game recap. Over the next five sections, I’ll break each metric down — what it measures, where to find it, and how I translate it into betting decisions.

DVOA Explained: Defence-Adjusted Value Over Average

DVOA — Defence-adjusted Value Over Average — is the metric I wish someone had explained to me in my first year of NFL betting. It would have saved me a season’s worth of bad takes based on raw yardage rankings.

Published by Football Outsiders, DVOA evaluates every single play of every NFL game and compares it to a league-average baseline, adjusted for situation (down, distance, field position, score) and strength of opponent. The result is a percentage: a team with a 15% offensive DVOA is performing 15% better than an average offence once you account for context. A team at -10% defensive DVOA is allowing 10% fewer expected points than average — and because DVOA is defence-adjusted, that number already accounts for whether the team has been facing elite or bottom-tier opponents.

That opponent adjustment is what makes DVOA more useful for betting than raw EPA or yardage stats early in the season. A team that posts strong raw numbers against three weak defences in its first month will look good in simple metrics but might be mediocre once you control for opposition quality. DVOA handles this automatically. By Week 6 or 7, when every team has faced a mix of opponents, DVOA stabilises into a reliable ranking that correlates strongly with future performance — and more importantly, with future point spreads.

How I use DVOA in practice: I compare each team’s offensive and defensive DVOA to the implied margin in the spread. If the DVOA gap between two teams suggests a five-point margin and the bookmaker is offering a three-point spread, I have a potential value bet on the DVOA-favoured side. The key word is “potential” — DVOA is one input, not the whole model. But it’s the input I trust most for evaluating overall team quality, particularly in the middle third of the season when the data is richest.

One limitation worth noting: DVOA updates weekly, and the midweek version sometimes lags behind the latest injury news. A team’s DVOA might reflect four games with their starting quarterback and one without, blending two very different performance levels into a single number. I always cross-reference DVOA with the current injury report before acting on it.

EPA/Play: Measuring Every Snap’s True Impact on the Scoreboard

If DVOA is the wide-angle lens, EPA per play is the microscope. Expected Points Added measures how much each individual play changes the scoring expectation for the team on offence. A five-yard gain on 3rd and 4 (converting the first down) adds far more expected points than a five-yard gain on 1st and 10 (still leaving 2nd and 5). EPA captures that distinction, and when you average it across all plays in a game — or a season — you get a snapshot of true offensive and defensive efficiency that strips away the noise of garbage time, field position luck and opponent strength.

The New England Patriots’ 2025 season is the best recent illustration of EPA’s predictive power. Their EPA per play jumped from -0.095 in 2024 to +0.119 in 2025 — an improvement of +0.214 that ranked second in the entire league. That swing was visible in the EPA data by Week 5, well before the mainstream narrative caught up. Bettors who tracked EPA caught the Patriots at favourable spreads for weeks before the market adjusted, because public perception was still anchored to the previous season’s results.

I track EPA per play in three splits: passing offence, rushing offence, and total defence. The passing EPA is the most volatile and the most predictive of future scoring. A team with elite passing EPA and mediocre rushing EPA is typically underrated in totals markets, because their ability to score quickly makes them likely to push games over the posted total even when their ground game looks uninspiring. Conversely, a team with strong defensive EPA across both rush and pass is a candidate for under bets, particularly in matchups where the opposing offence ranks below average in passing EPA.

Where EPA gets tricky is in small samples. A single game’s EPA can swing wildly based on a few high-leverage plays — a pick-six, a 70-yard touchdown on a busted coverage, a fumble at the goal line. I never use single-game EPA in isolation. The rolling four-game average is my minimum threshold for decision-making, and I prefer the full season-to-date number once we’re past Week 6.

PFF Grades: What They Capture That Other Metrics Miss

DVOA and EPA are play-level metrics. They tell you what happened on each snap. PFF grades answer a different question: how well did each player perform his assignment, regardless of the outcome? That distinction matters more than you might think.

Pro Football Focus employs a team of analysts who grade every player on every play of every game on a scale from -2 to +2. Those grades are aggregated into overall scores on a 0-100 scale. A quarterback who throws a perfect pass that his receiver drops gets a positive PFF grade even though the play resulted in an incompletion. A running back who hits the right gap and makes a defender miss gets credit even if a holding penalty nullifies the gain. PFF captures the process behind the result.

For betting, PFF grades are most valuable in two areas. The first is offensive line play. No other publicly available metric gives you a reliable read on how well an offensive line is blocking. If a team’s offensive line grades have dropped sharply over three weeks — due to injuries or a schematic change — the effect on the quarterback’s EPA and the team’s scoring output will lag behind by a week or two. The PFF grades flag the problem before the results fully reflect it, which gives you a window to bet against a team whose line is crumbling before the market reprices.

The second area is cornerback and pass-rush grading. A shutdown cornerback (PFF coverage grade above 80) can neutralise a top receiver and reduce the opposing quarterback’s EPA in that matchup. If the spread doesn’t account for that specific coverage dynamic — and it often doesn’t, because lines are set based on team-level data — you have an edge. I’ve found this particularly useful in player prop markets, where a receiver going against a top-graded corner is likely to underperform his yardage line.

The main criticism of PFF grades is subjectivity. Different graders might score the same play differently, and the scales aren’t fully transparent. I treat PFF grades as directional confirmation rather than primary evidence. If DVOA and EPA both flag a team as underperforming and PFF grades agree, I have high confidence. If PFF grades diverge from the other metrics, I investigate why before committing money.

Success Rate: The Consistency Metric Books Undervalue

Success rate is the metric that punishes big-play dependence — and for that reason alone, it deserves a permanent spot in your analysis toolkit. A play is considered “successful” if it gains 50% of the needed yards on first down, 70% on second down, or 100% on third or fourth down. Success rate measures how often a team achieves that threshold across all its plays.

Why does this matter for betting? Because a team that moves the chains consistently is harder to game-plan against and more likely to sustain drives in high-pressure situations. Big-play offences can look spectacular on the highlights, but they’re inherently volatile — if the 50-yard bombs stop connecting, scoring dries up rapidly. A team with a 48% success rate and modest EPA per play is often more reliable for spread betting purposes than a team with a 40% success rate and flashy EPA numbers inflated by a few explosive plays.

I lean on success rate most heavily in totals markets. High success rate on both sides of a matchup tends to produce longer drives, more first downs, and more total plays — all of which push games toward the over. When one team has a high offensive success rate and the other has a high defensive success rate (meaning they frequently limit opponents to unsuccessful plays), the game script tends to be tighter and lower-scoring, favouring the under.

The data is freely available on sites that track play-by-play results, and I’ll point you to the best sources in the final section. Success rate stabilises faster than EPA over the first few weeks of the season, which makes it particularly useful for early-season betting when other metrics are still noisy.

Building a Multi-Metric Model: Blending DVOA, EPA and PFF

No single metric tells the whole story. I’ve been burned enough times by trusting one number in isolation that I now refuse to make a bet based on fewer than two confirming signals — and my strongest plays come when three or four metrics align.

My blending approach is straightforward. For each game on the weekly slate, I pull DVOA rankings (overall, offensive and defensive), EPA per play splits (passing and rushing on both sides), PFF grades for key position groups (offensive line, pass rush, secondary), and success rate for both teams. I don’t run a formal regression model — that level of sophistication exists, and academic work in the Journal of Quantitative Analysis in Sports has shown that multi-factor models outperform single-metric approaches — but I’ve found that a structured checklist works nearly as well for the volume of bets I place.

Jason Ziernicki, the founder of Cleatz, put it well when he said that what happened to home-field advantage over thirty years can only be described as transformational — but certain stadiums maintained significantly stronger advantages, which told his team that specific environmental factors were still at play. That same principle applies to metrics. The aggregate picture matters, but the specific factors driving each team’s numbers matter more. A team with strong DVOA driven primarily by pass defence and a team with strong DVOA driven primarily by rushing offence will perform very differently against the same opponent, even if their top-line ratings are similar.

When metrics conflict — say, EPA ranks a team 8th in offence but DVOA ranks them 18th — I investigate the discrepancy rather than averaging the two. Usually the gap comes down to strength of schedule (DVOA adjusts for it, EPA doesn’t) or a few high-variance games that inflated EPA. Identifying the source of the conflict often reveals the edge: if the market is pricing based on the flashy EPA number while DVOA tells a more sober story, the spread is likely too tight for the overrated side.

The model doesn’t need to be perfect. It needs to be better than the market’s implied probability by more than the 2.38% margin the bookmaker takes. Over a full season, that’s a bar you can clear with consistent multi-metric analysis and disciplined execution.

Case Study: How Metrics Flagged the 2025 Patriots Turnaround

The 2025 New England Patriots were the best betting story of the season, and the metrics told it before anyone else did. Coming off a dismal 2024 campaign, the public consensus was that New England would be bottom-feeders again. Bookmakers agreed — the Patriots opened as underdogs in twelve of their first fourteen games. But the advanced data painted a radically different picture from the opening weeks.

By Week 4, New England’s EPA per play had climbed to +0.119, a staggering +0.214 improvement over their 2024 figure of -0.095. That jump was the second-largest year-over-year EPA improvement in the league, and it was driven by genuine schematic changes rather than a few fluky results. Their offensive success rate tracked upward in parallel, confirming that the improvement was systemic, not just a product of big plays.

DVOA told a similar story. New England’s offensive DVOA climbed into the top half of the league by Week 5, while their defensive DVOA remained strong. PFF grades flagged major improvements along the offensive line — a unit that had graded poorly in 2024 — which explained the quarterback’s increased time in the pocket and the corresponding jump in passing efficiency. Every metric pointed the same direction.

I started backing the Patriots in Week 5 as underdogs. The market was slow to adjust because the brand perception (“bad team, rebuilding”) lagged behind the performance data. By Week 8, the spreads had tightened and the value was gone — but those four weeks of mispricing were enough to generate several profitable bets. The Patriots went on to reach the Super Bowl, but the betting edge was captured months before that outcome became plausible to the wider public.

That’s the core lesson. Advanced metrics don’t guarantee you’ll pick every winner. They guarantee you’ll see team-quality shifts before the betting market does — and in a game where you need to clear 52.4% to profit, that informational lead is all the edge you need. If you’re tracking closing line value as part of your process, you’ll see the confirmation: bets placed on underpriced teams identified by metrics consistently beat the closing number.

Where to Find Advanced NFL Data for Free

You don’t need a paid subscription to access useful NFL analytics. The free tier of data available online is more than sufficient for a serious bettor who’s willing to spend 30 minutes per game on research.

For EPA and success rate, the primary source is the NFL’s own play-by-play data, which multiple sites aggregate and visualise. Football Outsiders publishes weekly DVOA rankings for all 32 teams with breakdowns by offence, defence and special teams — the summary tables are free, though the deeper dives require a subscription. PFF releases a selection of player grades each week for free, with their full database behind a paywall. The free tier is enough to check grades for key position groups (quarterback, offensive line, pass rush) in specific matchups.

I supplement these with social media. Several data-focused NFL analysts publish charts, tables and insights on X (formerly Twitter) throughout the week. Following five or six of the best ones gives you a curated feed of EPA splits, success rate trends and PFF grade highlights without trawling through the raw data yourself. The quality of free NFL analytics on social media is, frankly, better than what most paid tout services offer.

A word of caution: free data sources update on different schedules. Some refresh Monday morning after the weekend slate; others lag until Wednesday. Build your workflow around the update cadence of your preferred sources, and avoid making bets based on stale numbers. The data is only as useful as its timeliness — a DVOA ranking that hasn’t incorporated last week’s games is a DVOA ranking that could mislead you on a team whose trajectory shifted in its most recent performance.

NFL Advanced Metrics Betting Questions Answered

Prepared by the nfl Betting Strategies editorial staff.