How Fourth Down Data Works

What is EPA?

Every play in football changes the expected score. A 15-yard pass on 3rd and 10 is worth more than a 15-yard run on 1st and 10 from midfield. The first converts a crucial down and keeps a drive alive; the second is a nice gain but leaves the offense in roughly the same position. EPA — Expected Points Added — measures exactly how much each play changed the expected scoring outcome.

The foundation is the Expected Points model. We take every play from the 2014–2025 seasons and group them by three dimensions: down (1st through 4th), distance (bucketed into 1–3, 4–6, 7–10, and 11+ yards), and field position (10-yard increments from the 1 to the 99). For each situation, we track what happened on the ensuing drive and assign a point value: touchdowns are +7, field goals +3, turnovers –4, punts –1.

The Expected Points for a given situation is the historical average of these outcomes across tens of thousands of plays. Then EPA is simply the difference: the expected points after a play minus the expected points before it.

Success rate is the simplest derivative: the percentage of plays with positive EPA. A team with a 52% success rate is consistently moving in the right direction on more than half its snaps — and that correlates with winning more than yards per play, total offense, or any traditional stat.

Why Independent Computation Matters

Most analytics sites show you numbers from a third party. They pull EPA from someone else's model, display it on their interface, and call it analysis. We do something different: we compute every metric from scratch, using raw play-by-play data.

This matters for three reasons. First, transparency — you can trace any number on this platform back to the raw play data that produced it. Second, methodology control — we can tune how we compute EPA, adjust garbage-time filters, or weight recent games differently without waiting for someone else to change their model. Third, trust — when we say a team's EPA per play is +0.142, we can show you the 847 plays that produced that number.

Power Ratings

Our power ratings blend multiple layers of EPA analysis to produce a single composite number for every FBS team. The rating combines four components, each adding more sophistication to the raw numbers:

Layer 1 — Raw EPA Differential: The starting point is the per-play EPA gap between offense and defense. Layer 2 — Pass/Rush Weighting: Passing EPA is weighted at 65% and rushing at 35%, reflecting the outsized impact of passing efficiency. Layer 3 — Opponent Adjustment: 8 iterative passes adjust every team's EPA for the quality of their opponents. Layer 4 — Recency Weighting: Exponential decay of 0.88 per week, so recent games count more.

The final composite blends all four layers: 10% raw, 20% pass/rush-weighted, 30% opponent-adjusted, and 40% recency-weighted. The heavy weighting toward opponent-adjusted and recency metrics means the model rewards recent performance against quality opponents most of all.

The Lab

We don't just give you our number — we give you the tools to build your own. The Lab exposes every weight and component in the power rating system and lets you adjust them in real time.

Think Ohio State's quarterback situation makes their passing numbers unreliable? Drag the pass offense weight down. Think home field matters more in November than September? Crank it up. You can also override individual team metrics — reduce a team's offensive EPA by 20% to simulate a key injury, or boost a defense to account for a returning starter.

The spread recalculates instantly as you adjust. Every factor's contribution is shown transparently: the differential between the two teams, the weight applied, and the resulting impact on the spread. You see exactly where the number comes from and exactly what changes when you modify an assumption.

Situational EPA

Season-long EPA averages hide important information. A team that dominates between the 20s but stalls in the red zone will look better in aggregate than they actually perform when it matters most. Situational EPA breaks performance into 10 distinct game contexts:

Each situation is tracked for both offense and defense, with FBS percentile rankings. The Team DNA page synthesizes these into a radar chart that shows a team's unique playing identity at a glance.

The Vault

The Vault gives you access to 12 years of game-level betting data — every spread, every result, every weather report — and lets you query it with any combination of filters. Build a system, see the historical record, and decide if the pattern is meaningful.

Filter by home/away, favorite/underdog, spread range, weather conditions (rain, snow, wind, cold, dome), momentum (coming off a win or loss), EPA matchup quality (top 25 offense vs. bottom 25 defense), and season timing. The results are instant — every filter combination queries the full dataset client-side and shows you the W-L record, hit rate, and season-by-season breakdown.

Data Sources

Play-by-play data, game results, drive summaries, schedules, betting lines, weather data, and recruiting rankings are sourced through a commercial data partnership. Every EPA value and team rating shown on this platform is independently calculated by Fourth Down Data from that raw data — we never display pre-computed analytics from any third party.

Our code, methodology, and computations are proprietary. The distinction between the underlying play-by-play data and our analysis on top of it is the foundation of sports analytics as an industry: the numbers you see here are ours.