What EPA means and how to read the EPA page
EPA, expected points added, measures how much each play helped or hurt a team's chances of scoring. It is the stat most sharp bettors and analysts use to rate offenses and defenses, because it credits what actually matters on each play instead of treating every yard the same.
EPExpected points
Before every snap, the down, distance, field position and time left say how many points the offense can expect from the drive, on average, based on thousands of past plays. First-and-10 at your own 25 is worth about +1 point. First-and-goal at the 2 is worth about +6. Backed up at your own 2, it can be below zero, because the other team is more likely to score next.
EPAExpected points added
The change in expected points from one play to the next. A play that improves the offense's situation has positive EPA; a play that makes it worse has negative EPA. Averaged over many plays, EPA per play is a team's efficiency.
EPA = expected points after the play − expected points before it
SuccessSuccess rate
The share of plays with positive EPA. EPA per play rewards big plays; success rate shows how often an offense stays on schedule. A team with a good EPA but a low success rate is living on explosive plays, which tend to be less repeatable.
One play, start to finish
First-and-10 at the offense's own 25. The quarterback completes a pass for 20 yards to his own 45.
On the next drive the same team faces 3rd-and-5 near midfield and takes a sack that forces a punt. That play might be worth about −1.5 EPA: the drive's scoring chances are gone, and the other team gets the ball.
Why it beats yards and points
- Situation counts. A 4-yard run on 3rd-and-3 keeps a drive alive and earns positive EPA. A 9-yard catch on 3rd-and-12 gives the ball away on the punt and earns negative EPA. Yards per game counts the second one as better.
- Turnovers and sacks are priced in. An interception at midfield costs several points of EPA; yards ignore it.
- It is less noisy than points. Points depend on a few red-zone plays, field goal luck and defensive touchdowns. EPA spreads the credit over every snap, so it settles down faster.
- Passing EPA predicts best. How well a team passes, especially on 1st and 2nd down, carries over from week to week more than almost any other stat. Rushing EPA is noisier.
What "adjusted" means on this site
- Garbage time is left out. Plays with a lopsided score don't count: a lead over 43 points in the 1st quarter, 37 in the 2nd, 27 in the 3rd or 21 in the 4th. College blowouts are common, and starters often sit, so these plays say little about the teams. Kneels and plays wiped out by a penalty are left out too.
- Opponents are accounted for. One regression across every play this season gives each team an offense rating and a defense rating, with the strength of every opponent and home field separated out. A good offense that has faced three elite defenses moves up; one that has faced three bad ones moves down. FCS opponents are pooled into one team, so games against them still count, with an adjustment.
- Pass and run get their own ratings, fit the same way.
- Projections are pulled toward average. The EPA line on each game is each offense's rating plus the other defense's, times expected plays, plus 2.5 points for home field (none at a neutral site). Totals are pulled toward the league average; spreads are not, because the regression already shrinks the ratings and pulling them further made them worse.
For an offense, a positive number is good. For a defense, a negative number is good, because it means opponents gained less than expected. Ranks always run #1 = best.
Reading a game on the EPA page
Pick a game from the dropdown at the top of the EPA page. Read it top to bottom: the header gives the bottom line, the two panels show where it comes from, and the team cards show how much to trust it.
1. The header: the ratings vs. the market
| Item | What it means |
|---|---|
| EPA line | What the ratings alone say the spread should be, for example "DEN −1.5". |
| Market | The current sportsbook line. |
| Side gap | The distance between the two, in points. Gaps of 3 or more are highlighted: the ratings and the market disagree enough to look into. It is not a bet signal. |
| EPA total and gap | The same comparison for the over/under. |
| Net EPA rank | Adjusted offense minus adjusted defense, ranked among FBS teams. Two teams with similar net ranks produce an EPA line near a pick'em plus home field. |
2. The panels: each offense against the other defense
Each row shows the value, the league rank and a bar for both sides; a longer bar is a better rank. The Edge column names a team only when it ranks at least 26 spots better than the other side (about the same share of the field as 6 of 32 in the NFL), and otherwise says Even.
| Row | What it measures |
|---|---|
| Adj. EPA / play | Overall efficiency on every pass and run, adjusted for opponents. |
| Pass EPA / play | Efficiency on dropbacks, including sacks and scrambles. The most predictive row. |
| Rush EPA / play | Efficiency on runs. Matters, but swings more from week to week. |
| Early-down pass | EPA on 1st- and 2nd-down passes, before a defense knows the offense has to throw. |
| Success % | Share of plays with positive EPA: how often the offense stays on schedule. |
Look for the rows that decide the game. An elite run defense matters less against a team that throws 60% of the time; a bad pass defense against a strong passing offense matters a lot.
3. The team cards: context and warning signs
| Row | How to read it |
|---|---|
| Rank, raw → adjusted | How much the schedule changed the picture. ▲ means the unit has been better than its raw numbers once opponents are counted; ▼ means worse. |
| QB, EPA / dropback | Points added per dropback for the season's main passer. Check that he's the one starting this week. |
| Plays per game | Pace. Two fast teams push a total up; two slow ones pull it down. |
| Neutral early-down pass rate | How often a team throws on 1st and 2nd down in a close game: its true play-calling tendency. |
| Turnover margin | Turnovers swing on luck, especially fumble recoveries. ±5 or more gets a flag. |
Flags at the bottom of each card list results that usually even out: a big turnover margin, a lopsided record in one-score games, or more or fewer wins than the team's points would predict. A team with good flags has probably been a little lucky; one with bad flags, a little unlucky.
A routine for any game
- Start with the header gap. Under 3 points, the ratings and the market roughly agree. Highlighted means look closer.
- Find what drives it in the panels. Which rows have an edge, and are they the important ones? Overall, pass and early-down pass matter most.
- Check the cards for reasons to doubt it. Big raw-to-adjusted shifts, a different starting QB, flags for turnover or close-game luck, or news the ratings can't know: injuries, opt-outs, weather, rest.
- Decide whether the market has a reason. If you can find one, trust the market. If you can't, the game is worth a closer look.
Right now: the top-rated team
Limits
- It rates the team's play, not one player. A quarterback's EPA includes his line, his receivers and his play-caller.
- Small samples swing. Six weeks is about 350 plays per unit, and schedules vary enormously between conferences. Treat early-season ratings as rough.
- It doesn't know the news. Injuries, opt-outs, weather, rest and travel aren't in the ratings. The market knows all of them.
- The market is hard to beat. The walk-forward check at the bottom of the EPA page shows how the ratings have done against closing lines this season. Use them as context and as a way to find games worth a closer look, not as picks.
Data: ESPN play-by-play via cfbfastR for the current FBS regular season, updated with each weekly refresh. Expected points come from cfbfastR's model.