EPA, or Expected Points Added, tracks how much every play moves the scoring needle. Turn a game into a running tally of value gained or lost, snap by snap, and you get EPA.
What is EPA in football?
EPA stands for Expected Points Added. It tells you how much a play helped or hurt a team's scoring chances compared to what was expected before the snap. Every spot on the field, down, and distance has a known scoring value based on thousands of past drives. A 3rd-and-2 from the opponent's 40 is worth roughly 2 expected points to the offense. Convert for 8 yards and that value jumps to almost 4 expected points, so the play gets credited as +2 EPA. Get stuffed for no gain and now it's 4th-and-2, expected points crater, and the play shows up as a big negative number.
That's the whole point of EPA. It doesn't care about raw yardage, it cares about context. Gaining 4 yards on 3rd-and-3 saves a drive. Gaining 4 yards on 3rd-and-15 does almost nothing. Yards per carry treats both the same. EPA doesn't.
Why did yards per game stop being enough?
Yards per game rewards empty volume instead of winning plays. That's why analytics-driven front offices moved on from it years ago. A team can rack up 400 yards and still lose by two touchdowns if those yards come in garbage time or on drives that stall out in the red zone. Pro-Football-Reference data going back decades is full of 400-yard losers, because yardage totals ignore when the yards happened and how much they actually moved the scoring needle.
Take a team that runs for 5 yards on 1st down, 4 yards on 2nd down, then gets stopped on 3rd-and-1. That's 9 yards on the drive segment and a failed conversion. EPA catches the failure. A yardage stat just sees a decent gain and moves on. That gap is why coaches, GMs, and now plenty of fans lean on EPA instead.
How do you calculate EPA?
EPA compares the expected points value of the field position and down-and-distance before a play to the value after it, then credits the difference to whoever made the play happen. The baseline "expected points" chart comes from historical play-by-play data across thousands of games, mapping every combination of down, distance, and yard line to an average point value teams have scored from that spot going forward (this accounts for the chance the other team gets the ball back too).
Here's a simplified example. A team has 1st-and-10 at their own 25, worth about 0.4 expected points. A 15-yard completion moves them to 1st-and-10 at their own 40, worth about 1.2 expected points. That play earns +0.8 EPA. Do this for every play in a game and add it up, and you get a team's total EPA. Divide by the number of plays and you get EPA per play, which is the number most analysts actually quote.
How do quarterbacks use EPA to prove their value?
EPA per play is now the go-to stat for ranking quarterback performance because it accounts for down, distance, and opponent strength instead of just piling up counting stats. A quarterback who throws for 280 yards on 40 attempts against a soft zone defense in garbage time looks fine on a box score. A quarterback who throws for 220 yards but converts five 3rd-and-longs to keep drives alive is doing more real work, and EPA per play shows that gap.
Look at Patrick Mahomes' 2018 MVP season. His raw stat line was outrageous, 50 touchdown passes, but what analysts pointed to as proof of his efficiency was his EPA per dropback, which ranked at the top of the league according to data tracked by Pro-Football-Reference. That's the number that separated "compiled huge stats on a great offense" from "was actually making winning plays."
What does a good EPA per play look like?
A good offensive EPA per play sits above 0.10. An elite offense usually clears 0.15. Anything negative means the offense is actively hurting its own scoring chances on the average snap. On defense the scale flips: a good defensive EPA per play allowed is negative, meaning the unit is taking points away from opposing offenses play after play. Elite defenses often sit around -0.10 or lower.
To see how tight these margins are, the gap between a top-5 offense and a middle-of-the-pack offense in EPA per play is often just a few hundredths of a point. That's why a single explosive play or one disastrous turnover can swing a team's seasonal EPA numbers more than people expect, especially early in a season when sample sizes are small.
Where does EPA fall short?
EPA is built on averages, so it can miss context that never shows up in a play-by-play spreadsheet, a receiver dropping a sure touchdown, a perfect throw that gets tipped at the line. It also doesn't cleanly separate what a quarterback did from what his offensive line or receivers did on the same play. A running back who breaks three tackles for a 20-yard gain gets the same EPA credit as a running back who walks into the same gain untouched behind a dominant offensive line.
That's why the better breakdowns pair EPA with other tools, like Player Efficiency-style breakdowns for other sports (see our piece on how PER works in basketball for a similar idea applied to hoops), or with charting stats like yards after contact and pressure rate. No single number tells the whole story, and EPA is no exception. It's just a much better starting point than yards or touchdowns alone.
How does EPA connect to scoring predictions?
EPA matters for prediction because it's built on the same foundation as implied probability: both convert game situations into a number representing likelihood of an outcome. If you're trying to predict how many yards a running back will put up, or whether a defense will hold a team under a certain point total, recent EPA trends tell you a lot more than last week's box score. A team that's posted strong offensive EPA per play for three straight weeks, even with close final scores, is trending toward more production, not less.
If you like digging into these numbers before making a call on a player's week, that's basically the whole idea behind GAGE. You're predicting real stat outcomes against the field, not guessing blind. Download GAGE and put stats like EPA to use. For more on how box score numbers can mislead you elsewhere, check out why shooting percentage alone is misleading in basketball.
Is EPA the same as DVOA?
No. EPA measures the raw point value of individual plays. DVOA (Defense-adjusted Value Over Average) takes EPA-style data and adjusts it further for opponent strength and situational context like garbage time. Think of DVOA as a more refined, proprietary cousin of EPA, built by Football Outsiders.
Can EPA be negative for a whole game?
Yes, and it happens more than people think. A team can post negative total EPA even in a win if they scored on a few explosive plays or a defensive or special teams touchdown while grinding out mostly inefficient offensive drives the rest of the game. Good reminder that wins and EPA don't always move together.
Where can I find EPA stats for free?
Pro-Football-Reference and other advanced football analytics sites publish EPA and EPA per play data broken down by team, player, and situation, usually updated weekly during the season.