RYOE Explained: Rushing Yards Over Expected and the Truth About Running Backs

The Problem With Counting Rushing Yards

The rushing yards column is one of the oldest stats in football, and one of the most misleading. When a back finishes with 120 yards, we instinctively credit the back. But a rushing total is the end product of a long chain - the blocking up front, the scheme and play call, the box the defense showed, and the game script that decided whether the team ran 30 times or 12. The back is only one link. Rushing Yards Over Expected (RYOE) is the analytics community's attempt to pull the back's own contribution out of everything else that produced the yardage.

One-line idea: RYOE compares what a back actually gained on a carry to what a typical back was expected to gain from that exact situation - and credits the back only with the difference.

Why Raw Yards Mislead

Four big confounders sit between a running back and his yardage total, and none of them are really about the back:

Volume

A back with 25 carries will usually out-gain a back with 12, even if the second back is more efficient per touch. Season yardage rewards workload as much as talent.

Blocking

An offensive line that opens a clean lane hands the back five yards before anyone touches him. The same back behind a poor line is tackled at the line of scrimmage.

Scheme & box count

Running against a light box - say, six defenders when the defense is worried about the pass - is far easier than running into a stacked eight-man front on an obvious running down.

Game script

A team with a big lead runs the clock out on the ground against a defense that knows it. A team trailing abandons the run. The score dictates carries, and carries inflate or deflate totals.

Yards per carry fixes the volume problem by averaging, but says nothing about how hard each yard was to get, and a single long breakaway can swing it for weeks. We need a measure that asks, carry by carry, "how good was this run given the situation it started from?"

How the Expected-Yards Model Works

RYOE is a tracking-data metric, pioneered with NFL Next Gen Stats player-tracking chips. A model estimates the expected yardage on a carry from the field state at the moment of the handoff. Trained on thousands of historical carries, it has learned how many yards a typical back gains from any given starting situation.

At a high level:

RYOE = Actual Yards Gained − Expected Yards (given the situation at handoff)

Or in math form, summed over a back's carries:

$$ \text{RYOE} = \sum_{i=1}^{n}\left(\text{ActualYards}_i - \mathbb{E}[\text{Yards}_i \mid \text{situation}_i]\right) $$

The "situation" the model conditions on is rich, because the tracking data captures the whole field:

Input at handoff	Why it matters
Box count	How many defenders are near the line - a light box predicts more yards, a stacked box fewer.
Defender locations & speed	Where every defender is and how fast they are closing on the likely running lane.
Blocker positioning	How well the blocking has set up the play and where the open space is.
Down & distance	Third-and-1 and first-and-10 are completely different rushing environments.
Field position & gap	Yards to the end zone and the intended running lane shape the expectation.

Reading the number: a back with positive RYOE per carry is, on average, gaining more than the model expected given his blocking and the defense he faced - evidence he is adding yardage with vision, balance, and burst. A back near zero is roughly cashing in exactly what the situation offered. Negative RYOE suggests he is leaving expected yards on the field.

An Illustrative Example

Example: two carries, same yardage, very different value

Invented numbers to show how RYOE separates the back from the blocking - not real plays.

Carry 1 - easy lane

Light box, blocking opens a clean hole
Model expected ~6 yards
Back gains 6 yards
RYOE = 6 − 6 = 0

Carry 2 - stacked front

Eight in the box, defender in the gap
Model expected ~1 yard
Back breaks a tackle, gains 6 yards
RYOE = 6 − 1 = +5

The box score records two identical 6-yard runs. RYOE says the second was worth five yards of individual brilliance and the first was simply taking what the line gave him. Over hundreds of carries, those differences accumulate into a real signal about which backs create yardage their blocking did not.

Success Rate and EPA as Companions

RYOE answers "how many extra yards?" but yardage is not the only thing that matters on a run. Two companion metrics round out the picture, and serious rushing analysis uses all three together:

RYOE

Yards added beyond the expectation given blocking and defense

Success Rate

Share of runs that "succeed" for the down (consistency, not just bulk)

EPA / rush

Expected-points value added, which weights situational leverage

A back can have strong RYOE but mediocre success rate if his value is concentrated in a few long breakaways while too many carries go nowhere. Pairing RYOE (yardage) with success rate (consistency) and EPA per rush (situational value) keeps any one number from telling a misleading story. For the value framework behind EPA, see EPA vs. DVOA.

The Bigger Truth About Running Backs

RYOE did more than rank ball carriers - it helped confirm one of the most important findings in modern football analytics: running-back production is heavily environment-driven, and rushing is generally lower-leverage than passing.

When analysts strip away blocking and situation, the gap between a great back and a replacement-level back, measured in points added, turns out smaller than the gap between a great and average quarterback. A back's raw yardage swings enormously with his line and scheme, which is why a new back can step into a strong rushing system and produce immediately. This is the analytics case behind the league-wide reluctance to pay premium prices at the position - much of what looks like individual production is the system around the player.

The nuance: this does not mean backs are interchangeable or that talent is irrelevant. RYOE itself is the proof that some backs consistently beat their expected yards. It means the running back's marginal impact on winning is, on average, smaller and more replaceable than the passing game's - so context matters enormously when you read a rushing line.

The Caveats

It's model-dependent

RYOE is only as good as the expected-yards model behind it. Different providers train different models on different features, so two versions of "RYOE" can disagree. There is no single official formula the way there is for ANY/A.

Back vs. scheme isn't fully separated

The model controls for the situation at the handoff, but blocking that develops after the handoff, scheme design, and a back's specific fit are hard to fully isolate. Some credit that belongs to the system can still leak into the back's number.

It ignores pass-game value

RYOE measures running only. A back who is a dangerous receiver or a reliable pass protector has real value that rushing-over-expected never captures. For modern backs, receiving can be the larger part of the job.

Tracking data is required

Because it needs player-tracking inputs like box count and defender locations, RYOE cannot be computed from a box score and is only available where tracking data exists. It is not a back-of-the-envelope stat.

Use it in concert. RYOE is the best single answer to "did this back beat his blocking?" but it is not the whole evaluation. Combine it with success rate, EPA per rush, receiving value, and a look at the back's offensive line before drawing conclusions.

The bottom line

Raw rushing yards conflate the running back with his volume, blocking, scheme, and game script, so they overstate how much of the production the back created. RYOE uses player-tracking data to estimate the expected yards on each carry from the situation at the handoff - box count, defender locations, blocking, down and distance - and credits the back only with actual minus expected. Read alongside success rate and EPA per rush, it isolates which backs add yardage beyond what their environment gave them, and in doing so it underpins the broader finding that running-back output is largely environment-driven and rushing is lower-leverage than passing. Just respect the limits: RYOE is model-dependent, never fully separates back from scheme, requires tracking data, and ignores a back's value in the passing game.