Pass Rate Over Expected and Game Script: How Teams Really Call Plays

Why Raw Run/Pass Splits Lie

Open any team page and you will find a run/pass split: this offense ran 48% of the time, that one only 38%. It is tempting to read those numbers as identity - the first team is "run-heavy," the second is "pass-happy." That reading is usually wrong, and the reason is game script: the score and time situation that quietly dictates play-calling far more than a coach's philosophy does.

The mechanism is simple and well understood. A team with a big lead in the second half runs the ball to bleed the clock, regardless of how aggressive its coordinator is by nature. A team trailing late throws on nearly every down to catch up and stop the clock. So a great offense that spends Sundays ahead will rack up runs in garbage time, inflating its rushing share; a bad offense that is always behind will pass constantly out of desperation, inflating its passing share. The raw split is contaminated by the scoreboard - it tells you as much about whether a team was winning as about how it wants to play.

One-line idea: teams pass when they are losing and run when they are winning, so the only way to see a coach's true lean is to compare his calls against what the situation expected - not against 50/50.

The Fix: Pass Rate Over Expected

The cure is to model the expected pass rate for every situation and then measure how far a team deviated from it. A model trained on years of play-by-play learns the league-average probability that any team passes given the situation, using inputs like:

Down & distance

3rd-and-8 vs 2nd-and-1

Score margin

up 14 vs down 10

Time remaining

1st quarter vs two-minute drill

Field position

own 5 vs red zone

For each snap the model outputs an expected pass probability. Pass Rate Over Expected (PROE) is then just the actual choice minus that expectation, averaged over a team's plays:

PROE = Actual pass rate − Expected pass rate

Or per play, the deviation that gets averaged up:

$$ \text{PROE} = \frac{1}{N}\sum_{i=1}^{N}\left(\text{pass}_i - \hat{p}_i\right) $$

where pass_i is 1 if the team passed on play i and p̂_i is the model's expected pass probability for that play's situation. A team at +5% PROE passed five percentage points more often than a league-average team would have in the same spots; a team at −5% ran more than expected. Because the expectation already absorbs score and time, PROE strips out game script and leaves behind the coach's genuine pass-or-run lean.

The same trick you have seen before: PROE is an "over expected" metric, the play-calling cousin of CPOE for completions and RYOE for rushing. Each one replaces a raw rate with a deviation from a situation-aware baseline, so context stops doing the lying.

Neutral Game Script and Early Downs

PROE removes the average effect of situation, but analysts often go a step further and look only at neutral game script - the snaps where the scoreboard is not yet forcing anyone's hand. The usual filters are some combination of:

Score within one possession (roughly a one-score game), so neither team is bleeding clock or in catch-up mode.
First and second down, before the down-and-distance of third down dictates the call.
Outside the final minutes of each half, away from two-minute and clock-management distortions.

The logic is that a coach's true identity is revealed when he is free to choose. On early downs in a close game, nothing is forcing a pass or a run, so the call reflects philosophy. Some analysts report PROE over all plays; others report an early-down, neutral-script pass rate directly. Both are after the same thing - filtering out the situations where the scoreboard, not the coordinator, is calling the play.

Example: The "Run-Heavy" Team That Loves to Pass

The cleanest way to feel why this matters is a team whose raw split and whose true lean point in opposite directions.

Example: raw split hides a pass-happy offense

Imagine an excellent offense that spends most of the season with a lead. Its full-season, all-situations numbers look run-leaning:

Split	Pass rate	What it reflects
All plays, all game (raw)	~46%	Dragged down by clock-killing runs while ahead
Expected pass rate for its situations	~42%	The league baseline given all those leads
Neutral game script, early downs	~58%	What the coach actually wants to do when free

The raw 46% screams "run-first team." But its expected pass rate was only ~42%, because it spent so long protecting leads - so its PROE is positive (about +4%): it passed more than its situations called for. And in neutral early-down spots it threw on ~58% of plays. This is a pass-happy offense wearing a run-heavy disguise; the leads it built, by passing effectively early, are exactly what produced all the clock-killing handoffs that fooled the raw split. The figures here are round and illustrative, not any real team's totals.

The trap in one line: a low raw pass rate can be the result of an offense being good - leads create garbage-time runs. PROE and neutral splits separate "chose to run" from "was winning, so ran out the clock."

What High vs Low PROE Actually Says

Once you trust the metric, PROE becomes a clean read on offensive identity and aggressiveness.

High PROE (pass-leaning)

The coordinator chooses to throw more than the situation demands, often on early downs where the league still leans run. This usually tracks with modern, aggressive offenses - and there is a reason for it: on early downs, passing has tended to generate more expected points per play than running, so a positive PROE is frequently the more efficient lean.

Low PROE (run-leaning)

The team runs more than expected even in neutral spots - a deliberately conservative, ground-based identity. It can be the right fit for a dominant offensive line or a limited passer, but a deeply negative PROE on early downs is often where efficiency is being left on the table.

PROE pairs naturally with fourth-down aggressiveness as a measure of how a staff approaches risk: both ask whether a coach follows the efficiency math or defaults to convention. A team that is pass-leaning on early downs and goes for it on fourth-and-short is showing a coherent, aggressive philosophy - the same instinct surfacing in two different decisions, as our fourth-down guide explores. One caution: PROE measures tendency, not quality. A high pass rate executed badly is not a virtue; PROE tells you what a coach wants to do, while EPA tells you whether it worked.

The bottom line

Raw run/pass splits are misleading because game script - score and time - dictates play-calling: teams run to bleed the clock when ahead and pass to catch up when behind. The fix is to model the expected pass rate for each situation (down, distance, score, time, field position) and compute Pass Rate Over Expected (PROE) = actual minus expected, which strips out game script to reveal a coach's true pass-or-run lean. Analysts sharpen this further with neutral-game-script and early-down filters, which isolate the snaps where the coordinator is actually free to choose. High PROE marks an aggressive, pass-leaning identity that often aligns with the early-down efficiency edge of passing; low PROE marks a conservative, run-first one. The example above - a "run-heavy" team at ~46% raw but +4% PROE and ~58% in neutral early-down spots - uses round, hypothetical numbers to show how a strong offense's leads disguise a pass-happy core. And remember PROE measures tendency, not whether the plays succeeded - that is EPA's job.