Pass Rate Over Expected (PROE)

Losing teams throw and winning teams run. PROE takes that out, so what is left is what the coach chose.

Rank byPass over expectedEarly downs onlyFourth down vs peersFourth down vs the modelRaw pass rate

LeagueAll FBSACCAmerican AthleticBig 12Big TenConference USAFBS IndependentsMid-AmericanMountain WestPac-12SECSun Belt

Season202620252024202320222021202020192018201720162015

Pass rate minus the league's pass rate from the same down, distance, field position, score and clock. A statement about the staff rather than the schedule.

2026 is still being played — through week 2 · 2 games a team · 130 of 136 teams rated so far. These are early numbers on a small sample, not a season rating. Teams whose only game was against an FCS opponent are not here: those games are excluded from every metric on this page.

Most

+13.5%

Middle Tennessee

Least

−10.6%

Liberty

Spread

24.1%

top to bottom

Fourth downs

21%

model says 34% — why →

Pass over expected— against the spots faced

Across is what an average staff would have thrown from this team's actual snaps. Distance above the line is identity; distance along it is game script.

130 teams · dashed lines mark the league medianr = 0.49(r² = 0.24)

Pass over expected— the two ends

← runs more than expected±14%throws more →

2026 through week 2

#	Team	Conf	Pass	Exp	PROE	Early	Go	vs peers	vs model
1	Middle Tennessee	Conference USA	70.2	56.6	+13.5	+13.7	14	-15.4	-25.0
2	Sam Houston	Conference USA	66.4	53.5	+12.9	+13.8	14	-0.3	-42.9
3	Florida International	Conference USA	59.6	49.4	+10.2	+14.1	60	+24.3	+0.0
4	New Mexico State	Conference USA	63.2	57.4	+5.8	+6.6	27	+1.3	+0.0
5	Kennesaw State	Conference USA	63.6	59.1	+4.6	+8.7	20	-0.9	+0.0
6	Western Kentucky	Conference USA	54.3	53.3	+1.0	+4.5	29	-3.1	-50.0
7	Delaware	Conference USA	55.6	59.5	-4.0	-1.8	29	-4.8	+0.0
8	Jacksonville State	Conference USA	42.7	48.0	-5.3	-0.5	30	+10.1	-28.6
9	Liberty	Conference USA	44.3	54.8	-10.6	-12.0	43	+9.7	+0.0

How this number is computed

A gradient-boosted classifier predicts the probability that a given snap is a pass, from the game state alone — down, distance, yardsToGoal, margin, secs_left, period, venue, off_to, def_to — fit on 118,690 fourth downs and roughly a million scrimmage plays across 12 seasons. PROE is a team’s actual pass rate minus what that model expected from the same situations, so the schedule and the scoreboard are divided out before any team is compared with any other.

⚠ No team identity enters the model. That is what makes it a measure of choice rather than a measure of circumstance — and it is also why a team with few plays has a noisy number rather than a wrong one. Excludes overtime, garbage time, FBS vs FCS. Rebuilt 2026-09-13.

The reference table

What the league throws from each situation — the baseline every “over expected” number on this page is measured against. The model itself is a gradient boosting fit on down, distance, yardsToGoal, margin, secs_left, period, venue, off_to, def_to; this is the same thing in a form you can read.

Down	To go	opp 20	opp 21-50	own 20-49	own 1-19
1	1-2	12%	—	—	—
1	3-4	20%	—	—	—
1	5-7	26%	43%	40%	—
1	8-10	35%	46%	45%	36%
1	11+	47%	53%	52%	45%
2	1-2	17%	34%	33%	24%
2	3-4	30%	37%	38%	32%
2	5-7	39%	44%	47%	42%
2	8-10	50%	54%	56%	52%
2	11+	58%	62%	65%	59%
3	1-2	19%	18%	23%	26%
3	3-4	53%	51%	69%	67%
3	5-7	72%	69%	82%	78%
3	8-10	79%	78%	84%	77%
3	11+	76%	75%	77%	68%

Excluded from every number on this page: overtime, garbage time, FBS vs FCS. A call made down forty in the fourth quarter is not a call.