The One-Point Question: How Much Is Home-Field Advantage Really Worth?

Everyone prices home field at two and a half points. Ten seasons and 6,954 games say the folklore is broadly right — but the team rankings are far noisier than they look.

Average home-field value, points per game, 2021–2025 · three answers, six hundredths apart
6,954 games · 131 qualifying teams · 2026 watchlist · every figure reproduces from hfa.db

Part oneThe rumour of its death is exaggerated

The most repeated claim about home field over the last five years is that it's dying. In the NFL, the NBA and European soccer, that's well documented. In college football it simply is not happening, and if you measure carefully the arrow points the other way.

Every number here comes in up to three flavours:

• Raw — the home team's actual margin. Bundles home advantage together with the fact that home teams are usually better and usually play a softer home slate.
• Vs SP+ — subtracts what a neutral-field power rating expected, removing opponent quality.
• Vs the closing line — subtracts what the betting market expected. Since the market's number already contains a home adjustment, what's left is the market's error.

Weekly SP+ ratings only exist in our data from 2021, hence the blanks.

Pre-pandemic (2016–19) the home team won by 4.44. Post-pandemic (2021–25), 4.52. The two largest home margins of the decade are the two most recent seasons.

The middle column sharpens it. Measured against ratings, home advantage has grown +0.43 points per season (t = +2.30) while the market's own adjustment stayed flat near 2.5. That gap is why home teams beat the closing number in both 2024 and 2025 — the only back-to-back positive years in the sample.

⚠️ Required caveat. That is a trend fitted to five annual points, found after looking at them. Two seasons do not establish a structural change, and the market may have caught up already. Present it as something to watch.

2020 is the exception that proves the rule. Average attendance fell from ~42,000 to 6,108, and home field halved to +2.13. Take the crowd away and home field collapses; give it back and it returns. (2020 also had truncated schedules, opt-outs and severe roster disruption — suggestive, not a clean experiment.)

Part twoThree ways to weigh a home field

Take each team's average result at home minus its average on the road, and halve it. The halving matters: home advantage helps you at home and hurts you away, so the full home-to-road swing is worth twice the home effect.

The market's own adjustment averages +2.52. Results delivered +2.58. Across 3,616 games the market is off by six hundredths of a point.

Why that last number isn't as impressive as it looks

A fair question: did the market really get home field exactly right over five seasons?

Partly, that zero is bookkeeping. Every game gives one team a positive result against the line and the other the identical negative one. Add up every team-game and you get exactly zero — always, in any sport, in any season, however badly priced the market is. That number can never tell you anything.

The number that can is the home-only figure: +0.06 points per game. Nothing forces that to be small. If the market had systematically underpriced home teams it would read +2.00. That it lands six hundredths from zero across 3,616 games is a real result.

The team-by-team average of −0.00 in the table below is that same +0.06 restated, not a second independent confirmation. One finding, not two.

Callout: three sources, one answer

SP+ applies a flat 2.5-point home adjustment to every game, with no team or venue variation. We verified this by inverting SP+'s published game projections against its published team ratings across 2,710 games — the residual has a median of exactly +2.500 in 60 of 61 weekly sheets. So SP+ assumes 2.50, the market prices 2.52, and the games delivered 2.58.

The adjustments don't shrink teams — they sort them

You might expect that if the league average falls from 4.03 to 2.55 to zero, every team shrinks the same way. It doesn't.

The league average falls sharply. The gap between teams barely moves:

Each adjustment subtracts a similar amount from nearly every team — like taking 20 marks off everyone's exam. The average drops; the distance between best and worst doesn't.

What happens to individual teams is more interesting. Some shrink toward zero:

• Tennessee: +9.70 raw → +5.36 vs SP+ → +1.62 vs the line
• Kent State: +8.32 raw → +2.26 vs SP+ → −0.26 vs the line

Others move the opposite way and get more extreme at every step:

• Bowling Green: −1.71 raw → −4.26 vs SP+ → −6.65 vs the line
• Arkansas: +0.17 raw → −2.45 vs SP+ → −5.19 vs the line

A raw home record blends three separate things: genuine home advantage, how easy the home schedule was, and how good the team is. Tennessee's big raw number was largely a soft home slate; strip that out and little remains. Arkansas's ordinary-looking number was concealing genuinely poor home performances against beatable opponents.

So the ranking changes depending on which column you read. Any list of "best home fields" built on raw results is measuring schedule as much as home advantage.

Part threeDoes anyone actually own a better home field?

Set the three result columns beside a fourth: what the market thinks each team's home field is worth.

Results appear to span 12–15 points. The market spreads all 131 teams across a band 2.7 points wide, with 80% between +1.9 and +3.3. Somebody is wrong.

Testing it the hard way

Each team has 20–36 home games against a game-to-game noise level of about 15 points, giving each team's measured home advantage a standard error near ±2.05. When the spread across teams is about the same size as the typical error bar, most of what you see is chance.

There's a further catch: teams are not independent. If Alabama beats the line at home, its opponent simultaneously records a road result in the opposite direction — the same game, counted twice with opposite signs. Standard tests assume independence and get this wrong.

So we tested it properly. Keep the exact schedule and the exact set of 3,622 results, but shuffle which game each result belongs to. That destroys real team identity while perfectly preserving the schedule, the home/road counts, and the zero-sum structure.

Two things follow.

The raw numbers are useless. The observed spread in raw home-road performance is smaller than chance alone would produce. Ranking teams by raw home record is ranking noise.

Something real probably survives the adjustment — a genuine spread of roughly 1 point, meaning an exceptional home field is worth perhaps 1–2 points more than average, not 5 or 6. But at p = 0.08 this is a lead, not a conclusion. Which is why the next section sets up a test rather than handing you adjustments to use today.

Part fourThe teams, and a 2026 watchlist

Why we can't just check whether it repeated

The obvious test: do teams that over-performed at home in 2021–23 do it again in 2024–25?

Near zero — and it's tempting to call that the end of the argument. It isn't, and this is where most analysis of this question goes wrong.

Splitting five seasons in two leaves each team 10–18 home games per half, with standard errors of 2.60 and 3.11. Given a real spread near 0.9, the highest correlation this test could possibly produce is +0.093. We observed +0.021 — well inside the range consistent with a real one-point effect. The test cannot distinguish "no team effect" from "a one-point team effect," so it shouldn't be used to argue either.

The 2026 test

So we do it the honest way. The table below is a pre-registered watchlist, published before the season. If team home-field advantage is real and persistent, the teams at the top should lean the same way again in 2026. If it's an artefact, they'll scatter.

We'll publish the scoreboard either way.

How many extreme teams should exist by chance? With 131 teams, chance predicts 6.0 beyond two standard errors. We observe 9. At 1.5 standard errors, chance predicts 17.5 and we observe 27. A modest excess, exactly what a ~1-point real spread implies.

The table

Column guide:

• Raw home — average margin in home games only. This is mostly team quality: Ohio State is top because Ohio State is good, not because its home field is special.
• Raw H−R/2 — half the home-minus-road contrast. Removes team quality; still contains schedule imbalance.
• Vs SP+ — same contrast, after subtracting a neutral-field rating. The best single measure of home advantage here.
• Vs line — same contrast, after subtracting the market's expectation. What the market missed.
• ±se — the error bar on that team's "vs line" figure.
• Shrunk — the defensible estimate once sample size is accounted for. This is the only column that should inform a number you actually use.

A note on the shrunk column. We've been deliberately generous. The strictest reading of our data shrinks these figures by a factor of 0.165; we've used 0.279, taking the upper end of the plausible range for real team variation. This gives teams the benefit of the doubt. A stricter analyst would roughly halve every number in that column, and we couldn't argue with them.

Sorted by vs SP+. All figures 2021–2025. Qualifying = ≥15 home and ≥15 road FBS-v-FBS games across ≥4 seasons → 131 of 136 teams.

Not qualifying: Delaware, Jacksonville State, Kennesaw State, Missouri State, Sam Houston.

ConclusionStart at 2.5. Move rarely. Never by much.

Across ten seasons and 6,954 games, home field in college football is worth about two and a half points once you account for home teams being the better team. It is not declining. If anything it has grown since 2021 while the market's number stood still.

On team-specific home field, the honest answer is partly. About 80% of the apparent spread is sampling noise, and any ranking built on raw home records is meaningless. But roughly a point of real difference probably survives. Some teams genuinely are better at home — by far less than five years of results imply, and not reliably enough yet to say which.

1. Anchor at 2.5. Adjust by at most a point, only for teams at the extremes of the shrunk column, and only when there's a football reason as well as a number.
2. Don't add what's already in the line. Byes, altitude and travel are real and already priced.
3. Keep the uncertainty. A perfectly fair line still misses by a median of 10.5 points, and one game in twenty misses by 30 or more.

We'll revisit the watchlist after the 2026 season.