Underrated / Overrated: 131 Teams Ranked by How Badly the Market Missed Them

We ranked 131 teams by how badly ratings and bookmakers missed them. The list is real history. It is almost useless prediction.

Split-half correlation, 2021–23 → 2024–25 · quality persists; mispricing does not
131 qualifying teams · home and away combined · 30+ games across 4+ seasons

The premise

Our home-field piece asked whether any team plays better at home than on the road. To answer it we compared each team's home form to its own road form — deliberately, because subtracting the road number cancels out any general tendency to beat or miss expectations.

But that general tendency is interesting in its own right. So here it is on its own.

This piece asks a simpler question: across all its games, home and away, does a team beat what was expected of it? Same three measures, no venue split.

• Raw margin — how good the team simply was. Ohio State +21.7, Massachusetts −23.1.
• Vs SP+ — whether the rating system misjudged them.
• Vs the closing line — whether the betting market misjudged them.

That third column is the one everybody wants. It's a five-year list of who the bookmakers under- and over-rated. It is also, we're fairly confident, close to meaningless — and demonstrating why is more useful than the list itself.

Part oneTeam quality is real. Beating the market is not.

This is the cleanest contrast in either article.

Split the five seasons in two. Ask whether what a team did in 2021–23 tells you anything about what it did in 2024–25.

Good teams stay good. A correlation of 0.46 between the first three years and the last two is exactly what anyone would predict, and it's reassuring that the data says so — it tells us the method works.

Teams that beat the betting line, though, show no tendency whatsoever to do it again. A correlation of 0.03 means a team that beat the number by 3 points a game for three seasons projects to beat it by roughly a tenth of a point in the next two.

And the formal test agrees. The spread in team-vs-line records across all 131 teams is statistically indistinguishable from pure chance (Cochran's Q = 140 on 130 degrees of freedom, z = +0.61). Compare that with home field advantage, where about a point of real variation does survive testing. Here there is essentially nothing.

Why this differs from the home-field result

In the home-field piece we found a real spread of roughly 1 point and argued it was worth tracking. Here we find 0.54 and argue it isn't. The difference is that the home-field measure survived a randomization test at p = 0.08, while this one doesn't come close.

It's also what you'd expect if markets work. A persistent, identifiable group of teams the bookmakers systematically misprice would be free money, and free money does not sit untouched for five years.

Part twoThe one that looks real

Notre Dame beat the closing line by 6.4 points a game across 66 games. That's more than three times its own error bar, and by some distance the most striking single number in either piece. There's even a plausible story: an independent playing a national schedule, with unusual public attention and a fanbase that bets.

We still don't believe it, and the reason is worth understanding because it applies to every "system" you'll ever be shown.

When you rank 131 teams, the most extreme one is expected to sit around three error bars from average by chance alone. Notre Dame is right at that boundary. Given that the group as a whole shows no more spread than chance predicts, and that the measure doesn't repeat between periods, the defensible estimate after shrinkage is +1.42, not +6.39 — and even that uses a deliberately generous adjustment.

The same logic applies at the bottom. Tulsa's −4.91 is not a standing instruction to bet against Tulsa. It's what the tail of a distribution looks like.

Part threeReading the three columns together

The three columns tell different stories, and the gaps between them are where it gets interesting.

When raw and vs-line agree — Massachusetts (−23.13 raw, −4.14 vs line) — the team was bad and the market kept overrating it. Usually this means a team declined faster than its reputation.

When they disagree sharply — Old Dominion (−0.73 raw, +3.67 vs line) — the team was mediocre in absolute terms but consistently better than the number asked for. That is what "underrated" actually means, and it looks nothing like being good.

When vs SP+ and vs line disagree — Toledo (+7.37 raw, −3.80 vs SP+, −0.48 vs line) — the rating system and the market disagreed about a team, and the market was closer. Liberty (−4.39 vs SP+, −2.17 vs line) is another. These are cases where SP+ misjudged a team more than the bookmakers did.

That last group is arguably the most useful thing in the table: it flags where a purely statistical rating and a market full of money reached different conclusions.

The table

Column guide. All figures 2021–2025, all FBS-v-FBS games, home and away combined.

• Raw margin — average winning margin. Team quality. Repeats year to year.
• Vs SP+ — margin minus what a neutral-field power rating expected.
• Vs line — margin minus what the closing line expected. Does not repeat.
• ±se — the error bar on that team's vs-line figure.
• Shrunk — the defensible estimate after accounting for sample size.

On the shrunk column. We've used a generous adjustment (a factor of 0.209) rather than the strict one our data supports (0.070). Even so, nobody in the country clears 1.5 points, and the strict reading would put the entire table inside ±0.5. Treat this as a record of what happened, not a shopping list.

Sorted by vs line, most underrated first.

ConclusionThe verdict

There is a list of teams the market got wrong over the last five years. We've published it in full. What there isn't is any evidence that the list means anything going forward.

Team quality repeats — the correlation between one period and the next is 0.46, and any model that ignored it would be broken. Beating the closing line repeats at 0.03, which is another way of saying it doesn't repeat at all. The entire spread across 131 teams is consistent with chance.

The practical rule: use last season's results to tell you how good a team is. Don't use last season's results against the spread to tell you anything.

If you take one number from this piece, take the one at the top: 0.46 versus 0.03.