Revenge Games, Trap Games, and Other Stories the Market Already Read

Analysis · by Rick's Picks Analytics

Every college football Saturday comes with a script. The revenge game: a team that got embarrassed last year circles the rematch on the calendar and plays possessed. The trap game: a ranked favorite looks past a bad opponent to next week's showdown and gets clipped. The letdown spot. The team riding momentum. The squad coming off a bye, rested and scheming.

These stories are fun. They might even be true in some football sense. But the question that matters for analysis is different: does the betting market misprice these situations? If oddsmakers already know the story, it's baked into the spread, and the angle is worth nothing.

We tested every classic situational angle against 7,962 FBS games using our standard methodology: train on 2015-2022, test on held-out 2023-2025, break-even at 52.4% against the spread at standard -110 pricing, Bonferroni correction when a hypothesis file tests multiple variants. Angles only get promoted into our engine if they clear break-even in both windows.

Here's the scoreboard. It is not kind to the folklore.

The scoreboard

Zero promotions. Every row is a story the market already read.

The revenge game: watching 53.9% die in real time

Revenge is the best case study we have, because we got to watch the mirage form and evaporate.

When we first ran the revenge angle — back any team facing an opponent that beat it the previous season — the train window said nothing: 50.8% across 3,187 games. But the 2023-2024 test window came back at 53.9% on 776 games. Above break-even, one-sided p around 0.21 — not significant, but the kind of number that makes you look twice.

Our rules said no. Without train support, a test-only trend is exactly what overfitting looks like, so revenge stayed rejected and we flagged it for re-validation.

Then the 2025 season finished, and we scored the angle against 367 revenge games it had never seen: 50.7%. A coin flip. The pooled 2023-2025 test settled at 52.8% — still under break-even once you account for the vig, and with a 95% confidence interval (49.9%-55.7%) that comfortably contains "nothing here."

That's the whole life cycle of a betting myth compressed into two seasons: a noisy hot stretch, a plausible narrative to explain it, and then regression the moment new data arrives. If we had published "revenge teams cover 53.9%!" in early 2025, we'd have been selling noise.

The letdown spot: first incomputable, then backwards

The letdown angle — fade a favorite coming off a win over a top-10 opponent — has a special place in our integration log, because for a while we literally could not test it. Our historical rankings data was too sparse: the query returned zero qualifying games. An angle that touts have recited for decades, and the data to check it didn't exist in our database until we backfilled AP rankings.

Once computable, the result was stranger than a simple no. In train, fading the letdown favorite went 48.3% — the angle loses, and the opposite side (backing the favorite everyone expects to sleepwalk) was the better bet. In the 2023-2025 test window, fading went 57.0% on 149 games, and 60.4% in 2025 alone.

A train/test sign-flip on a small sample is the least trustworthy pattern in this entire line of work. Two windows, two opposite answers, neither significant after correction (the test p-value is 0.14 against break-even). We log it as a near-miss to re-check after 2026, and we bet exactly nothing on it. The trap-game cousin — fading a favorite with a ranked opponent looming next week — was quieter and just as dead: 49.7% train, 52.0% test.

Momentum: whichever side you take, the next season takes the other

In-season form is where the folklore contradicts itself, and our data cheerfully confirms both contradictions are worthless.

Theory one says ride the hot team. Theory two says regress to the mean — fade teams covering at an unsustainable clip. We tested both across multiple definitions:

• Regress-to-mean (fade teams that covered at least 2 of their last 3 while opponents covered 1 or fewer): 52.0% train, 50.8% test.
• Back the hot team (recent scoring-margin gap of 10+ points): 50.7% train, 52.0% pooled test — and then 50.4% in the 2025-only slice. The minimum train p-value across every momentum variant was 0.64.

Every momentum flavor we defined flips sign across holdout years. That's not "we haven't found the right threshold yet." That's a market repricing teams within a week of new information, which is exactly what an efficient market does. Three games of form is not private information; it's the first thing the line-setters look at.

The bye week: a textbook out-of-sample death

Rest advantage was our most respectable-looking casualty. In the train window, teams coming off a bye against opponents on standard rest covered 53.0% of 1,115 games — above break-even, and the train p-value passed. For a while this was our "closest call."

The test window executed it. Pooled 2023-2025: 49.6% (398 of 802). The 2025 season alone: 46.2% — the sign fully inverted. An angle that had eight seasons of gentle support gave it all back in three. The extreme-rest variant (15+ days versus short rest) was worse and comically small: 44.4% in train on 18 games, 23.1% in test on 13. Extra practice time is not a secret. It's printed on the schedule.

Coaching angles: the market reads the press conferences too

Fading a first-year head coach at home — the "system installation chaos" theory — ran 51.3% in train, 50.2% in test, and 46.7% in 2025. Backing the massively more tenured coach (10+ years of experience edge) hit 51.7% in train and then sign-flipped to 49.0% in test. Coaching changes are the most publicized events in the sport. Of course they're priced.

Why these myths refuse to die

If none of this works, why does every broadcast and every tout thread keep repeating it?

Survivorship of the anecdote. You remember the revenge game where the underdog covered by 20, because it made a great story and a highlight reel. You don't remember the 49% of revenge games where nothing narratively interesting happened. Memory keeps the hits and discards the denominator. Our denominator is 3,187 games, and it keeps everything.

Small samples masquerading as signals. Elsewhere in this same research round, a line-movement angle hit 81.8% in train — on 11 games. It scored 44.7% in test. Situational angles are especially prone to this because the "situation" can be defined narrowly enough to produce any win rate you want in hindsight.

The stories are real; the mispricing isn't. This is the subtle one. Players probably do get up for revenge games. Favorites probably do look ahead sometimes. But the question was never whether the effect exists in football — it's whether it exists and the market missed it. These angles have been in print since before the internet. Every one of them is public information, and public information is what a closing line is made of.

No one publishes the failures. Content that says "this angle is dead, here are 1,143 games proving it" doesn't sell picks. So the ecosystem recycles the same angles annually, and the negative results — like the ones on this page — mostly go unwritten. We think publishing them is the entire point.

The bottom line

We tested the classics — revenge, letdown, trap, momentum, regression, bye weeks, coaching tenure — across 7,962 games with a strict train/test split, and promoted exactly zero of them. Revenge's 53.9% test blip collapsed to 50.7% the moment 2025 arrived. The bye-week edge inverted to 46.2%. Momentum flips sign whichever way you define it. The letdown spot went from incomputable to backwards.

None of this means situational football is fake. It means the betting market read these stories long before you did, and the closing spread is where the story ends up. When one of these angles ever does clear both windows and a Bonferroni-corrected test, we'll promote it into the engine and tell you. Until then, our analysis treats every "everyone knows" angle as what the data says it is: already priced.

Rick's Picks publishes statistical analysis of college football for informational and entertainment purposes. Nothing here is betting advice. 21+. If gambling is affecting you or someone you know, call 1-800-GAMBLER.