Fact or Fiction? Fifty College Football Beliefs Run Through the Same Test

Fifty familiar beliefs about college football games, run through the same test: does the effect survive team quality, and is anything left after the line?

Baseline points per game · dramatic raw stories usually shrink after adjustment
3,616 games · roughly 50 splits · every figure oriented toward the home team

How to read every table in this article

Each row is a type of game. Four numbers, and the gaps between them are the whole story.

The trick to reading these is simple:

• Raw is big but vs SP+ is ordinary → the effect is just team quality. Fiction.
• Vs SP+ is big and market charged is similar → real, and already priced. Fact, but no edge.
• Vs SP+ is big and market charged is small → real and possibly underpriced. Interesting.
• Everything is ordinary → nothing there.

⚠️ marks a row with under 80 games, where the error bar is wide enough to swallow most conclusions.

Part oneThe ones that are pure fiction

These are the beliefs that evaporate the moment you account for who was playing.

"Home field is worth more in conference play." Look at the raw column: home teams win non-conference games by 9.97 and conference games by 2.43 — a four-fold difference that looks enormous. Now look at vs SP+: +2.59 and +2.58. Identical to a hundredth of a point.

The entire gap is scheduling. Teams buy home non-conference games against weaker opponents. Home field is exactly the same in both.

"Ranked teams dominate at home." They win by 19.92 — and beat expectations by 1.60, which is below the league average of 2.58. Being ranked tells you the team is good. It tells you nothing extra about home field. The market's own charge (+1.64) matches almost exactly.

But note the last row. When both teams are ranked, the home side beat expectations by 5.66 and beat the market by +2.56 ±0.94 across 197 games. Big games at home appear to be worth more than the line allowed. That's one of only a handful of live leads in this entire piece — and it needs an out-of-sample test before anyone acts on it.

"Big favourites don't cover." Raw margins swing 42 points from one end of this table to the other. Vs market swings by about one point, and every row's error bar contains zero. Nothing here.

"Grass is a real home-field weapon." Grass venues do show a bigger raw margin, and even a slightly bigger vs-SP+ figure. But grass venues are disproportionately SEC and Big Ten stadiums — different teams, different climates, different opponents. The market prices in most of the difference and the residual (+0.33 ±0.48) is nothing.

"Coming down from altitude is a disadvantage." Visitors descending 700m+ lose by 8.00 — and beat expectations at a perfectly ordinary rate (+2.13 vs SP+, +0.08 vs market). The raw number just reflects who lives at altitude and who they visit.

Part twoReal, and the market knows

These effects survive the SP+ adjustment. They also show up almost perfectly in the price.

"A bye week matters." Fact. Home team off a bye: the market charges +3.37 instead of the usual +2.52. Away team off a bye: it charges +1.59. That's a swing of about 1.8 points depending on who rested, and it is deliberate pricing, not an accident.

Results move the same way (+4.75 vs +2.49). The leftover is positive in both directions, which is a hint the market may still be slightly under-adjusting — but both error bars are near ±0.95 and the difference between them is only 0.48. Treat as priced.

"Altitude is a weapon." Half fact. What matters is not how high the stadium is — it's how far the visitor climbed. Games at the very highest venues (1,500m+) produce a negative residual: the market charges the most there (+3.67) and home teams underperform it.

Big climbs are a different story: +5.96 vs SP+ against a charge of +3.73. But 105 games with a ±1.62 error bar is not a foundation.

Part threeThe live leads

Three splits where the result outran the price by more than the error bar.

The five-hour flight west. This is the strongest single finding across all our work. A visitor crossing two or more time zones westward beat expectations by 6.98 against a charge of 3.55, leaving +3.42 ±1.09.

The symmetry is what makes it interesting. Teams travelling the same distance east show nothing (+0.18). It isn't distance — it's direction. And it isn't simply "West Coast teams are underrated," because those same host teams are dead flat (−0.04) in their other home games.

Still unproven: it's concentrated among a handful of hosts, it was negative in 2022, and it has never been tested out of sample. We'll know more after 2026.

"Heat is an advantage; cold is an advantage." Half right. Genuine heat (85°F+) produced +1.43 ±0.92 beyond the line, and the market charged nothing extra for it (+2.59, the league average). Cold produced exactly nothing — and note that the market reduces the home charge in freezing games (+1.47), correctly, since results follow.

The heat finding is heavily concentrated in Texas and Arizona venues, so it's as much about specific hosts as about temperature.

"Home field fades late in the year." Possibly. The market charges a flat ~2.5 all season, but results run +0.58 in weeks 9–12 and −0.90 in weeks 13 and beyond — rivalry week and conference championships. That's a 1.5-point swing the price doesn't reflect. Neither figure clears its error bar alone, but the pattern is orderly and worth a look.

Part fourWeather, mostly overrated

Wind, rain and humidity do essentially nothing to the spread. Every residual is inside its error bar. Wind is a genuine and important factor — but for totals, not sides: sustained wind above 15mph knocks around 1.5 points off the closing total, and scoring follows.

That "rain over 2mm" row is a trap worth showing readers. −3.29 looks dramatic. It is 35 games with an error bar of ±2.86. If we'd gone looking for a heavy-rain angle, that's the number we'd have found, and it would have been meaningless.

Part fiveRest, travel distance, and time of day

Short weeks are unmeasurable. Both figures are negative, which makes no sense as a rest effect — if a short week hurt, the two rows would point in opposite directions. With 60 and 53 games, this is noise. Anyone claiming a short-week angle is working from samples this size.

"Long road trips are hard." Nothing at all until 2,000km — the market charges the same ~2.45 from a local derby to a 1,900km flight, and results agree. Beyond 2,000km something appears, but it overlaps heavily with the westward-travel finding above and shouldn't be counted twice.

The short-trip row is a mild surprise: visitors travelling under 250km (usually local rivals) produce −0.75 for the home team. Nearby rivalry games lean slightly to the visitor.

"Night games are different." Mildly. Home teams do a little better under lights (+0.42) and a little worse in morning kickoffs (−0.43), and the market charges slightly less at night. An 0.85-point spread across the day, none of it individually significant, but the ordering is what you'd expect.

Part sixDoes the line movement tell you anything?

This is the cleanest illustration of a hard idea. When the line moves two or more points toward the home team, the market ends up charging +4.11, and the home team delivers +4.37. The move was right.

And it was worth nothing. The residual is +0.26 ±0.65 — the same as every other row, because by the time you can see the move, the price has already moved with it.

Steam being correct and steam being profitable are different things. Both movement rows show a positive residual, which is mildly interesting, but neither clears its error bar.

Conclusionwhat actually survived

Fiction (raw looks dramatic, disappears against SP+): conference vs non-conference, ranked teams at home, big favourites, grass fields, domes, descending from altitude, cold weather, rain, humidity, short weeks.

Fact but fully priced: bye weeks, extreme altitude venues, moderate travel, line movement.

Live leads worth watching in 2026:

1. Westward trips of 2+ time zones — +3.42 ±1.09, the strongest signal we have.
2. Both teams ranked — +2.56 ±0.94, big games at home.
3. Heat above 85°F — +1.43 ±0.92, though concentrated in a few venues.
4. Late season fade — weeks 13+ at −0.90, an orderly seasonal pattern.

None of these is proven. All four were found by looking at 50-odd splits of the same data, which is exactly how false positives are manufactured. We're publishing them as predictions so next season can judge them.

The single most useful habit this article can give you: whenever someone quotes a dramatic number about a type of game, ask whether they've accounted for who was playing. Nine times out of ten — as with those non-conference home games at +9.97 — the effect vanishes entirely.