Every fan base has said it. We sleepwalk through the cupcakes and play out of our minds against the top-10 team. The 30-point favorite needs a fourth-quarter stop to survive.
We checked. DAWG grades every Power 4 game against what Vegas expected: DAWG = points better or worse than Vegas expected, garbage time removed. That lets us ask the question fairly. Not "did they win," but "did they do better or worse than expected, and did that depend on who they played?"
It looks real
We took every Power 4 team-season from 2014 through 2025, about 800 of them. We split each schedule in half by the point spread: the tougher half, where the team was the bigger underdog or the smaller favorite, and the easier half. On average, the two halves sit about 21 points of spread apart.
The tougher half wins. Teams graded about 2 points better against their tougher opponents than their easier ones: +1.1 against the tougher half, −0.9 against the easier half. It happened in every season we have, 2014 through 2025.
Teams look better against the tougher half of their schedule
The gap, season by season (2014–2025)
The twist
Look at how "tougher" was decided: by that game's spread. That isn't the same thing as how good the opponent is. An opponent's strength is about that team, whoever it plays. A spread is Vegas's prediction for one particular game. It reflects more than just how good the two teams are.
So we measured opponents a different way. Before each game, we rated both teams using only what the betting market had said about them in their earlier games that season, with home field taken out. Then we split each schedule again, this time by how strong the opponent looked going in.
The pattern disappears. On the same games, splitting by the spread showed teams about 2 points better against tougher opponents. Splitting by the opponent's form showed a difference of −0.4 points, too small to tell from zero. Same teams, same seasons, same games. The only thing that changed is how we decided which opponents were tough.
Sort by the opponent instead of the spread, and the gap disappears
- That game’s spread
- Vegas’s prediction for this one game
- Opponent’s strength before kickoff
- How good the opponent had looked in its earlier games, home field taken out
What's left is the part of each spread that's specific to that one game: whatever moved the line beyond the two teams' form so far and home field. When that piece made a team a bigger favorite than its form suggested, the team tended to fall short. When it made the team a bigger underdog, the team tended to do better.
That's what the data show. What it means is less certain. One likely explanation is that the line for a single game sometimes moves further than the game turns out to justify. Then a team looks like it "played down" or "rose up" when the number was what was off. We can't prove that. What we can say: we found no sign that an opponent's strength, on its own, changes how teams perform against expectations.
What about your team?
Every season has a team that seems to prove the theory:
- Miami 2019 graded about 9 points better than expected in its tougher games, which were close to pick'em on average, and about 18 points worse than expected when it was favored by around 22.
- California 2025 graded +8.8 against its tougher half and −12.3 against its easier half.
- Vanderbilt 2024 was a 22.5-point home underdog to Alabama and won 40–35, a +25.3 grade. Across its five toughest games by the spread it graded +18.9; across its five easiest, −3.5.
And every season has the reverse. Maryland 2021 graded about 16 points worse than expected against its tougher half and about 8 points better against its easier half.
Those seasons happened. Split a 12-game season into two halves of six, and ordinary ups and downs alone can open a big gap. Across about 800 team-seasons, the gaps vary almost exactly as much as game-to-game luck would produce. 4% land in "extreme" territory, against the 5% you'd expect by chance. Across the full sample, extreme gaps weren't more common than chance would predict.
Pick a team to see its tougher and easier halves, season by season.
Show every game this season ▾
| Game | Spread | DAWG | Half |
|---|
Spreads are from the team’s side. Research seasons (2014–18) use research grades, not official DAWG grades. DAWG = points better or worse than Vegas expected, garbage time removed.
The takeaway
The next time your team needs a late field goal to escape a 24-point underdog, it isn't proof they played down. When they push a top-five team to the wire, it isn't proof they rose to the moment. Over hundreds of seasons, how strong the opponent looked going in doesn't explain how teams performed against expectations.
We haven't shown that no team ever plays up or down. We've shown that the pattern fans see follows that game's spread, and that it disappears once you measure the opponent another way.
How we did it
- The games. Regular-season games for today's 68 Power 4 programs, 2014 through Oct. 10, 2026: 9,857 team-games with a betting line and a DAWG grade. "Power 4" means today's 68 programs in every season, including years before some of them joined.
- The halves. Each team-season with at least 8 games is sorted by that game's spread and split in half. With an odd number of games, the middle game counts in neither half. Games with the same spread on both sides of the halfway line count in neither half, and each half needs at least 3 games.
- Opponent strength before kickoff. The average of a team's spreads in its earlier games that season, with home field taken out. Home field is estimated each season from the lines alone. This comparison uses games against Power 4 opponents once both teams had played two games, so the first weeks of each season aren't in it.
- Uncertainty. The ranges in Exhibit 2 come from resampling team-seasons 4,000 times. A range that crosses zero means we can't tell the difference from zero.
- 2014–18 grades are research grades: computed with the DAWG formula, not official DAWG grades.
Receipts
- How it was checked
- Exploratory: the question and how to measure it were chosen while looking at the data, not planned in advance. Every number in this article was then reproduced independently from the raw data by a second analysis that never saw the first one’s code or results.
- How this case changed
- LAB 001 started as a different question: do programs carry a play-down pattern from one season to the next? That version was planned in advance; it found no evidence that they do. We changed the question to the within-season version here after seeing those results. Before exploring it, we had already seen one related number: how much a team’s big-favorite results vary from season to season beyond chance, which was very little.
- Data
- Data through Oct 10, 2026 · snapshot sha256:52c915ceb1b9
- Method
- DAWG V3.0 · case method v1.0
- Research-only data
- 2014–18 research grades: Graded with the DAWG V3.0 formula unchanged, using frozen curve fits (the 2017 fit for 2014–17, the 2018 fit for 2018). The same script reproduces every published 2019–25 grade exactly. Lines for 2014–16 can’t be checked against an independent book. Includes research grades for 2014–2018: computed with the DAWG formula, not official DAWG grades.
Version history
- v1.0 · Oct. 18, 2026 · First published
No corrections. Any correction is dated here and on the corrections page.