How often football underdogs win: 1,193 NFL and college football games at kickoff prices on Polymarket

Football upsets happen about as often as the kickoff price says they will. Across 1,193 NFL and college football games checked against ESPN's final scores, NFL underdogs priced at 33.0 cents won 32.2% of the time and college underdogs at 27.9 cents won 26.9%. The chart below splits them by season, and no season strays far from its price.

$1 on every underdog came out behind after Polymarket's taker fee, inside the noise. Backing upsets as a habit buys you nothing you can measure.

This study first ran on September 14 and counted 1,391 games. It kept a game when its moneyline's final volume reached 10,000, which it called $10,000. That volume includes the trading during the game, which nobody choosing a bet before kickoff can see, and Polymarket counts it in shares, not dollars.

On October 10, 2026 we kept a game only when $10,000 or more changed hands on its moneyline before kickoff. That drops 191 college games and 7 NFL games and adds none, so the sample is now 1,193 games, and the biggest upsets list lost the games that were thin before kickoff.

The dropped college games leaned toward upsets, as the chart below shows. A game that draws most of its trading during play is likely one where the underdog took the lead. UCLA's 5-cent win over Penn State on October 4, 2025 drew $5,026 before kickoff and most of its volume after it.

The finding stands: underdogs in both leagues won about as often as their kickoff price said. Everything after this section is the corrected study.

The NFL sample is 562 regular-season and playoff games from September 6, 2024 to September 10, 2026, 25 of them in the playoffs. The college sample is 631 games from September 11, 2025 to September 12, 2026, 44 of them bowl and playoff games. 66 games were played at a neutral site; they count everywhere except the home and road split.

Rates carry a binomial half-width. The edge is the win rate minus the average kickoff price, and its interval resamples games. The chart below shows what the export held and why the rest was left out: college football lost far more to thin trading than the NFL did.

NFL underdogs came in 0.80 points short of their average price of 33.00 cents over 562 games, with an interval from minus 4.59 to plus 2.95. College underdogs came in 0.98 points short of their 27.92 cents over 631, from minus 4.25 to plus 2.34. Win rates carry half-widths of 3.86 and 3.46 points.

So upsets come at the rate the market charges for them: about one NFL game in three and one college game in four. The favorites mirror it, 67.8% in the NFL at 67.0 cents and 73.1% in college at 72.1 cents.

Split by the favorite's kickoff price, no band was reliably too high or too low. College favorites priced 90 cents or more, 88 games at 95.3 cents, won 94.3%, so about one in 18 of them lost. NFL favorites at 80 to 90 cents won 87.0% of 54 games at 84.4 cents.

The widest gap was college favorites at 80 to 90 cents, who won 89.1% at 84.5 cents, 4.58 points ahead of the price over 101 games, and its interval runs from minus 1.62 to plus 10.21. The NFL's 10 favorites at 90 cents and up all won, too few games to read.

Oklahoma State beat Oregon 39-31 on September 12, 2026 at a kickoff price of 5.5 cents, the longest shot to come in among moneylines that traded $50,000 or more before kickoff. Sam Houston beat Oregon State at 7.5 cents on November 8, 2025, and UAB beat Memphis at 8.5 cents on October 18, 2025.

The NFL's biggest was the Panthers over the Packers on November 2, 2025, 16-13, at 13.5 cents, and its biggest playoff upset the Commanders over the Lions on January 18, 2025, at 18.5 cents. None of these says the price was wrong. At 5.5 cents the market expects about one such team in 18 to win, and across 631 college games, some will.

$1 on every NFL underdog at its kickoff price returned minus 4.68% before fees and minus 6.69% after the taker fee Polymarket charges on football games, with an interval from minus 18.72% to plus 5.99%. In college football the same bet returned minus 4.99% before fees and minus 7.15% after. Backing every favorite came out near even: plus 0.04% in the NFL and plus 0.62% in college after the fee.

Every one of those intervals includes zero, so none of it is an edge to chase.

With neutral-site games left out, NFL home underdogs won 32.1% of 221 games at 35.3 cents, 3.21 points short, and NFL road underdogs won 32.1% of 327 at 31.3 cents. College home underdogs won 29.5% at 30.8 cents and road underdogs 24.5% at 25.3 cents. None of the four clears zero, so the market already prices where a game is played.

The first cut, with every college game kept, told a different story: road underdogs priced at 20.4 cents won 19.3% of 658 games and $1 on each lost 21.85% after the fee, an interval that excluded zero. Most of that loss came from the 303 of those games whose moneyline traded under $10,000 before kickoff, where road underdogs were priced at 14.7 cents and won 13.2%. With them out, the loss is 10.13% with an interval from minus 33.86% to plus 20.00%.

Treat the kickoff price as a fair estimate of the upset chance. An NFL underdog at 33 cents won about a third of the time, a college underdog at 28 cents about a quarter, and a college favorite at 95 cents lost about one game in 18.

So back an underdog when you have a reason to think the price is wrong about that game, not because upsets happen. Then mind the fee. On NFL and college football games it is 0.03 × price × (1 - price) per share, so per dollar staked it weighs most on the cheap side, as the chart below shows. A limit order that rests on the book and fills pays no taker fee, and that matters most on an underdog.

It covers moneylines only. It does not model injuries, weather, or rest, and it does not say whether a spread bet on the same underdog would have done better.

The kickoff price is the last point in Polymarket's public price history. A buyer pays the ask, which sits above that price by up to half the spread, plus the taker fee. The fee here is today's football rate, 0.03, applied to every game. Until October 21, 2026 this study applied the 0.05 Polymarket lists for sports, which put each after-fee return here up to 1.7 points lower.

The NFL sample is parts of three seasons and the college sample one season and the start of another, so price bands of 50 to 200 games carry half-widths of five to nine points, as the chart below shows. Games whose market settled on Polymarket without a recorded result in 0xinsider's database are not counted.

A Polymarket moneyline pays $1 a share on the team that wins, so a team at 30 cents is a 30% chance. We took each game's price at kickoff: the last point in Polymarket's public price history before the game started, a median of 0.9 minutes before. The favorite is the team priced above 50 cents at that moment, the underdog the team below it, and an upset is the underdog winning.

Every game was matched to ESPN's final score, season type and venue. ESPN listed all 679 NFL games and 1,043 of 1,049 college games, and Polymarket's settlement named ESPN's winner on every one of them except a 2024 preseason NFL tie. Preseason games are left out, 109 of them in the NFL.

We also left out 404 college games and 7 NFL games whose moneyline traded under $10,000 before kickoff, counted from Polymarket's public trade history up to the same cutoff as the price. A market that thin can print a last price nobody could trade at size: the South Dakota Mines and Drake moneyline had no trade at all before kickoff.

College moneylines are thin far more often than NFL ones, as the chart below shows. We set the floor after a first cut that kept every game, and the one result it changes is reported above.

The NFL board and the college football board list live and upcoming games with an active Polymarket market, each with both teams' prices and the move over the last 24 hours.

On Pro, a game's event page adds a Markets board: the moneyline beside the spread, total and prop families, each with its volume. The large-trade feed shows fills of $10,000 or more, live on Pro and 24 hours behind without it. The NFL and college football guides cover every market on a game and how it settles.

Past the moneyline, spread and total, a football game's other markets drew under 2% of its volume in the chart below, so read their prices as thin.

The study rests on five published runs: a read-only export of the settled moneylines at 23:30 UTC on September 13, 2026, a match of each game to ESPN's public scoreboard, a pull of each game's kickoff price from Polymarket's public CLOB API, a pull of the dollars traded on each moneyline before kickoff from Polymarket's public trade history at 11:44 UTC on October 10, 2026, and an analysis script that reproduces every table from those four files.

The season split and the pre-kickoff bands come from one more script over the same files, and the volume by market family from a read-only query at 12:23 UTC on October 11, 2026. All of them are at github.com/0xinsider/research, under polymarket-sports-markets, in the files whose names carry the word football. The chart below is the check against ESPN.

The universe is one settled full-game moneyline per NFL or college football regular-season or postseason game that traded $10,000 or more before kickoff and settled on ESPN's winner. Expect the count to grow every week of the season.