Laying or taking the points? 30,158 large Polymarket spread bets, and how often games are decided by one

Sportsbook folklore says take the points: the public backs the team giving them, so the value sits with the team getting them. On Polymarket in 2026 it ran the other way. Across 30,158 large spread buys from April 2 to September 13, laying the points came in 4.68 points ahead of the price and taking them 2.82 behind.

Basketball, where most of that money went, shows it plainest in the chart below: two sides at almost the same price, and only one of them won more often than it paid for. So judge a spread on its price, and drop the habit of taking points because they are points.

The count covers every Polymarket spread market with a title in that form, a kickoff on record and a settled result, from the first on March 26, 2025 to September 13, 2026: 79,769 markets on 23,765 games. Soccer holds half of them because it has half the games, 12,402.

The pricing half uses the large-trade alerts from April 2, 2026, the first day the alert table's outcome labels are trustworthy (the favorite-longshot bias study documents why): 30,158 buys by 2,392 wallets across 2,972 markets, $565.9 million in stakes, 28.6% of them after the start. The smallest recorded buy was about $3,000 until July 2, 2026 and $10,000 after.

Edge is the share of buys that won minus the average price paid, in percentage points. Intervals resample markets rather than buys, because thirty large bets on one line are one result.

The laying edge rests on 13,902 buys on 2,030 markets, and its interval runs from plus 0.25 to plus 9.12 points. Taking the points, over 16,256 buys, runs from minus 6.84 to plus 1.28. The chart below splits both by sport.

The low end of the laying interval sits a quarter of a point above zero. That is enough to stop paying extra for the team getting points, and not enough to lay every spread you see.

Basketball here means the NBA: 13,314 of its 13,886 buys. Laying the points came in plus 5.32 and taking them minus 5.41, plus 11.8% and minus 10.2% on the dollar. Both intervals stop just short of clearing zero.

Soccer is mostly the 2026 World Cup, 10,568 of 13,243 buys on 355 World Cup markets, and its buyers put most of their money on taking the points, as the chart below shows. Taking came in minus 0.74 and laying plus 5.11, both inside the noise.

MLB run lines came in behind on both sides, minus 0.64 laying and minus 3.13 taking. NHL puck lines split widest on the fewest bets: plus 9.21 laying and minus 7.65 taking, across 443 buys.

Football is missing for a reason: the pricing window starts April 2, after the 2025 NFL and college football seasons ended, and the 2026 seasons had just begun by September 13. The whole window holds 94 large football spread buys on 54 markets, too few to read.

Laying came in ahead in both phases. Before the start it paid 48.7 cents and won 52.9% of the time, plus 4.20 points on an interval from minus 0.85 to plus 9.11. During play it came in plus 6.00 over 3,700 buys, from plus 2.01 to plus 9.93: the only split by phase, price or grade that clears zero on its own.

Taking the points came in minus 3.03 before the start and minus 2.35 during play. A live spread moves with the score, and in this window the buyers backing the team giving points during the game were paid for it more often than the price said. The two phases' laying edges sit within two points of each other, so read this as one lean measured twice, not two.

Most large spread buys land between 30 and 70 cents, and that is where the lean lives. Laying the points at 30 to 50 cents came in plus 5.64 over 5,313 buys, and at 50 to 70 cents plus 4.53 over 7,250. Taking the points at 50 to 70 cents, the busiest band of all with 8,039 buys, came in minus 4.78.

At 70 cents and up, taking the points came in ahead too: it paid 83.9 cents and won 86.2% of the time, plus 2.36. No single band's interval clears zero; the bands are the all-sports result cut four ways.

Corrected October 7, 2026. This section first graded each buy with a ranking row 0xinsider can rewrite after the trade, and reported S, A and B wallets at plus 5.46 laying the points. The figures now grade each buy as its wallet stood when it traded, from June 1, over the markets settled by October 7, 2026. The pattern holds; the numbers moved.

Laying came in ahead for both cohorts by about the same amount, plus 2.67 points for S, A and B wallets and plus 3.54 for D and F. So the lean belongs to the side of the market, not to the skill of the trader. A wallet's grade will not tell you which side of a spread to take.

Taking split wider, plus 1.46 for the graded wallets and minus 6.14 for the poorly graded ones. Every interval here spans zero on a few hundred markets per cohort, and a gap between two noisy cohorts is noisier still.

A run line of -1.5 loses whenever the game is decided by one run, whichever team wins. Polymarket listed minus lines for both teams in 74.1% of the MLB games with spreads, and a -1.5 line for both teams in 1,882 games, where neither covering means a one-run game. Of those 1,882 games, 548 were decided by one run: 29.1%, with a half-width of 2.05 points.

The rate holds through the season: 26.3% in April, between 29.3% and 29.9% from May through July, 27.3% in August, and 33.9% over the first 168 games of September.

That is the whole run-line trade. Backing a team at -1.5 instead of on the moneyline gives up every win by exactly one run, about three games in ten, for a better price. The price has to be better by enough to cover the one-run wins it gives up.

Soccer lists handicaps for both teams in 80.3% of matches, which makes the winning margin countable. On 7,528 matches listing -1.5 and -2.5 for both teams, with the match's draw market, 25.7% ended level, 38.6% were won by one goal, 20.2% by two, and 15.6% by three or more.

So a -1.5 handicap, a bet on a two-goal win, loses to a draw and to every one-goal win. Across the 9,016 matches that listed -1.5 for both teams, one team or the other won by two or more 35.3% of the time. A single team's chance is lower, and that is the number to read the handicap price against.

Know the margin you are buying. A -1.5 run line loses every one-run game, about three in ten, including the ones its team wins, and a -1.5 soccer handicap loses to every draw and every one-goal win.

In the NFL the line you pick matters most, because margins bunch on 3 and 7. 15.1% of the NFL games in the chart below were decided by exactly 3 points, so laying -3.5 instead of -2.5 gives up every one of those wins. When a game carries several lines, price each half point against how often the margin lands on it.

And judge a side on its price. Large buyers who took the points came in behind the price in 2026, most clearly during play, so the team getting points is not the value by default.

It does not say how often favorites cover. Polymarket lists a ladder of lines for both teams, and this study does not pick one line as the game's line; the price large buyers paid is the measure it uses instead.

Its sports come from particular stretches: NBA games from April 2 to June 14, soccer mostly from the World Cup, MLB from April to mid-September. Football pricing waits for a season of data, and a rerun over one could change the picture.

The one-run and margin counts need games that list both teams at the same line, and the chart below shows how few sports do. MLB and soccer do in most games. The NHL, basketball and football do not, so their subsets are not counted as their sports.

Many splits were tested here, which raises the odds that one clears zero by chance. Every position is scored as held to settlement at the buyer's price, before fees.

Polymarket lists a point spread as its own market, titled with the team that gives the points: "Spread: Detroit Tigers (-1.5)", with the two teams as the outcomes. The rules resolve it to the Tigers "if the Detroit Tigers win the game by 2 or more runs", and to the other team otherwise, a tie included. Buying the named team is laying the points. Buying the other team is taking them.

Polymarket often lists several lines for a game, for either team. The Bills and Texans game on September 13, 2026 carried 35 spread markets, from Bills -0.5 to Bills -21.5 and from Texans -1.5 to Texans -21.5, all listed before kickoff. So the team at minus on a given line is not always the team the market favored, and a count of how often favorites cover would depend on which line you call the game's line.

The price settles that. Every large spread bet was bought at a price, and every market settled, so whether laying or taking the points was the cheap side is a comparison of prices with results.

Spreads are a football market. Outside the NFL and college football they draw a small slice of a game's volume, as the chart below shows, so football is where the most-traded lines are.

On Pro, each game's event page opens a Markets board: every spread line with both teams' prices and its own volume, under a Volume by family bar that ranks the game's market families. Without Pro the board shows sample data.

The large-trade feed shows fills of $10,000 or more with each wallet's grade, live on Pro and 24 hours behind without it. The point spread guide covers how spreads, run lines, puck lines and soccer handicaps settle.

The pricing, the market counts and the MLB and soccer margins come from one read-only run of a set of queries against the production database at 20:45 UTC on September 13, 2026, and the grade split from a rerun on October 7, 2026. The NFL margins come from ESPN's final scores for the games in the football underdogs study, and the volume by league from a read-only query at 12:20 UTC on October 11, 2026.

All of it is public at github.com/0xinsider/research, under polymarket-sports-markets, in the files whose names carry the word spreads: the query text, the raw output, the market-level export and the bootstrap output. The export is committed, so the intervals reproduce with the bootstrap alone and no database. The chart below is the direction check: wherever the named team covered and the game had a moneyline, that team won.

The market universe is every settled Polymarket spread market titled "Spread: <team> (-X.5)" whose first outcome is that team, with a kickoff on record, in the eight sports categories named above. The pricing universe is every large-trade alert for a Polymarket buy on one of those markets, settled after the trade, at 2 to 98 cents, from April 2 to September 13, 2026.

The grade is the latest daily ranking dated on or before the trade day and last written at or before the trade (re-run October 7, 2026; the files whose names start grade-at-trade-2026-10-07). Expect the counts to grow and football to fill in as its season runs. The laying lean is the figure a rerun is most likely to move, since its interval sits close to zero.