Betting against the public: why it works at sportsbooks, and what Polymarket shows

When the poorly graded money on Polymarket goes one way, it goes to the favorite. In 1,150 sports markets where 90% or more of the D and F wallets' pre-kickoff money sat on one side, that side cost more than 50¢ 904 times. On this much, the folklore has the crowd right.

Sportsbooks publish, or leak, the share of bets on each side of a game. When 75% of tickets are on the Cowboys, the Cowboys are the public side, and betting against the public means taking the other team. The logic is old and simple: casual bettors back favorites, home teams, famous teams and overs at whatever price is offered, so a side most of them want has a price pushed past its probability.

The habit has other names: fading the public, contrarian betting, going against the square money, where a square bets on feel and a sharp bets on price. All of them describe one trade: find the crowd's side, and take the other one when the price is right. That last clause is the whole skill. An unpopular side is not good for being unpopular. It is good when its price is below its probability.

A sportsbook does not try to predict the game. It sets a price at which the money on each side leaves it a margin whichever side wins, and where it cannot balance the money it shades the price toward the side it expects the public to want. The Cowboys open at -3, the book knows the public will bet them, and it hangs -3.5. That half point is the public's first tax, and the bettor on the other side collects it.

The public pays the second tax to itself. Bettors overweight small probabilities and love big payouts, so longshots and long parlays are overpriced wherever they are sold; the favorite-longshot bias guide covers the racetrack and sportsbook evidence. They like overs and the teams they watch on television too, and each preference pushes a price a little past its probability.

Polymarket's large buyers paid neither tax in a form you can measure. There is no book to shade the line, and no price band in the chart above, longshots included, has an interval that excludes zero.

At a sportsbook both taxes are real and both are small: a half point of spread, a point or two of probability. The habit works there over hundreds of games or not at all.

The ticket percentage counts bets and the money percentage counts dollars. When 75% of tickets are on one side and only 45% of the money, a few large bets are on the other side, and a sportsbook bettor assumes they belong to bettors the book respects. Reverse line movement is the same split read off the price; the line movement guide covers steam, reverse movement and what each one means.

On Polymarket you can see who writes the big tickets, and the biggest came from wallets with no graded record at all. Their share of the dollars beat their share of the buys by more than any other cohort's did, the S, A and B wallets included. A money-against-tickets read would have flagged the newcomers before the sharps.

Both kinds of big ticket came in at their price: the ungraded wallets at +0.68 points and the S, A and B wallets at +0.50, each on an interval that includes zero. So the money percentage marks size, and size is not skill. The habit is a proxy for knowing who is on each side, and on Polymarket you do not need the proxy.

The public is on favorites, and favorites win. Fading every popular side means betting a lot of underdogs, and an underdog at +150 that wins 38% of the time was priced right; the crowd on the other side did not make it cheap. A fader who does not check the price against a probability of their own is a longshot bettor with a theory.

The shading is often already in the price the fader gets. If the book hung -3.5 expecting public money, the bettor taking +3.5 collected the public's half point; the bettor taking +4 a day later, after the line moved again, may have collected nothing. And the fader still pays the book's hold, 4.55% at -110, on every bet.

One month proves nothing, and Polymarket's own crowd shows why. Across the whole fade study, the side the poorly graded money agreed on trailed its price by 2.25 points, and the chart above shows where that came from: June, the month with the most one-sided markets. Take June out and the crowd's side came in at its price. The sample-size section of the sports betting analytics guide has the arithmetic, and a fader who started in July had nothing to collect.

Polymarket removes the inference. Every buy on a Polymarket sports market is a wallet address, every wallet's settled positions are public, and 0xinsider grades each wallet S to F on its settled P&L. There is no ticket count to read through: you can see whether the money on each side belongs to wallets that have been right or wallets that have been wrong.

The fade study uses that to test the habit directly. It takes every large buy on a Polymarket sports moneyline, spread or total from June 1 to September 12, 2026, placed before kickoff, and grades each buyer by the record it held when it bought.

For each settled market it adds up the D and F wallets' dollars on each side. Where those wallets had $5,000 or more on the market across 2 or more buys, the study asks how one-sided their money was, and how the side they favored did against its price.

The one-buy answer is what the sharp money study found: a single D or F buy before kickoff came in at +0.40 points against its price, a single S, A or B buy at -0.76, and the grade separates nothing before the game. The market-level answer leans the other way, and only a little.

In 1,150 markets, 90% or more of the D and F wallets' pre-kickoff money sat on one side. That side was priced at 64.5¢ on average and won 62.3% of the time: -2.25 points, with a 95% interval from -5.11 to +0.60.

In 592 markets the D and F money was 70 to 90% one way, and that side came in at -0.85 points with an interval across zero. In 767 markets it split under 70%, and the heavier side came in at -0.05. The penalty rose with the agreement, and only the top band is far from zero.

In dollars, a flat stake on the crowd's side in the 1,150 one-sided markets, held to settlement, returned -3.81%. A flat stake on the other side, priced at one minus the crowd's average entry, returned +15.94% before fees, a figure that swings on a few cheap contracts and shares the edge's interval.

Read the interval as seriously as the point. -2.25 with an upper bound of +0.60 is a lean, not a finding. This study first published minus 3.68 with an interval clear of zero; it graded buyers with ranking rows 0xinsider can rewrite after the trade, and graded at the trade the penalty shrank. So a one-sided poorly graded crowd is a reason to check the price on the other side, not a bet on its own.

Split the 1,150 markets by which side of 50¢ the crowd sat on. In 904 markets the D and F money was on the favorite, at 70.4¢ on average, and the favorite won 68.8% of the time: -1.59 points, interval -4.80 to +1.61.

In the 246 markets where the crowd was on the underdog, at 42.9¢, the underdog won 38.2%: -4.68 points, interval -10.86 to +1.50. The first version of this guide said the crowd's error was the favorite; graded at the trade, neither side's interval clears zero and the underdog crowd did worse on a quarter of the sample.

By sport, among those with 40 or more one-sided markets, the crowd side came in at -2.86 points in soccer and -2.08 in baseball, and at its price in tennis (+0.48) and esports (-0.22).

The same split for the S, A and B wallets finds nothing to fade. In 1,391 markets where 90% or more of the graded wallets' money sat on one side, that side was priced at 68.0¢ and won 66.7%: -1.25 points, interval -3.61 to +1.11. Their consensus before kickoff was in the price. It was neither a discount nor a tax.

Where the two cohorts disagreed, in 1,057 markets with $5,000 or more from each, the graded side was priced at 51.7¢ and won 51.8%, +0.10 points, and the crowd side was priced at 48.7¢ and won 48.2%, -0.55. Where they agreed, 1,193 markets on the same 67¢ favorite, both lost about 2 points. Taking either side of a disagreement was a fair price, no better.

Identification is the difference. A sportsbook fader bets against ticket counts and hopes the sharps are on the other side. A Polymarket fader can see whether they are. Before kickoff the data says the graded wallets confirm the price, and a poorly graded crowd alone is a reason to look again, not a reason to bet.

Most games you open will not have a crowd to fade. Of the settled sports markets with a large buy before kickoff, about one in seven had $5,000 or more of D and F money sitting 90% or more on one side. Treat the pattern as a filter you run across a slate.

Open the game from its league board, MLB, tennis, esports, soccer or the rest, and read the rail's Top holders tab. It lists the wallets holding each outcome with the grade each has earned; switch the outcome to see both sides.

A column of D and F wallets on one side before kickoff, with the S, A and B wallets absent or on the other side, is the study's 1,150-market pattern. Both cohorts on the same favorite is the agreement case, and graded wallets alone is the price. The Profitable wallets block on the board and the event header draws the graded side directly, as a sharp money bar beside a Market bar.

The large-trade feed shows each large buy with the buyer's grade. With Pro the feed is live, so a run of D and F buys onto one side of tonight's game shows as it forms; without Pro it runs 24 hours behind. The sharp money guide covers all 3 pages in order.

Then check the price. The study scored the fade at the complement of the crowd's average entry, and the price at kickoff may have moved. A fade is a bet that the other side is cheaper than its probability: write down the probability, read the ask, and take it only when the gap is there after the fee. A crowd on the other side is the reason to look. The price is the reason to bet.