How to find the sharp money on a Polymarket game, step by step

At a sportsbook, sharp money is a guess: the line moved against the ticket count, so somebody the book respects must have bet the other way. On Polymarket it is a list. Every buy is a wallet, every wallet's settled positions are public, and 0xinsider grades each wallet S to F on its settled P&L. The S, A and B wallets are the sharp money, and the D and F wallets sit at the other end of the same scale.

The list's first surprise is in the chart above: the sharp money mostly waits for the game. Half of the S, A and B wallets' large sports buys landed after kickoff, and the game is where their grade earns anything.

Across 161,128 large Polymarket sports buys from June to September 2026 (the smallest about $3,000 until July 2, 2026 and $10,000 after), each graded as its wallet stood when it traded, the graded wallets' in-play buys came in at +1.22 points against the price and the D and F wallets' at -1.23: a gap of 2.45 points, on an interval from +0.50 to +4.40. Before kickoff the order flipped, -0.48 against +0.37, inside the noise.

So a sharp read before the game confirms the price, and a sharp read during the game is the one the data backs. That finding, from the sharp money study as corrected on October 7, 2026, orders every step below. None of the steps is a pick. The sports betting analytics guide turns a probability into a bet, and each signal here is one input to that probability.

Expect soccer. In the week from October 11, 2026, soccer had more games on the boards than every other sport combined, 688 against 636, and each NFL game carried hundreds of markets of its own. The board is where you narrow that down to one game and one market worth your time.

Open the sports boards and pick the league: NFL, NBA, MLB, NHL, soccer, tennis, UFC, esports and the rest. Each board lists the day's games with the Polymarket moneyline, spread and total for each, the kickoff countdown, and the live score once the game starts.

The prices are in cents, and a price is a probability. A team at 62¢ is a 62% chance in the market's view, and buying it pays $1 a share if it wins. There is no line to convert and no vig to back out; the spread between bid and ask on these markets is 1¢ at the median, and the sports taker fee at 50¢ is 2.5% of the stake, falling to 0.5% at 90¢ on most sports; NFL and college football games carry 0.03 and NBA games no fee, as of October 21, 2026. The fees guide has the schedule.

Before you read anything else, write down the probability you believe. Every step after this one checks that number against wallets that have been right before, and the check needs a number.

Under each game on a board is a block labeled Profitable wallets. It draws 3 bars for each side of the game. The sharp money bar is that side's share of the profitable wallets' money on the game. Market is the chance the side's price implies. Wallets is how many profitable wallets hold that side, split by grade, S first.

A sharp money bar longer than its Market bar means the profitable wallets back that side more than the price does. Bars of about the same length mean they agree with the market. One side's name and sharp money bar read green: the side the profitable wallets back against the price when they back one, otherwise the side holding most of their money.

Read the bars for what they are worth at that moment. Before kickoff, a well-graded consensus is already in the price. The fade study looked at 1,391 markets where 90% or more of the S, A and B money sat on one side before kickoff: that side was priced at 68.0¢ and won 66.7%.

So before the game, a long sharp money bar says the price is probably right. If it matches your own number, you have confirmation, and you still pay the full price.

During the game the bars mean something else. The graded wallets place half their sports buys in-play, and that is the one place the grade separated buyers. A sharp money bar that moves toward one side after kickoff, while the price is still catching up, is the signal the sharp money study measured.

Pro shows the bars live, with the wallets behind them: each profitable wallet's side, dollar size and win rate in the sport. Without Pro the bars run on a delay, and the card prints how long.

Click the game to open its event page. The header carries the same sharp money and Market bars, with the trader column beside them. The rail on the right has 2 tabs: Polymarket activity, the game's trade feed, and Top holders, the wallets holding each outcome with each one's grade. Switch the outcome to see the other side.

Pro and Max show up to 100 holders per outcome, each with its grade, and a toggle for S, A and B wallets. Game analytics require a paid plan.

The holders list is where the second cohort becomes visible. The fade study found that when 90% or more of the D and F wallets' pre-kickoff money sat on one side of a market, across 1,150 markets, that side was priced at 64.5¢ and won 62.3%: 2.25 points short of its price, with an interval that reaches zero. When the D and F money was split, the heavier side came in at its price.

So read the two columns together. A Top holders list with S, A and B wallets on one side and D and F wallets on the other is the study's disagreement case, 1,057 markets, where both sides came in close to their prices: +0.10 points for the graded side and -0.55 for the crowd side.

A list where both cohorts are on the same favorite is the agreement case, where the favorite lost about two points to a 67¢ price. A list with only D and F wallets, all on one side, is the crowd the fade study measured, and it is a reason to check the price rather than a bet. The betting against the public guide covers that habit in full.

The large-trade feed shows recorded large Polymarket trades admitted by the feed rules, live with Pro and 24 hours later without Pro: size, side, price, the market, and the grade the buyer has earned. Filter it to the sport or the game. Before kickoff it shows the money the timing study calls late: 22.8% of large sports buys land in the last hour before the game and another 22.4% in the first hour of play.

The first hour of play is the one to watch. From June 2026, graded as of each trade, S, A and B wallets came in at +2.23 points on their first-hour buys and D and F wallets at -1.70, 3.93 points apart on the same games in the same hour.

A run of S or A buys on one side in the minutes after a goal or an injury, while the price is still moving, is the signal with the most support in the data. A run of D and F buys on the side that just scored is the losing half of that same split.

For pre-game reading the delay costs little, because a buy from yesterday evening is the same fact today. For in-play reading it costs the signal, because a live price moves in seconds. The timing comes from the study of when large sports bets land.

A grade is a summary. Click any wallet in the holders list or the feed and its profile shows what the summary is made of: settled markets, win rate, average entry price, the calibration edge, realized P&L, and badges for the categories it has a record in.

A wallet graded A on baseball with 12 settled soccer markets is a strong wallet and a weak soccer signal. A wallet graded B with 900 settled tennis markets and a positive edge in tennis is the signal you want on a tennis game.

The wallet census says what the letters mean in sports dollars. Of wallets with 20 or more settled sports markets on October 11, 2026, S wallets were in profit 89.6% of the time with a median sports P&L of $155,419; B wallets 85.9%; D wallets 30.5%; F wallets 15.8%, median $4,087 down. The grade is built from that record, so this is a description of the letters rather than a forecast, but it tells you what kind of trader you are reading.

Two things to weigh with it. Size is not skill: the smart money vs whale size study found the dollar amount of a buy says less than the grade of the buyer, and a $50,000 buy from an F wallet is still an F wallet's opinion. And a short record is a short record. A wallet with 30 settled markets and a 70% win rate has shown you a good month; the sample-size section of the sports betting analytics guide has the arithmetic for why.

Pick of the Day runs this workflow every day and publishes the result, up to 25 picks a day. Its first 443 settled picks, with kickoffs from June 25 to October 10, 2026, won 63.9% at an average price of 60.3¢: 3.59 points over the price, on an interval from -0.66 to +7.84 that includes zero. On 443 picks, the record has not yet proved an edge.

Each pick is a position S, A and B wallets already hold, published before kickoff after its checks pass, with the side and price saved at publication. Wherever the capture succeeds, the archive records the market's last price before kickoff beside the entry, so each pick is graded on price as well as on result. On price the picks have come in slightly behind, 0.25 points a pick against the close across 436 measured picks, which the closing line value guide takes apart.

Read each pick two ways. As a pick, it is a game where the profitable wallets' side has been checked for you, at a recorded price. As a worked example, it shows the shape of a read: the side named, the price recorded before the game, the result from the market's own settlement, and the close beside the entry. The stats page breaks the record down by sport, month and entry price. At 443 picks, judge the method on price and result together.

Corrected on October 17, 2026: three Pick of the Day picks, from June 23, September 2 and October 1, 2026, were withdrawn from the record because they were published after their games had started. The figures here now leave them out. This section said 446 settled picks with kickoffs from June 23, won at an average price of 60.4¢: 3.47 points over the price, on an interval from -0.76 to +7.70.

Set the minimum where it changes what you get. The feed only admits trades of about $10,000 and up, so a lower alert minimum filters nothing, and in soccer even $25,000 still sends dozens of trades a day. Watching a feed through every game is a job; an alert does the watching. A custom alert fires when a large trade matches your filters: a category such as soccer or the NBA, a minimum size, a minimum grade, and whether the wallet is new.

Add a minimum grade for your sport and the first-hour runs from step 4 arrive as notifications instead of scrolling past; the custom alerts guide walks through the form. Following a wallet does the same for one trader. Follow a wallet from step 5 whose sports record you trust, as the real-time trade following guide describes, and its fills on your games arrive with the rest.

One signal clears zero: an S, A or B buy in the first hour of play, +2.23 points over its price on an interval from +0.10 to +4.35, which only just clears it. Every row in the chart above is graded as its wallet stood at the trade, and every pre-game signal sits inside its own noise, the graded wallets' consensus included.

So before the game, nothing on this site beats the price reliably. The profitable wallets confirm it, and a one-sided poorly graded crowd is a reason to check it. During the game, the strongest signal is a profitable wallet moving first, and at about 2 points it is small enough for the taker fee and the spread to take a real share of it.

Read the boards and the holders before kickoff to check the price. Read the feed after kickoff to find the move.

Buying a week early because the sharp money is on it. Large sports buys placed one to seven days before kickoff came in 2.19 points under their price from April to September 2026, on an interval that stays below zero, and the few placed more than a week out met the widest books in the chart above. The graded wallets' early buys came in above their price on too few buys to tell, so there is no case for following them early.

Following a large buy because it is large. The size of the stake is on the alert and the buyer's grade is beside it, and the grade is the one with a measured signal, mostly in play.

Treating a long sharp money bar before kickoff as a discount. It is a confirmation. Where 90% or more of the graded wallets' money sat on one side, the fade study found that side priced at 68.0¢ and winning 66.7%. Paying the spread to join it at the ask is paying to agree.

Treating 20 settled markets as a record. Of the wallets in the census with 20 to 49 settled sports markets, 43.4% were in profit, about the same share as wallets with 100 to 1,999 (40.9% to 44.6%), so a record that short does not separate skill from luck.