Where the numbers come from
Everything here comes from Polymarket, a public prediction market where people bet real money on sports and every result settles in the open.
We pull two things. The first is Polymarket's own leaderboard, which reports each trader's profit and total money wagered. The second is Gamma, Polymarket's market data feed, which tells us when a game starts, whether it has finished, and how it settled.
We have no private information, no insider access, and no model that predicts games. We watch public money and we keep score.
What a signal is
A signal is a bet that a tracked trader actually placed.
When one of them buys a position in a sports market, we record what they bought, the price they paid in cents, how much they put up, and when they did it.
A signal keeps updating while its market is still open. If more tracked traders take the same side, or one of them adds to a position, the combined stake and the number of sharps behind it both grow. Once the market settles, the record locks and never changes again.
A price in cents is what the market thinks a bet is worth. A side trading at 60 cents means the market gives it roughly a 60% chance to win. Buy at 60 cents, and a winning bet pays you a dollar.
Why the numbers keep getting smaller
We follow the 100 traders at the top of Polymarket's All Time sports leaderboard.
21 have enough history for us to test whether their results are skill.
7 pass that test.
Nobody needs 100 people's bets. The number shrinking at every step is the filter doing its job.
What consensus means
Consensus means more than one tracked trader landed on the same side of the same market.
One sharp bettor taking a position tells you something. Three of them arriving at the same side separately tells you more. We group those together and show you the combined amount at stake along with how many people are behind it.
Edge Rating
Edge Rating is our score for whether a trader's results look repeatable.
Raw profit is easy to produce with luck and a large bankroll, so Edge Rating adjusts for two things. It accounts for how much risk a trader took to earn their return, which stops one huge win on a small sample from outranking steady work over thousands of bets. It also pulls short track records toward the group average, because a hot streak over 40 bets is not evidence of skill.
A positive Edge Rating means a trader beat the market after those adjustments. Clearing that bar is not the same as proving skill, and the next section covers the harder test.
Confirmed edge
Confirmed edge is a much higher bar, and almost nobody clears it.
Here is the problem it solves. Put 100 people in a room flipping coins and somebody will hit heads six times running, which proves nothing about that person. Test enough traders for skill and a few will look brilliant by pure chance in exactly the same way. Statisticians correct for that, and we apply the correction across every wallet we score.
7 wallets currently clear it, out of 21 with enough history to be tested at all. Their results are unlikely to be luck even after accounting for how many traders we checked. The rest have not proven anything yet, and we say so on their pages.
Price beats picks
Every bet has a price, and the price decides whether winning often enough is even sufficient.
We ran 2,499 settled consensus bets two ways. The first risks a flat $100 every time. The second sizes each bet according to its price. Same picks in both, and the outcomes split apart.
| Price range | Settled bets | Win rate | Flat return | Return when sizing by price |
|---|---|---|---|---|
| Under 30c | 181 | 9.9% | -56.3% | -47.7% |
| 30 to 60c | 1,070 | 43.8% | -9.9% | -8.1% |
| 60 to 85c | 898 | 76.1% | +5.1% | +6.5% |
| 85c and above | 350 | 94.6% | +4.3% | +3.7% |
Both price bands at or above 60 cents made money, and both bands below it lost money. Longshots feel good and pay badly. Those are not our opinions, and you can check every one of these bets on the performance page.
Pre-game, in progress, and finished
Every market here carries one of four states, and we never guess between them.
A market is pre-game before the event starts, in progress while it is being played, and finished once it is over. When Polymarket's data does not clearly say which, we mark it unknown instead of picking one.
We show two kinds of open market. A pre-game signal means the event has not started, so the price is closest to what the sharp actually paid. A live signal means the game is underway and the market is still trading, so the price has moved since that bet went in. We label them differently for exactly that reason. Once a game finishes, or once the market closes, the signal becomes a record rather than something to act on.
The leaderboard, and what it does not control
The leaderboard and the signal feed do two different jobs, and the difference is worth knowing.
Signals come from the 100 traders at the top of Polymarket's All Time sports leaderboard, and that set shifts as the rankings move.
The leaderboard page shows Polymarket's rankings across four windows: All Time, This Month, This Week, and Polymarket Today. Those shorter windows can surface traders who sit outside the tracked hundred, so you may see somebody ranked highly this week whose individual bets we are not following.
Polymarket reports each trader's profit and total money wagered. We calculate the third column ourselves, which is return on volume, and it is the one worth reading. A trader who made $2 million on $200 million of volume ran at 1%. A trader who made $500,000 on $5 million ran at 10%. The second one is better at this, and raw profit hides it.
Receipts
We publish how every tracked bet finished, including the losers.
The performance page runs on 57,327 settled bets. Every signal page shows how that specific bet ended once its market resolves. Nothing gets pulled down for being embarrassing.
Most services skip this part, and it is the only way you can check whether any of this works.
How to actually use this
- Start with pre-game signals because their prices are closest to what the sharps paid. We also show open markets after play begins, but every Live card tells you that its tracked entry is not the current market price.
- Read the price before you read the pick, since our own data says the price does most of the work.
- Treat anything under 60 cents as a bet swimming against a losing historical pattern.
- Give more weight to a market where several tracked traders landed on the same side than to one person acting alone.
- Come back and hold us accountable, because we publish the record either way.
What this does not tell you
A few limits, stated plainly.
We do not predict games. We report what other people bet and how it turned out. A positive Edge Rating says nothing about whether a trader's next bet wins. Prices move fast, so the price a sharp paid may be gone by the time you see it. We only see Polymarket, which means money moving through traditional sportsbooks is invisible to us.
We follow the All Time top 100, so a trader having an enormous month who has not yet climbed that far is invisible to our signal feed. You may spot them on the leaderboard's shorter windows without ever seeing their individual bets here.
The staking comparison above rests on 2,499 settled consensus bets. That is a real sample and it is not an infinite one, so treat the size of the effect as an estimate rather than a constant. Past results, very much including ours, guarantee nothing about the future.
Free and Pro
Free accounts see 3 full signals plus 5 locked previews with the details hidden, leaderboard ranks 1 through 10 on the All Time window, and filtering by bet size.
Pro opens up to 100 signal rows per feed request. The default board shows open pre-game and Live markets, while Show closed adds the historical record. Pro also includes leaderboard ranks 1 through 100, all four time windows, sport filters, every leaderboard sort, Sharp Score threshold filters, and the detailed statistics behind each trader's Edge Rating.
Glossary
- Signal.
- A bet a tracked trader actually placed, recorded with what they bought, the price, the amount, and the time. It keeps updating while the market is open and locks once the market settles.
- Consensus.
- More than one tracked trader landing on the same side of the same market.
- Price in cents.
- What the market thinks a bet is worth. A side trading at 60 cents implies roughly a 60% chance. A winning bet pays a dollar.
- Price band.
- A group of bets that shared a similar price, which lets us compare how cheap bets performed against expensive ones.
- Return on volume.
- Profit measured against total money wagered rather than as a dollar figure. Making $500,000 on $5 million beats making $2 million on $200 million.
- Risk-normalized return.
- Profit compared against how much risk was taken to earn it. Two traders can both return 10%, and if one did it on even-money favorites while the other fired at longshots where most bets die, they did not do the same thing. We score the first one higher.
- Shrinkage.
- Pulling a short track record toward the group average. A trader who is 30 for 40 looks incredible, and 40 bets is nowhere near enough to separate skill from a hot streak. Shrinkage treats a small sample as partly luck and moves that score toward the middle until there is enough history to say otherwise. The more bets somebody places, the less we shrink. The formal name is empirical-Bayes shrinkage.
- Edge Rating.
- Our score for whether a trader's results look repeatable, combining risk-normalized return with shrinkage.
- Bonferroni correction.
- Raising the bar when you test many people at once. Test one trader and a strong result means something. Test a hundred and a few will look brilliant by pure chance. The correction makes the standard harder in proportion to how many were tested.
- Confirmed edge.
- A trader whose results survive the Bonferroni correction, meaning the performance is unlikely to be luck even after accounting for how many traders we tested.
- Eligible wallet.
- A tracked trader with enough settled bets for a skill test to mean anything. Traders below that threshold are not tested.
- Pre-game, in progress, finished, unknown.
- The four states we assign a market. Open pre-game and in-progress markets appear on the current board, with pre-game signals ranked first when space is limited. Finished, closed, and unknown markets belong to the historical record. Unknown means Polymarket's data did not clearly tell us, and we refuse to guess.
- Receipts.
- The published record of how tracked bets finished, wins and losses both.
- Settled bet.
- A bet whose market has finished and whose result is recorded as a win or a loss.
- Tracked cohort.
- The current 100 traders at the top of Polymarket's All Time sports leaderboard whose activity feeds the signal board.
- Sharp Score.
- A score used to sort and filter signals by wallet rank, bet size, and recency. Pre-game signals rank before Live signals when a current surface has limited space.
- Max bet.
- A tracked stake more than five times that wallet's median recent BUY ticket. For consensus, we compare the combined stake with the participating wallets' medians added together.
- Big bet.
- A tracked stake more than twice that wallet's median recent BUY ticket. We only show this badge when every needed wallet has at least 50 tickets in its fresh behavior profile.
- Closed.
- A Polymarket market that no longer accepts trades. Market state is separate from game phase, so a game can be in progress while its market is already closed.
- Polymarket.
- A public prediction market where people buy and sell positions in outcomes with real money.
- Gamma.
- Polymarket's market data feed, which supplies market timing and settlement information.