EdgeFinder asks that question of every sports market it can match to a sharp line. It takes the price the sharpest sportsbook in the world is quoting, strips out that book’s profit margin, and checks whether Polymarket is selling the same outcome for less.
The answer so far is no — and the interesting part is by how little. This page shows the measurement rather than the conclusion, so you can disagree with it.
Four steps. Plain words first, the language of the field second.
Pinnacle takes the largest bets from the sharpest bettors, so its price is the closest thing to a true one that exists.
Odds are pulled per event from The Odds API. Where Pinnacle has no line, the fallback is the median devigged probability across the other sharp books, each devigged individually first so a fat-margin book cannot drag the consensus.
A book's prices always add up to more than 100% — the extra is its profit margin. Removing it leaves the probability the book actually believes.
Devigging maps raw implied probabilities q to fair p summing to 1. The default is the power method, p_i = q_i^k, which shrinks longshots harder and matches the observed longshot bias better than proportional scaling.
Line up the fair probability against what a share actually costs on Polymarket right now.
Events are matched by fuzzy team/fighter name and start time, and every match carries a confidence score. Unverified matches are recorded and excluded rather than silently traded.
Subtract the fee, the gap between the buy and sell price, and the slippage from filling a real size — then rank what is left.
EV per share is p − c where c folds the taker fee into the price, walking the order book to the size actually being bought. Survivors are staked by fractional Kelly under portfolio caps.
Nothing has been measured yet.
The desk reads a snapshot of EdgeFinder’s SQLite ledger through the private analytics interface, and either that snapshot is missing or no scan in it has priced an outcome. Run npm run edgefinder:export and make sure ANALYTICS_BASE_URL points at it. Showing nothing is deliberate — an empty desk would read as “the scanner found nothing”, which is a different claim.
The costs are measured, not assumed. Spread and fee come from the order book at scan time, so a Polymarket fee change or a structurally tighter book would move the toll and could reopen the taker route on its own.
Matching can be wrong. Events are paired by fuzzy name and start time. A bad match produces a fair line for the wrong contest; unverified matches are recorded and excluded rather than traded, which protects the result but also means the scan sees fewer markets than exist.
Pinnacle is treated as the truth. The whole method rests on the sharp line being closer to correct than Polymarket. Where Pinnacle has no line the consensus median stands in, which is a weaker instrument — the fair-source breakdown above says how often that happened.
The maker route is unsettled, not proven. It prices positive because it pays no fee, but a resting order only fills when someone chooses to sell into it. Until the ledger has a usable sample of real fills, its edge is a quote, not a return.
A bookmaker's prices deliberately sum to more than 100%. That surplus is its margin. Devigging removes it and rescales to a real set of probabilities.
One that welcomes large, informed bets instead of banning the people who place them. Its line moves on information, which is what makes it worth copying.
A taker crosses the spread and buys immediately, paying the fee. A maker posts a price and waits, paying nothing — but only trades when somebody chooses to trade against them.
Whether the price moved toward you or away from you between your bet and the event starting. It is the highest-power test of an edge available, because it uses every fill rather than only the settled ones.
How many standard errors a result sits from zero. Below about 2, the result is indistinguishable from chance no matter how large it looks.
The bet size that maximises long-run growth for a given edge. Used here at a fraction of full Kelly, which is what a serious bettor does, because full Kelly assumes the edge is known exactly.