Sharp money analysis: follow the bettors the books respect

Bettors come in two kinds. Squares bet on favorites, home teams, big names and gut feelings, and the sportsbook treats their money as noise. Sharps bet when the price is wrong, in sizes that matter, and the book treats their money as information: when a known sharp takes one side, the book moves the line, because it has just learned its number was off.

That is why a line can move with no injury news and no weather change. Someone the book respects just bet it.

• Sharp sportsbooks — a few books (Pinnacle and Circa are the classic examples) welcome winning bettors and keep their margins thin, so their prices reflect sharp opinion faster than anyone's.
• Exchanges and prediction markets — on Novig, ProphetX, Kalshi or Polymarket, bettors trade against each other and real money sits on each side. The price is set by whoever is willing to put the most behind it.
• Line movement — when retail books move toward a sharp book's number, the sharp side is usually the one the line moved toward.

The simplest sharp-money play is to find a retail sportsbook still offering yesterday's price on a market the sharp books have already moved, and to bet the sharp side there before it catches up.

A big win proves nothing; anyone can hit a parlay. What separates a sharp from a lucky square is closing line value (CLV): did they consistently get a better price than the one the market closed at? The closing line is the market's final, best-informed opinion. Beating it over and over is the one thing that reliably predicts long-run profit.

A sharp book makes money on volume with a small margin and limits nobody, so winning bettors go there and their money shapes its line. A retail book protects a bigger margin by limiting winners and leans on the sharp books' numbers to set its own. The upshot is that the sharp price is the one closest to the truth, and the retail price is the one most likely to lag.

Exchanges and prediction markets have the same property for a different reason: nobody is setting the price, so it sits wherever real money on both sides meets, and the depth behind it tells you how firmly the market believes it.

Put a retail book's price on one side of a market next to a sharp book's price on the same side, in implied probability. When the retail book pays more than the sharp reference implies, that is the sharp-money play: you are betting the side the sharp market favors, at a price the sharp market would not give you.

• Depth matters — a sharp price with $50 behind it is an opinion; one with $5,000 behind it is a position. Require real liquidity before trusting an exchange price.
• Check the other side — if the sharp book is also better on the opposite side, the two books simply disagree on the vig, and there is no signal.
• Timing — the edge is largest right after the sharp move and shrinks as retail books catch up, often within minutes.

• Beat-close rate — the share of bets that closed better than you bought them. Fifty percent is break-even; a sustained 55% or more is strong.
• Sample size — ten bets tell you nothing, fifty start to, a few hundred are convincing. Any ranking of bettors has to say how many picks it is based on.
• Lanes — a bettor can be sharp on NBA totals and ordinary on NFL sides. Measure by sport and market type, not by lifetime total.

• Signal set — Novig, ProphetX, Kalshi, Polymarket, OG and the sharp sportsbooks Pinnacle and Circa form the fair benchmark. Their devigged prices (worst-case devig, median across the set) set fair probability exactly as on the Positive EV board; only the book set differs.
• Placement set — FanDuel, DraftKings, BetMGM, Caesars, Hard Rock, Fanatics, ESPN BET, BetRivers, Bet365 and Fliff are the books whose prices are scored against that benchmark.
• What counts as a play — a row is shown only when a signal book holds the opposite side (so the sharp market has a real position against the retail price) and the best available price is not itself a signal book. The sharp side's resting liquidity is shown per book.
• Defaults — exchange liquidity of at least $2,000 behind the sharp side, so it is a position rather than a resting order someone forgot, and prices between −200 and +150. The board refreshes every 30 seconds.
• Same gate — the off-market gate and swap detector run first, so a stale exchange ask or a flipped market does not read as sharp disagreement.

• Discovery — a Polymarket wallet enters the tracked set when it places a single sports trade of $1,000 or more. The market index refreshes every ten minutes; the 600 highest-volume markets are polled in rotation, 60 every 20 seconds; tracked wallets are re-read every two minutes.
• Lanes — every market is classified into a (sport, market type) lane by a rule-based classifier that is deliberately willing to answer UNKNOWN rather than guess. A wallet is measured per lane, because sharpness in one lane says little about another.
• Picks — a pick is one entry into an outcome, tracked on an average-cost ledger that mirrors Polymarket's own average price, so partial exits and re-entries do not manufacture picks.

• Wallet metrics — adjusted ROI (realised P&L ÷ wagered), beat-close percentage, average CLV in cents, win-loss record, volume, and a 95% confidence interval on ROI. Rates are withheld below five picks; a wallet needs 20 scored picks before CLV can rank it.
• Sample-size dots — five dots at thresholds of 10, 30, 80, 200 and 500 settled picks, so the size of the evidence sits next to the number.
• History window — Polymarket exposes roughly one month of trade history per wallet; scoring further back is not yet built.

Ranking by lifetime profit mostly ranks bankroll. The board instead asks whether a wallet beats the closing line in a lane you care about, and every gate below must pass before a wallet is listed.

• At least 50 settled picks in a lane, at least $10,000 of volume, and at least 30 days active.
• Adjusted ROI above zero where it can be measured, and a beat-close rate of at least 50% where it can be measured.
• Tiered sort — proven edge ranks above unmeasured, which ranks above proven bad, with beat-close as the default order. The default window is 30 days (7, 90 and all-time are available), capped at 200 rows.

The gates exist because of a specific failure: before them, a wallet down $26,358 at −31.4% ROI, beating the closing line 38% of the time, was published as the number one sharp on the strength of its volume.