How no-vig odds work

A sportsbook price includes a margin (the vig). The NO-VIG column removes it: our estimate of the true probability of each side, built in the open from the sharpest prices we receive. Here is exactly how.

1. Sources

The no-vig price is pooled from 4 sources: Pinnacle (PIN, base weight 1), Circa Sports (CIR, base weight 1), Polymarket (PM, base weight 1), Kalshi (KAL, base weight 1). These books and exchanges move first and take large bets, so their prices carry the most information. They feed the no-vig price; Kalshi is also ranked as a book, on by default, at its after-fee price, where it is allowed and at least $2,000 is available at that price on that side. Polymarket's two order books (international and US) count as one source, PM: together they weigh what the heavier of the two would alone.

2. De-vig each source

Each source's two prices are converted to implied probabilities, and the margin is removed the way that source prices that market:

Moneylines (Pinnacle, Circa): power. We find the exponent k so that pA^k + pB^k = 1. Sharp books put proportionally more margin on longshots than on favorites (the favorite-longshot bias); the power method allows for that, and in our own test on 725,000 results it beat multiplicative and Shin on Pinnacle’s prices.
Soccer three-way (home / draw / away): power across all three prices, so the three no-vig probabilities sum to 100%.
Spreads and totals: power too. Near even (−110 / −110, −105 / −115) every method gives the same answer to within a tenth of a point. A juiced line, such as a puck line at +200 / −250 or a live total at −300 / +240, is really a moneyline on covering, so the longshot side loses more of the margin, as on moneylines.
Books backup: multiplicative on US-league moneylines, otherwise power. When no sharp source has a usable price, the board shows a Low-confidence no-vig price from the median of the US books, each de-vigged on its own. Two-way moneylines in the NFL, NBA, MLB and NHL (and college football, college basketball and the WNBA, which follow them) use multiplicative there, which tested more accurate on retail prices; everything else stays on power.
Exchanges (Polymarket, Kalshi): multiplicative. Their small overround is the bid/ask spread, not a bookmaker's margin.

3. Weight each source

Freshness. Prices arrive only when they change, so a quote's age is how long it has sat unmoved. During a game it keeps full weight for 10s, falling linearly to zero at 1 min, in basketball, hockey and soccer, where the clock runs and prices move constantly; in football, baseball, tennis and MMA, which stop between plays, pitches, points and rounds, the window is 20s to 3 min. Pregame: 15 min to 720 min. Anything the board marks stale (a silent feed, a price from before the latest score) or suspended is left out entirely. A sharp source that sits still while the retail books re-price around it is kept: they follow it. It is aged from its last change.
Depth (exchanges). No weight below $2.0k of resting order-book depth, full weight at $25k or more, linear in between. Unknown depth counts 50%. This is both sides of the top of the book; whether Kalshi ranks as a book is a separate test, the dollars at its buy price on one side.
Spread (exchanges). A midpoint is only as good as the bid-ask around it: full weight up to a 2¢ spread, falling to nothing at 10¢ ("spread 12¢").
Same line. A source quoting a different number than the board's main line is dropped during games ("line 40, board 40.5"). Before football and basketball games, a source up to 1 point away is moved to the board's number with a model of final margins and totals (key numbers like 3 and 7 included) at 50% of its weight; its chip says "moved from". When the model is unsure (a full point through 3), it is dropped instead.
Sanity. A source is dropped when a side is missing or off, or its implied probabilities sum outside 97–115%. A source sends a side only when it changes, so its other side counts as confirmed at the same moment.

The weighted average of the sources' no-vig probabilities is the no-vig price. The chips under each no-vig price print every source's share ("PM 67 · PIN 33"); dropped sources are struck through with the reason.

4. Confidence

High: sources agree within 1.5 points and at least two are fresh and solid, or a deep fresh exchange is backed by Pinnacle or Circa. A source moved from another line can back a High but never makes one.
Med: sources within 4 points with two solid sources, or one fresh sharp book or deep exchange.
Low: one weak source, or sources more than 4 points apart. EV is hidden against a Low no-vig price.
None: no usable source, so no number is shown.

5. Edges

EV% = no-vig probability × decimal odds − 1, using the most cautious no-vig price: for the side being bet we de-vig with multiplicative, Shin and power and take the lowest probability, and a source moved from another line counts at the least favorable end of the model's range. Near even the methods agree; on longshots and heavy favorites they can differ by several EV points, so an edge has to hold under all of them. That cautious price can differ from the NO-VIG column, which shows one method (power for sportsbook moneylines), so an edge's tooltip gives both: a sharp −466 / +418 shows −449, the cautious price is −426, and a book's −384 reads “+EV 2.1% vs the cautious no-vig −426”, then “3.1% vs the no-vig shown, −449”. A price is marked as an edge only at the main line, or (for spreads and totals) on a nearby number at its worst case over the model (below), only when the book's price is live (not stale or suspended), only against a Med or High no-vig price, and only at 1.0% or more. Above 10% we flag it "Verify": prices that good are often errors that books void. Arbitrage and middles use the same freshness rules.

Before a game, BEST is the best value on each side among the lines books post, sometimes on a better number. Usually that is the best price on the main line. A nearby number takes its place when a book offers a better number at a price no worse, or, in football, basketball, NHL totals, MLB and the big five soccer leagues, when our key-number model (pushes included) rates it worth more in every variation of the model. During live games BEST is the best price on the main line. The NHL puck line, MLS, Liga MX, the Champions League, soccer totals other than 2.5, preseason, spring training and summer league games, tennis games and MMA rounds have no checked model: there BEST moves only to a better number at a price no worse. A price on another number is valued by that model from the no-vig price on the main line, or on the sharps' own number when they post another, at its worst case over the model's variations. In hockey, baseball and soccer the worst case also allows the no-vig price to be off by 2 points. The Edges page lists those under "Other numbers". Hold, arbitrage, middles and the no-vig price stay on the main line.

New to no-vig prices? Start with what no-vig odds are, try the no-vig calculator, see what the no-vig price says line shopping is worth with worst line vs best line, or browse all guides.

Edges close fast — confirm at the book. Not betting advice.