Predictium

No Vig

A sportsbook's two prices on a game add up to more than 100%. The extra is the book's margin, called the vig. We take it out before we book a result, and on most of our boards before we price expected value. Tennis is the exception today, covered below. We trade on prediction-market exchanges, where the price you pay already has no vig.

Taking the vig out

Take a spread priced −110 on both sides. A −110 price means risking 110 to win 100, which works out to a 52.4% chance. Both sides at 52.4% add to 104.8%, and that extra 4.8 points is the margin.

Removing it means scaling the two probabilities back so they add to exactly 100%. Here that gives 50% and 50%. On a lopsided game the same scaling keeps the favourite as the favourite and just shrinks both sides. Some of our models use standard refinements of this (the Shin and power methods) that take a little more off longshots, where books tend to load more of their margin. Futures with a whole field of runners work the same way, with the margin spread across every runner instead of two sides.

How results are booked

Every result on a record page is booked to win 1 unit at the vig-free price of the side taken. If that price is a probability p, the bet risks p ÷ (1 − p) units. A win is +1 unit, a loss is minus what was risked, and a push is 0.

Say an underdog's vig-free price is 28.6%, which is +250 with no margin in it. It risks 0.40 units to win 1. A favourite at a vig-free 71.4% (−250) risks 2.50 units to win 1. At even money those would look like the same bet, which hides where the money is made. Most of our prices are +100 or longer, so the hit rate alone says little.

The breakeven is the price itself. A side taken at a 40% price breaks even if it wins 40% of the time. It isn't 50%, and it isn't the 52.4% a −110 bettor needs.

How expected value is priced

Forward expected value compares our probability with the vig-free probability of the side being bought, which is what an exchange with no vig charges. On the NFL and WNBA boards it's quoted as a return per dollar staked: our probability ÷ the price − 1. If we give a side 60% and its vig-free price is 50%, that's +20% per dollar staked. It's positive only when we give the side a better chance than the price does.

It's never priced at a flat even money. At even money a 30% shot is worth −40% however good its real price is, so a flat rule would call every longshot a disaster.

On Kalshi the price is already vig-free, so nothing is taken out. On the NBA Kalshi board, the golf board and the WNBA Kalshi lines, the expected value is worked out at the ask with Kalshi's fee included (see exchange prices).

Tennis works differently today. Its market board shows each sportsbook's expected value at that book's own quoted price, with the margin still in it, and its Kalshi row leaves the fee out. Read those numbers with that in mind.

When there is no expected value

Taking the vig out needs both sides of the market. A market where a book quotes only one side has no vig-free price, so we publish no expected value for it. The page shows a caption saying why, never a zero.

Sportsbook prices still appear on some pages as context. On the NFL and WNBA boards the expected value beside them never assumes you pay those prices. On tennis it does, as described above.

Expected value is computed once, by the backend for each venue, and published as a number. The site shows that number and never works it out again from a quoted price.

Related

Early line value is the measure every record leads with, and the EV methodology page goes into more detail on how each sport prices it.