De-vig explained: what a price really says about probability
By Adam · Updated 11 July 2026 · 5 min read
A bookmaker's price is two things welded together: an opinion about probability, and a fee for the service. Until you separate them, you can't compare your number with theirs, measure closing line value, or say anything useful about what the market believes. Separating them properly is the part almost everyone gets wrong.
The margin, made visible
Price a coin toss fairly and both sides are 2.00, each implying 50%. A real bookmaker offers 1.91 and 1.91. Convert those to probabilities and each side implies 52.4%, for a total of 104.7%. Probabilities that sum past 100% describe an impossible world, and the excess 4.7 points are the point: that's the overround, the vig, the margin. Balance the money on both sides and the book banks it whoever wins.
Every market carries this surcharge. To recover what the bookmaker actually thinks, you have to decide how those extra percentage points were distributed across the outcomes, and this is where the trouble starts.
The obvious method, and its flaw
The tempting fix is proportional: divide each implied probability by the 104.7% total and everything sums neatly to 100% again. Tidy, quick, and built on a false assumption — that the margin was spread evenly across the outcomes.
It wasn't. A century of racecourse and bookmaker data shows the same pattern, known as the favourite–longshot bias: punters systematically overpay for long odds, and bookmakers load the margin where the overpaying happens. The favourite's price sits fairly close to fair value; the longshot's price is clipped hardest. Scale everything proportionally and the longshot keeps too much probability — the number comes out flattering, and every conclusion built on it inherits the flattery.
What the flattery costs, in numbers
Take a card market: a player at 4.50 to be booked, 1.22 that he isn't. Implied probabilities: 22.2% and 82.0%, summing to 104.2%. The proportional method hands back 21.3% for the booking. A bias-aware method, recognising that most of those 4.2 margin points sit on the longshot side, puts the true figure nearer 19.5–20%.
A point and a half sounds academic until you price with it. Suppose another firm offers 5.00 on the same booking. Judged against 21.3%, that's a +6.7% edge and an obvious bet. Judged against 19.8%, it's roughly break-even — the edge was manufactured by sloppy arithmetic, not found. Do this across a slate of player bets, which are nearly all longshots, and the lazy method will hand you a portfolio of phantom value and a genuinely mysterious losing record.
Shin's method, in a paragraph
The approach SharpXI uses comes from the economist Hyun Song Shin, who asked how a bookmaker should set prices knowing some fraction of the money arriving is better informed than they are. The mathematics that answers his question produces margin concentrated on longshots, matching what real prices have always done. Running it in reverse takes a set of quoted prices and returns probabilities with the bias already accounted for. You don't need to believe in a particular volume of insider money for it to be useful; it simply reads prices the way prices are actually written.
When de-vigging matters, and when it doesn't
If you compute edge directly — your model's probability times the quoted odds — no de-vigging appears in the formula, and the margin quietly works against you inside the price itself. Where de-vigging becomes unavoidable is every time the market serves as a benchmark: sanity-checking a model against what the market believes, measuring closing line value, comparing prices across firms. Those comparisons are only as honest as the de-vig behind them. Get it wrong in the usual direction and the market looks beatable everywhere you check — which is, historically, an expensive thing to believe about longshots.
Frequently asked
What is the vig, or overround?
The bookmaker's margin. Add up the implied probabilities of every outcome in a market and the total lands above 100% — a coin toss priced at 1.91 both sides implies 104.7%. The excess is the fee you pay to play, whichever side you take.
Why not just divide each probability by the total?
Because the margin isn't spread evenly. Bookmakers concentrate it on longshots, where punters overpay most reliably. Proportional scaling leaves longshot probabilities looking higher than they really are, which matters enormously for player bets.
What is Shin's method?
A de-vigging approach, from work by the economist Hyun Song Shin, that models how bookmakers defend themselves against informed money. Its practical effect is to assign more of the margin to longshots, which matches how real prices behave.
Do I need to de-vig if I have my own model?
Yes, whenever you use the market as a benchmark — to sanity-check the model, to measure closing line value, or to ask what the market believes. Compare against a lazily de-vigged number and longshots will look like value far too often.
Keep going
SharpXI models probabilities; it doesn't promise profit, and neither should anyone else. Betting involves risk — never stake more than you can afford to lose. 18+ — please gamble responsibly.