Implied Probability and Devigging NHL Betting Odds

Removing Margins from UK Hockey Bookmakers

I once put a tidy four-figure stake on a Vegas moneyline at 1.85 because the model said the team was a 55 % favourite and the price implied 54.1 %. That looked like edge. It was not. What I had failed to do was strip the bookmaker’s vig out of the other side of the market, and once I worked through the maths properly the true market probability was closer to 56 %. The book was actually pricing Vegas as the marginally less likely winner of a coin flip — and I was on the wrong side of the implied number I had not bothered to calculate. The bet lost. The lesson stuck.

Devigging is the single most useful skill an NHL bettor can develop, and almost nobody bothers with it. This article is the version of the explanation I wish I had had ten years ago.

What implied probability actually means

Implied probability is the percentage chance a price assumes will happen. For decimal odds, the formula is 1 divided by the decimal price. A 2.00 price implies 50 %. A 1.50 price implies 66.7 %. A 3.00 price implies 33.3 %. That is the part everyone knows.

The part most bettors skip is that those numbers do not add up to 100 % across a two-way market. They add up to 103, 105, sometimes 110 — the surplus is the overround, also called the vig, the juice, or the margin. UK books typically run overrounds of around 4 to 6 % on standard NHL moneylines and significantly more on props, totals and futures. Until you remove the overround, the implied probability of either side is inflated above the book’s true estimate.

That matters because your edge is calculated against the book’s true estimate, not against the displayed implied probability. If a moneyline says 2.10 — implied 47.6 % — and you think the team is a 50 % proposition, you might assume you have 2.4 percentage points of edge. After devigging, the book’s true estimate might be 49.5 %, in which case your real edge is half a percentage point. That difference is the gap between a profitable bet and a coinflip with a fee.

The simple proportional devig

The most common devigging method is proportional: take each side’s raw implied probability and divide it by the sum of both sides’ implied probabilities. The result is a vig-free fair probability for each outcome.

Worked example. Edmonton is priced at decimal 1.75. Vancouver at decimal 2.20. Raw implieds: 57.1 % and 45.5 %, summing to 102.6 % — a 2.6 % overround. Edmonton’s devigged probability is 57.1 / 102.6 = 55.7 %. Vancouver’s is 45.5 / 102.6 = 44.3 %. These now sum to 100 % and represent the book’s true probability estimate stripped of margin.

If your model has Edmonton at 58 %, your true edge over the book is 58 – 55.7 = 2.3 percentage points. That is enough to bet. If your model has Edmonton at 56 %, your edge is 0.3 percentage points and the bet is a wash after accounting for estimation error. Without devigging, you would have read your edge against the 57.1 % raw implied and concluded you had no edge at all — even though you have a real one.

Why proportional devigging is not perfect

Proportional devigging assumes the bookmaker applies the same percentage margin to both sides. That assumption holds reasonably well on close-to-even matchups. It breaks down on lopsided ones.

Imagine a 1.10 / 8.00 market. Raw implieds: 90.9 % and 12.5 %, sum 103.4 %. Proportional devigging gives 87.9 % and 12.1 %. But experienced market-makers know that vig on a heavy favourite is typically lower in percentage terms than vig on a long underdog — because the bookmaker takes more risk on the dog side, they pad the favourite side less. The true probability on the favourite might actually be closer to 89 %, with the underdog correspondingly squeezed.

The fix is called the “power” or “logit” method, which applies a different rebalancing weighted by the magnitude of the raw probabilities. For most NHL bets the proportional method is close enough — within half a percentage point of the more sophisticated approaches — and the simplicity is worth the small loss in accuracy. I use proportional devigging on close lines and switch to a power method on extreme prices longer than 4.50 or shorter than 1.30.

Three-way markets and the regulation problem

NHL three-way regulation markets — home win, away win, tie after sixty minutes — add a complication. You now have three implied probabilities to sum, and the overround is typically higher because the book is offering three settlement paths instead of two.

Worked example. Price quoted as 2.10 / 4.20 / 3.30 for home / tie / away. Raw implieds: 47.6 %, 23.8 %, 30.3 %. Sum: 101.7 %. Proportional devig: 46.8 %, 23.4 %, 29.8 %. Compare your model’s three-way probabilities against those numbers, not against the raw implieds.

The three-way market is one of the few places where the devigging exercise tends to produce the biggest “aha” moments for bettors. The tie probability — typically around 22 to 25 % in the NHL when sixty-minute regulation is settled before overtime — is consistently misjudged by recreational bettors who assume ties are rarer than they are. Devigging the price reveals exactly what the book thinks, and the gap between the book’s estimate and the casual bettor’s mental model is often substantial.

Using devigged prices to find +EV bets

The professional-grade workflow looks like this. Build a model that produces a probability estimate for each side of a market. Pull the raw prices from the book you intend to bet at. Devig the prices to get the book’s vig-free probability estimate. Compare your model’s probability to the devigged book probability. If your probability is meaningfully higher than the devigged book probability — by at least one and ideally two or more percentage points to account for estimation error — the bet has positive expected value.

The same workflow can be done across multiple books to find the sharpest line. The sharpest line is the one closest to the true probability after devigging, which is almost always Pinnacle, Circa or another low-margin operator. UK punters can use the sharp book’s devigged price as a proxy for true probability and then look for a soft book to actually place the bet at a longer price.

Professional bettors typically nail down a 2 to 5 % edge per bet using fractional Kelly and flat percentages of bankroll. That edge is, almost by definition, the gap between your probability estimate and the devigged market estimate. Without the devig step, you are guessing at the gap rather than measuring it.

The mental shortcut that saves time

If you do not want to do the maths on every bet, internalise this shortcut: subtract roughly half the overround from the raw implied of the favourite, and add roughly half the overround to the raw implied of the dog, to get the devigged estimates. On a 4 % overround, that means subtracting 2 percentage points from the favourite’s raw implied and adding 2 percentage points to the dog’s. It is not perfect but it is fast, and it is far better than betting off the raw implied number.

For longer NHL series prices, futures and grand-salami totals, the overround is much wider — sometimes 15 % across the field — and the shortcut breaks down. There you genuinely need to run the proper devig. My futures bets always have a devigged price next to them in the tracking spreadsheet before I commit.

The habit that compounds

Devigging is not glamorous. It is a thirty-second arithmetic step that converts a displayed price into the book’s true opinion of the matchup. Done consistently, it changes the way you evaluate every bet you place — you start seeing edges where there are real ones and rejecting bets that looked profitable but were not.

The transition from “I think this team is good value at 1.90” to “the devigged fair price is 50.6 % and I have it at 53 % — that is 2.4 points of edge and worth one unit at quarter Kelly” is the transition from punting to betting. The numbers do not lie. Make them say what they actually mean before you commit your money.

For the natural next step — turning a devigged estimate into a bet size — the Kelly criterion for NHL betting piece picks up where this one leaves off.

Published by the Ice Hockey Betting Tips team.