How to remove vig from betting odds

No-vig odds calculator

Enter every outcome from the same sportsbook market. The calculator shows the quoted overround, then removes it four different ways.

The basic no-vig formula

Sportsbook odds contain an implied probability and a margin. Convert every price in the market, add the probabilities, then bring the total back to 100% using a declared method.

For odds of 1.55 and 2.55, the raw probabilities are 64.516% and 39.216%. They total 103.732%, so the quoted overround is 3.732 percentage points.

Four ways to remove the vig

Every method forces the adjusted probabilities back to 100%. They disagree on where the bookmaker placed the margin.

Which method should you use?

Start with multiplicative when you need a clean baseline. Use another method because it improved out-of-sample calibration for the market you model—not because it creates a larger edge.

Two-way additive and Shin results are equivalent under the standard Shin calculation. Differences become more useful to inspect in uneven and multi-outcome markets.

The methods produce different fair odds

Two-way market: 1.55 / 2.55

Three-way market: 2.10 / 3.20 / 3.30

The implied probabilities total 109.172%. A three-way book makes the allocation assumption easier to see.

The gap is not an error. It measures model risk: each method makes a different claim about how the 9.172-point overround was distributed.

Match the exact market first

The maths is useless when the inputs describe different bets. Require every mutually exclusive outcome from one complete market and match the normalized market, period and line fields.

• 1
  Same bookmaker and timestamp

  Do not combine one current price with an opposing side captured five minutes earlier.

• 2
  Same event and period

  Full-game odds cannot complete a first-half market.

• 3
  Same market and line

  Over 224.5 and under 225.5 are not opposite outcomes. Home −3.5 and away +4 are not a complete spread.

• 4
  Same player and metric

  A player-points over needs the under for the same player, line, period and settlement rule. The NBA player-prop workflow in Python shows the full selection key.

• 5
  Every possible winner

  A football 1X2 market needs home, draw and away. Leaving out the draw does not turn it into a two-way market.

You cannot remove vig from one selection. The missing outcomes still occupy probability space.

Calculate all four methods in Python

The function takes one complete market in decimal odds. It returns fair decimal odds for every valid method and refuses malformed prices.

from math import sqrt


def remove_vig(decimal_odds):
    if len(decimal_odds) < 2 or any(o <= 1 for o in decimal_odds):
        raise ValueError("Use at least two decimal prices above 1.00")

    raw = [1 / o for o in decimal_odds]
    book_total = sum(raw)
    if book_total <= 1:
        raise ValueError("The market has no positive overround to remove")

    probabilities = {
        "multiplicative": [p / book_total for p in raw],
    }

    margin_each = (book_total - 1) / len(raw)
    additive = [p - margin_each for p in raw]
    if all(p > 0 for p in additive):
        probabilities["additive"] = additive

    def power_total(exponent):
        return sum(p ** exponent for p in raw)

    low, high = 1.0, 2.0
    while power_total(high) > 1:
        high *= 2
    for _ in range(80):
        midpoint = (low + high) / 2
        if power_total(midpoint) > 1:
            low = midpoint
        else:
            high = midpoint
    probabilities["power"] = [p ** high for p in raw]

    def shin_probabilities(z):
        return [
            (sqrt(z * z + 4 * (1 - z) * p * p / book_total) - z)
            / (2 * (1 - z))
            for p in raw
        ]

    low, high = 0.0, 0.999999999
    for _ in range(80):
        midpoint = (low + high) / 2
        if sum(shin_probabilities(midpoint)) > 1:
            low = midpoint
        else:
            high = midpoint
    probabilities["shin"] = shin_probabilities((low + high) / 2)

    return {
        method: [round(1 / p, 3) for p in values]
        for method, values in probabilities.items()
    }


print(remove_vig([1.55, 2.55]))

Request no-vig fields from the API

For normalized sportsbook data, request the available no-vig fields with the quoted odds. Keep null values as unavailable evidence.

curl --request GET \
  --header "X-API-Key: $ODDS_API_KEY" \
  "https://api.odds-api.net/v1/events/{event_id}/odds/snapshot?price_fields=odds,novig&include_unavailable=true"

Supported lines can include odds_no_vig, odds_no_vig_multiplicative_method, odds_no_vig_additive_method, odds_no_vig_power_method and odds_no_vig_shin_method. A null means the response did not contain enough comparable evidence for that calculation.

Build a weighted market consensus

Do not average raw sportsbook prices. One bookmaker may quote a 3% overround and another 9%. Averaging them first carries both margins into the model.

Possible weights include source sharpness, price freshness, liquidity, historical calibration and coverage. Require a minimum source count and cap or reject outliers. Fit those choices on earlier data, then test them on later events.

Keep the de-vig method fixed during a backtest. Switching methods because one creates a better-looking result is another form of overfitting.

Continue with the full sports betting model guide.

Compare bookmaker vig without inventing a leaderboard

A defensible bookmaker comparison needs complete markets captured at the same fixed horizons. A single attractive selection price does not prove that the bookmaker runs a lower-margin book.

First, build an exact bookmaker comparison so each row represents the same bet across sources.

• Choose operators by a dated, cited country-level measure. “Europe” is not one bookmaker market.
• Pre-register the sports, leagues, market types and collection horizons.
• Calculate overround within each bookmaker before comparing operators.
• Separate two-way and three-way markets.
• Report event counts, complete-market counts, medians, distributions and event-level confidence intervals.

No-vig odds questions