How to Calculate and Check a Value Bet

A value bet is a bet with positive modeled expected value (EV), based on your estimated probability and the odds you actually receive. Value betting compares probability with price; it does not promise a win or profit. Even the most likely outcome can be poor value at the offered odds.

What should you compare?

Compare your own estimated probability with the odds’ implied probability, then calculate expected value separately. The probability gap and EV describe different things and use different units.

For decimal odds O and an estimated probability p expressed as a fraction:

EV is positive when p > 1 / O. Fees and other costs can change that conclusion; these formulas exclude them.

Worked example: odds of 2.10, a 52% estimate and a $100 stake

These inputs give a probability gap of +4.38 percentage points and modeled EV of +9.20%, or +$9.20 on a $100 stake. The calculation assumes that the 52% estimate is correct and excludes extra costs. It is not realized profit.

First, distinguish the gross return from the net result:

• A win returns 100 × 2.10 = 210: a gross return of $210, including the original stake.
• The net profit on a win is 210 − 100 = 110, or +$110.
• A loss produces a net result of −$100.

The calculations below use unrounded inputs and round only the final results:

A single bet in this simplified win-or-loss example does not pay the +$9.20 mean: it either nets +$110 or loses $100. Expected value is the probability-weighted average of those possible net results.

How much does a probability error matter?

A small reduction in your probability estimate can erase a small positive EV. At the same decimal odds of 2.10, the calculation changes as follows:

Each row uses p × 2.10 − 1, converted into a percentage and a monetary amount. These are hypothetical estimates, not observed win rates. The table shows sensitivity to the input; it does not establish which estimate is correct.

How do you account for bookmaker margin?

Use the complete set of mutually exclusive outcomes in a market to calculate its overround. The overround is the amount by which the sum of their implied probabilities exceeds 100%.

Consider an illustrative three-outcome market with decimal odds of 2.05, 3.60 and 3.90. These are example prices, not a live bookmaker quote.

The implied probabilities total 102.20%, giving an overround of 2.20 percentage points. Values are rounded after calculation.

The ToolsGambling calculator’s default devig method uses multiplicative, or proportional, normalization: it divides each implied probability by the sum of all implied probabilities. For one outcome, the formula is (1 / outcome odds) / sum of implied probabilities.

This produces a no-vig estimate, not a known true probability. It is one model for allocating margin. Interpreting normalized implied probabilities as probability estimates requires additional market assumptions, as discussed in University College Dublin working paper WP24/03.

Keep the two tasks separate: manual mode uses your own probability estimate, while devig mode provides a market-based reference from a complete set of prices.

How do you use the ToolsGambling calculator?

Open the Value Bet Calculator, enter decimal odds and your own probability estimate, and read the break-even probability and EV results. The tool performs the arithmetic; it does not validate your estimate.

To reproduce the example:

1. Select Manual / Own probability.
2. Choose the decimal odds format.
3. Enter 2.10 in the odds field.
4. Enter 52 in the probability field, meaning 52%.
5. Enter 100 in the stake field.
6. Leave or enter 1000 in the bankroll field.

English Value Bet Calculator inputs showing decimal odds of 2.10, a user probability estimate of 52%, a $100 stake and a $1,000 bankroll

This screenshot from the local development server shows the example inputs, not a production test. You supply the 52% estimate. The $1,000 bankroll and $100 stake are illustrative inputs, not a staking recommendation.

The field labeled Your true win probability means your own estimated probability here. Neither the label nor the calculation establishes the outcome’s true probability.

In the results, implied probability appears as 47.6% because the interface rounds it. EV is +9.20%, and the corresponding modeled amount on the $100 stake is +$9.20.

Calculator result tiles showing +9.20% expected value and a modeled amount of +$9.20

This local development screenshot shows the same EV in percentage and monetary terms. Neither result is realized profit, and +9.20% is not the probability gap.

The calculator does not discover events, fetch live prices, predict sports outcomes or validate a model.

For a separate no-vig calculation, use Devig mode with the complete set of mutually exclusive outcomes for one market. It normalizes the implied probabilities proportionally; the result does not replace an independent check of your probability estimate.

Where should your probability estimate come from?

Your estimate should come from a stated, repeatable method and be timestamped before the event. A bare figure such as 52% is not enough: positive calculator output only reflects the assumption you entered.

Use this workflow:

1. Define the market and settlement rules. Specify the outcome being assessed and when the bet wins, loses or is refunded.
2. Record the estimate before the event. Save the date, time, data sources, method and model version so you can reconstruct the decision without knowing the result.
3. Compare the estimate with the price. Use the break-even threshold 1 / odds and, optionally, a complete-market no-vig estimate as a separate reference.
4. Record the accepted odds and time. The best price you see may differ from the price actually accepted. Base your calculation on the accepted price.
5. Recalculate when the price changes. Positive EV at an earlier price does not automatically remain positive at a new one.
6. Evaluate on new data. Do not judge the method only on the events used to fit or tune it.

Calibration asks whether predicted probabilities match observed outcome frequencies. Group forecasts into probability ranges and compare predictions with outcomes on out-of-sample data. The proportion of correctly selected outcomes is not a substitute for this check. Wilkens’ paper listed by the University of Bath examines the distinction between predictive accuracy and probability calibration; its empirical results should not be treated as a universal claim about betting models.

What should you record, and how should you assess results?

Keep a decision log before each event, then assess probability calibration, price quality and realized returns separately. None of these checks replaces the others.

Closing line value (CLV) compares the odds you actually accepted with a comparable closing price. The market, outcome and settlement rules must match. CLV is a price-quality reference, not proof that your probability estimate is correct or that future bets will be profitable.

Calculate realized returns separately from net results and stakes, including costs. Wins and losses alone do not tell you whether a model is well calibrated.

There is no universal number of bets that proves an edge. Small samples are noisy, and a larger log does not correct a systematic estimation error. Neither CLV nor a winning run guarantees future profit.

How does value betting differ from arbitrage?

Value betting depends on a probability estimate for one selected outcome; arbitrage combines prices to cover every outcome in a market. With value betting, positive expectation depends on the estimate’s quality, and individual results remain subject to variance.

Arbitrage uses multiple prices, potentially from different operators. The calculation depends on every required price actually being accepted and all bets covering the outcomes under compatible settlement rules. One attractive price is not enough to establish arbitrage.

What risks remain after the calculation?

Positive EV does not guarantee profit: the conclusion depends on the probability estimate, price availability, accepted odds, fees, settlement rules and uncertainty in the inputs. Variance remains even when a model is well specified.

Use the calculator to check arithmetic and your log to review decisions. Never chase losses or bet money needed for essential expenses.