OddlyLikely Compare partners

Strategy guide

Value bets: how to find them and test your edge

Value bets are prices above your fair estimate. Finding value bets takes a defensible probability, the right comparison and a record that can expose a weak method.

What value betting means

Value betting means taking a price whose break-even probability is lower than your estimate of the outcome's fair probability. At decimal odds of 2.50, the break-even probability is 40% because 1 / 2.50 = 0.40. If your estimate is 44%, the price has positive expected value according to that estimate.

Fair probability is an estimate, not a known fact before the event. It may come from a statistical model, a no-vig market reference or a documented judgement based on relevant information. The quality of that estimate matters more than how strongly you feel about the outcome.

Price value and confidence answer different questions. A 35% chance is low confidence in a win, but odds of 3.10 produce positive expected value if the 35% estimate is sound.

Use the implied-probability calculator to convert a quoted price before comparing it with your estimate.

Estimate probability before checking the offer

Write down your probability before looking at the price you plan to bet. State which data, assumptions and market definition produced it. For a team market, that could mean ratings adjusted for venue and confirmed absences. For an event prop, it could mean a distribution built from comparable attempts rather than a recent highlight.

Compare like with like. The event, market, settlement rule and information cutoff must match. A full-time football price cannot be compared directly with a price that includes extra time, and an opening estimate should be updated when team news changes the inputs.

Betting odds can be useful probability forecasts, but forecast accuracy differs by bookmaker and market size. Research across 37 competitions also found that simple normalisation was less accurate than the Shin method in the sample studied. Treat a no-vig price as a market estimate, not ground truth. Štrumbelj, International Journal of Forecasting (2014)

Remove the bookmaker margin

Decimal odds convert to raw implied probability with 1 / odds. In a complete market, add the outcome probabilities. The amount above 100% is the overround, also called the bookmaker margin or vig. To make a basic no-vig estimate, divide each raw probability by the total.

Consider a two-way market priced at 1.80 and 2.10:

Outcome Odds Raw implied probability No-vig probability No-vig odds
A1.8055.5556%53.8462%1.8571
B2.1047.6190%46.1538%2.1667

The raw probabilities total 103.1746%, so the overround is 3.1746%. Outcome A's normalised probability is 55.5556% / 103.1746% = 53.8462%. Its corresponding fair decimal price is 1 / 0.538462 = 1.8571.

The vig calculator performs this calculation for two or more outcomes. Basic normalisation assumes the margin is distributed in proportion to the raw implied probabilities, which may not describe how a bookmaker actually shaped the prices.

Calculate expected value

Expected value (EV) combines your estimated probability with the offered price. For decimal odds:

EV per unit staked = (estimated probability x decimal odds) - 1

Suppose another bookmaker offers 1.95 for Outcome A from the worked market, and you adopt the 53.8462% no-vig estimate. The calculation is (0.538462 x 1.95) - 1 = 0.05, or 5% EV per unit staked. A 20-unit stake has an expected profit of 20 x 0.05 = 1.00 unit.

Estimated probability Decimal odds EV per unit Reading
53.8462%1.95+0.0500Positive by the estimate
50%2.000.0000Break-even before other costs
50%1.80-0.1000Negative by the estimate
35%3.10+0.0850Positive by the estimate; 65% loss probability

Run your own numbers in the value-betting calculator. A positive output describes the inputs you supplied. It does not validate the probability estimate or recommend a bet.

Why a positive-value bet can lose

EV is the probability-weighted expected profit per unit under the stated inputs; it does not predict the next result. If the fair probability is 35%, the loss probability for that bet is 65%. At odds of 3.10, however, its EV is (0.35 x 3.10) - 1 = 0.085, or 8.5% per unit.

A losing result does not prove the bet lacked value, and a winning result does not prove the estimate was good. Short sequences can sit far from their expected average. Judge the estimate and decision separately from the result.

A repeatable process for finding value bets

  1. Define the exact event, outcome and settlement rule.
  2. Estimate the probability before viewing the target bookmaker's price.
  3. Convert all comparison prices to implied probabilities.
  4. Remove the margin from a complete reference market, noting the method used.
  5. Shop the same market and rules across available bookmakers.
  6. Calculate EV at the best executable price, then test how a small change in your probability affects it.
  7. Check stake limits, maximum payouts and void rules before treating the quoted price as available.
  8. Record the estimate, price, stake, closing price and result without rewriting the original reasoning.

Line shopping improves the price without changing your forecast. If the same selection is available at 1.95 instead of 1.85 under matching rules, the higher price increases EV. It does not remove the chance of losing.

Review probability quality, not just profit

Keep the probability you issued, the available price, the price taken, the market close and the outcome. This lets you review three different questions: whether your probabilities were calibrated, whether you found better prices, and whether the realised return matched the expected return over the sample.

The Brier score is one established way to assess probability forecasts by comparing each forecast with the observed outcome. It is more informative than grading every pick simply as right or wrong. Brier, Monthly Weather Review (1950)

Closing line value compares your taken price with the closing market. It can show that you repeatedly secured a better number, but it cannot prove that either the closing price or your probability estimate was correct.

Limits of value betting

  • Your fair probability can be wrong because the data, model or assumptions are wrong.
  • A basic no-vig calculation removes the displayed overround mathematically; it does not reveal how the bookmaker allocated margin across outcomes. The 2014 forecasting study compared alternative removal methods.
  • A quoted price may move or disappear before acceptance. Different settlement rules can make two similar-looking markets non-equivalent.
  • UK-licensed operators must publish rules covering voids, errors, withdrawals, maximum payouts and charges. Check the rules that apply in your jurisdiction before betting. UK Gambling Commission betting-rules code
  • UK Gambling Commission guidance treats a significant reduction in available markets or odds as material when a player has already started qualifying for a free-bet promotion. UK Gambling Commission guidance
  • Variance can produce long losing periods even when the inputs are accurate. No sample size turns an uncertain estimate into proof.

Set spending and time limits independently of any perceived edge, and do not chase losses. GambleAware's safer-gambling advice recommends setting limits in advance and stopping rather than trying to win losses back.

Calculators and related guides

Compare available prices

Once you have an independent probability and a clear market definition, compare the prices and rules offered by Oddly Likely's reviewed betting partners. Oddly Likely may earn a commission from partner links. Check the operator's current price, eligibility and terms before betting. Read the full affiliate disclosure.

Compare partners

Sources