Favourite–Longshot Bias Explained for Football Betting
An evidence-led explanation of favourite–longshot bias, including implied probability, bookmaker margin, expected returns and why long odds do not automatically represent value.
The favourite–longshot bias is the tendency for longer-priced bets to deliver worse average returns than shorter-priced favourites. In affected markets, outsiders win less often than their quoted odds would suggest, or carry a larger effective margin than favourites.
This does not mean every favourite is underpriced or every longshot is a bad bet. The bias is an average pattern observed in some historical datasets, not a rule governing every football match, bookmaker or market. A selection at 10.00 can still offer value if its true probability is greater than the 10% break-even probability. Equally, a favourite at 1.40 can be overpriced despite being highly likely to win.
The correct question is never simply “favourite or outsider?” It is whether the available price is greater than the selection’s independently estimated fair price.
What Is the Favourite–Longshot Bias?
In betting terminology, a favourite is an outcome with a relatively high estimated probability and therefore shorter odds. A longshot is an outcome assigned a low probability and a correspondingly larger potential return.
The favourite–longshot bias describes a systematic difference between those quoted probabilities and observed results. In its conventional form:
- favourites perform better relative to their odds than longshots;
- longshots perform worse relative to their odds than favourites; and
- the bettor’s expected loss increases as the quoted price becomes longer.
The comparison is relative. Favourites can still produce a negative average return after bookmaker margin. They merely lose less, on average, than longer-priced selections in a market exhibiting the bias.
Odds can be converted into break-even probabilities using the method explained in GoalIQAI’s guide to football betting odds and implied probability:
Break-even probability = 1 ÷ decimal odds
A price of 2.00 requires a 50% success rate to break even before considering differences between quoted and obtained prices. A price of 10.00 requires a 10% success rate.
A Simple Football Example
Suppose a team is offered at decimal odds of 8.00 to win an away match.
The quoted break-even probability is:
1 ÷ 8.00 = 0.125, or 12.5%
If a well-calibrated independent estimate gives the team only an 11% chance of winning, the fair odds would be:
1 ÷ 0.11 = 9.09
The 8.00 market price is shorter than the 9.09 fair price. The outsider may produce a large payout when it wins, but the price does not fully compensate for how rarely that outcome is expected to occur.
The expected return for a £1 stake is:
Expected return = (estimated probability × decimal odds) − 1
Using the 11% estimate:
(0.11 × 8.00) − 1 = −0.12
That represents an illustrative expected loss of 12 pence per £1 staked, or an expected return on stake of −12%.
The long price changes the size of the potential payout. It does not, by itself, establish positive expected value.
Break-Even and Expected Returns Across Different Prices
The following comparison uses three separate hypothetical selections. The estimated probabilities are illustrative and are not claims about a specific league, bookmaker or historical dataset.
| Price type | Decimal odds | Break-even probability | Illustrative estimated probability | Estimated fair odds | Expected return |
|---|---|---|---|---|---|
| Short-priced favourite | 1.50 | 66.67% | 68% | 1.47 | +2% |
| Medium price | 3.00 | 33.33% | 32% | 3.13 | −4% |
| Longshot | 8.00 | 12.50% | 11% | 9.09 | −12% |
This is an illustration of the pattern, not evidence that the same return gradient exists in every real market. It shows how relatively small probability differences can produce materially different expected returns.
At 1.50, an estimated probability of 68% is slightly above the 66.67% break-even point. At 8.00, an estimated probability of 11% is only 1.5 percentage points below the quoted 12.5%, but that difference produces an expected return of −12%.
Probability-point errors are therefore not economically equal across the odds range. At longer prices, a small absolute overstatement can represent a large proportional error in the outcome’s true probability.
What Does “Expected Return” Mean?
Expected return is the theoretical average result if comparable decisions could be repeated many times under the same assumptions. It is not a forecast of what will happen to one bet.
Using the illustrative table:
| Decimal odds | Expected return per £1 | Illustrative return from £1,000 staked | Illustrative profit or loss |
|---|---|---|---|
| 1.50 | £1.02 | £1,020 | +£20 |
| 3.00 | £0.96 | £960 | −£40 |
| 8.00 | £0.88 | £880 | −£120 |
Actual results would vary substantially, particularly for the 8.00 selection. Longshots win infrequently, so realised results can remain far from their theoretical expectation over surprisingly large samples.
How Bookmaker Margin Can Contribute to the Pattern
A bookmaker creates an overround by offering odds whose implied probabilities total more than 100%. The excess represents the market’s quoted margin, although realised profit also depends on customer behaviour, liabilities and price movement.
GoalIQAI’s guide to bookmaker margin and overround explains the basic calculation.
The margin does not have to be distributed proportionally across every outcome. Consider a simplified two-outcome market with fair probabilities of 80% and 20%:
| Outcome | Fair probability | Fair odds | Offered odds | Quoted implied probability | Expected return at fair probability |
|---|---|---|---|---|---|
| Favourite | 80% | 1.25 | 1.20 | 83.33% | −4% |
| Longshot | 20% | 5.00 | 4.50 | 22.22% | −10% |
The offered probabilities total 105.55%, but the expected loss is not identical for both selections. At the assumed fair probabilities, the favourite loses 4% in expectation while the outsider loses 10%.
This is only an illustrative margin allocation. Real markets contain multiple operators, changing prices and heterogeneous customers. It nevertheless shows why simply removing the overround proportionally may fail to capture how pricing error is distributed across outcomes.
Why Might Bettors Overbet Longshots?
Researchers have proposed several possible explanations. These explanations are not mutually exclusive, and their importance can vary between markets.
Preference for a large payout
A small stake with a large possible return may provide entertainment or excitement beyond its financial expectation. Some bettors may knowingly accept a worse expected return in exchange for the possibility of an unusually large win.
This resembles demand for lottery-style outcomes. It does not necessarily mean the bettor misunderstands the probability; the bettor may value the payoff profile itself.
Overweighting small probabilities
People may treat a low-probability outcome as more plausible than the evidence supports. A team priced at 15.00 can feel “too big” because victory is imaginable, even when the difference between a 6% and 7% chance is difficult to judge intuitively.
Possibility is not the same as probability. Almost any football result is possible, but a price offers value only when it adequately compensates for its frequency.
Narrative appeal
Longshots often come with compelling stories: a new manager, derby motivation, recent giant-killing result or supposedly underestimated home advantage.
Those factors may be relevant, but a plausible story does not show that the market has underpriced the outcome. The same narrative may already be reflected in the odds or may receive more attention precisely because it supports an exciting outsider.
Uneven bookmaker margin
Operators may face different customer demand, competitive pressure and liability across the price range. Popular favourites can be highly visible and easy to compare, potentially creating stronger price competition than obscure outsiders or secondary markets.
The resulting margin distribution may make longer prices less attractive even when the headline market overround appears reasonable.
Information and liquidity
Shorter-priced outcomes may attract more informed attention or greater trading volume in some markets. Longer-priced selections can be harder to estimate and may have wider disagreement between models.
That does not establish a universal causal explanation. Liquidity, limits and participant quality differ across competitions, operators and market types.
What Does the Football Research Show?
A peer-reviewed study by Cain, Law and Peel, published in 2000, reported evidence of favourite–longshot bias in UK fixed-odds football betting, including match results and correct scores. The bibliographic record is available through the Lancaster University research directory.
Later research has also identified the pattern in various fixed-odds sports markets. For example, Berkowitz, Depken and Gandar examined favourite–longshot bias in fixed-odds markets and reported that heavy favourites produced the strongest average returns in their samples, although those returns were close to zero rather than evidence of an effortless profit strategy.
The important limitation is that results depend on:
- the sport and competition;
- the period studied;
- whether opening, average, best or closing odds are used;
- the bookmaker and customer population;
- the market type;
- how margin is removed; and
- whether selection and transaction constraints are included.
An historical average cannot prove that the same effect remains exploitable now. Once a pattern is documented, operators, bettors and trading models can adapt. Market structure also changes over time.
Why the Bias Is Not a Universal Shortcut
Backing every favourite can still lose money
The conventional bias says favourites may offer better returns relative to longshots. It does not say favourites necessarily have positive expected value.
If favourites return −3% and longshots return −12%, the favourites have performed better, but both strategies have lost money.
Some outsiders are genuinely underpriced
A longshot can offer value when the available odds exceed a defensible estimate of fair odds. Rejecting every outsider would be as simplistic as backing every one.
The correct framework is explained in the GoalIQAI guide to value betting: compare estimated probability with the probability required by the price.
The pattern can vary between markets
A mature Premier League match-odds market is not equivalent to a lower-league player market, correct-score market or early outright market. Liquidity, margin, limits and information quality may differ materially.
A favourite–longshot pattern established in one market should not be transferred automatically to another.
Best available price matters
Research using one bookmaker’s price may not describe the return available to someone comparing multiple operators. Improving an outsider from 8.00 to 9.00 changes its break-even probability from 12.5% to 11.11%.
Price shopping can reduce an apparent longshot penalty, although it cannot transform an outcome into value when its underlying probability remains too low.
Historical results are vulnerable to noise
Longshots create high-variance samples because wins occur infrequently. A few additional or missing winners can materially change an estimated return.
Any analysis should report the number of bets, time period, uncertainty and out-of-sample performance. The GoalIQAI guide to sample size in football analytics explains why a large row count does not automatically guarantee a reliable conclusion.
Calibration and the Favourite–Longshot Bias
Calibration asks whether outcomes assigned a particular probability occur at approximately that frequency over time.
If football teams priced as 20% chances win roughly 20% of comparable matches after appropriately removing margin, the probabilities are well calibrated around that range. If they win only 17%, the market has overstated their chances on average.
A calibration analysis can group selections into probability or odds bands and compare:
- average quoted or margin-adjusted probability;
- observed win frequency;
- average expected or realised return;
- sample size; and
- uncertainty around the observed frequency.
Wide odds bands can hide important differences. A single “longshot” category might combine 5.00 selections with outcomes priced at 30.00, despite their very different probabilities and variance.
How to Test for Favourite–Longshot Bias
A credible test requires more than calculating whether favourites won more matches. Favourites are expected to win more often because their prices already imply higher probabilities.
A better process is:
- Collect the odds available at a consistent point, such as market close.
- Record the bookmaker, competition, market and result.
- Convert odds into implied probabilities.
- Remove or model the bookmaker margin using a stated method.
- Divide selections into sufficiently granular price bands.
- Compare predicted probabilities with observed frequencies.
- Calculate realised return on stake within each band.
- Report sample size and statistical uncertainty.
- Test the pattern on a later out-of-sample period.
Changing the margin-removal method can change the conclusion. Proportional normalisation assumes the overround is allocated proportionally, while other methods allow for a non-linear relationship between odds and margin. The method should be disclosed rather than treated as an invisible technical choice.
How GoalIQAI Uses the Concept in Predictions
The favourite–longshot bias should act as a safeguard against weak reasoning, not as an automatic selection rule.
When a high-priced underdog appears attractive, the analysis should ask:
- What probability does the available price require?
- Has the bookmaker margin been considered?
- Is the independent probability materially above the break-even point?
- How sensitive is that estimate to team news and modelling assumptions?
- Is the market liquid and competitive?
- Is the quoted price still available?
- Could narrative appeal be causing the outsider’s chance to be overstated?
“Anything can happen in football” is not a value argument. Neither is “the price looks too big”. A longshot requires the same disciplined comparison between probability and price as every other selection.
Its higher variance also means that process cannot be judged from a few memorable wins or a long losing run. GoalIQAI’s guide to separating process from results provides the appropriate evaluation framework.
Key Takeaways
- The favourite–longshot bias describes worse average returns on longer-priced selections relative to favourites.
- It is a relative pattern: favourites can still have negative expected returns.
- A large potential payout does not make an outsider good value.
- Small absolute probability errors can create large proportional errors at long odds.
- Possible explanations include demand for large payouts, probability weighting, narrative appeal and uneven margin allocation.
- The strength and direction of the effect can vary by bookmaker, competition, period and market type.
- Backing every favourite is not a valid response to the bias.
- Every selection should still be assessed by comparing its estimated probability with its break-even probability.
- Historical patterns require adequate samples, calibration testing and out-of-sample validation.
Related Guides
Stay Ahead of the Market
Subscribe to GoalIQAI for evidence-based football predictions, probability analysis and educational guides focused on price, uncertainty and better decision-making.