In one sentence
The favourite-longshot bias is the pattern where big-priced selections win less often than their odds suggest, while short-priced favourites win slightly more often.
How it works
Punters like big payouts. An outsider at 51.0 is a bigger thrill than one at 1.50, so money flows to longshots beyond what their chances justify. Bookmakers also load more margin onto outsiders, partly because they carry more risk of being hit by well-informed money.
The result is that a £1 bet on a longshot loses far more on average than a £1 bet on a favourite. The pattern has been documented in horse racing for decades and appears in football and many other sports, though its size varies by market.
On betting exchanges the bias tends to be smaller, because margins are thin and anyone can lay an overpriced longshot. It has not disappeared, though, and it is a key reason proportional margin removal can mislead.
The maths
There is no single formula for the bias itself. It is an empirical pattern, measured by comparing implied probability with actual win rate across price bands.
- j: a price band, for example all selections priced 20.0 to 30.0.
- W j: the number of winners in that band.
- N j: the number of selections in that band.
- q bar j: the average implied probability in the band.
In plain English: in each price band, does the actual win rate beat or fall short of what the odds implied?
Worked betting example
Football, illustrative figures, not real market data. Suppose your records suggest:
- A Correct Score longshot such as 3-3 at 51.0 implies 1.96% but truly lands 1.2% of the time.
- A Match Odds favourite at 1.50 implies 66.67% but truly wins 68% of the time.
- Longshot EV per £1: 0.012 × 51.0 − 1 = −0.388, so −38.8%.
- Favourite EV per £1: 0.68 × 1.50 − 1 = +0.02, so +2.0%.
Laying the longshot on Betfair at 51.0 for a £2 backer's stake:
- Your liability is £2 × 50 = £100.
- You win £2 with 98.8% probability and lose £100 with 1.2%.
- EV = 0.988 × £2 − 0.012 × £100 = £1.976 − £1.20 = £0.776, before commission.
- After 2% commission the £2 win is £1.96, so EV = 0.988 × £1.96 − £1.20 = £0.736.
That looks attractive, but you are risking £100 to make £2, and a couple of landed longshots wipe out many weeks of profit.
Where it's good
- Choosing a margin-removal method: power, Shin and odds-ratio all reflect this bias.
- Avoiding the biggest leaks: most recreational losses come from longshots and multiples.
- Building systematic lay strategies on exchange outsiders, carefully.
- Calibrating a model: if your model agrees with longshot prices, it may inherit the bias.
- Pricing Correct Score, where most of the big-priced scorelines sit.
Limitations and pitfalls
- The bias is strongest in bookmaker racing markets. In major football markets on the exchange it can be too small to profit from after commission.
- Knowing the pattern does not make favourites profitable. A slightly less bad bet is still usually a losing bet after margin.
- Measuring it needs large samples, because longshots rarely win. A few thousand selections at 50.0 give very noisy estimates.
- It varies by league, market and price range, so a figure from one market does not carry over.
- Laying longshots gives small steady wins and occasional large losses; bankroll and liability limits matter.
- The bias can shrink as more traders exploit it.
How to build it
- pandas to bin selections by implied probability; statsmodels for a logistic regression of wins on log odds.
- Data: many thousands of selections with prices and results, ideally exchange prices at a fixed time before kick-off.
- Practical tip: plot actual win rate against implied probability on a log scale. A curve sagging below the diagonal at the long end is the bias.
Related methods
- Margin removal: methods that account for the bias.
- Overround: the margin that longshots carry more of.
- Calibration: the tool for measuring the bias.
- Behavioural biases: the punter psychology behind it.
- Market efficiency: the bias is a known crack in efficiency.