Evidence for the Favourite–Longshot Bias in Football Betting Odds
Posted 25th April 2013
Writing in the Scottish Journal of Political Economy (The Favourite–Longshot Bias and Market Inefficiency in UK Football Betting, 2000, vol. 47/1), its authors Michael Cain, David Law and David Peel in the Scottish Journal of Political Economy discussed evidence for the favourite–longshot bias in a sample of English and Scottish football league matches played during the 1991/92 season. Betting odds were from the bookmaker William Hill. Their findings summarised below.
|Betting Odds||Number of bets||Average Return|
|Shorter than 1.66||598||98%|
|1.66 < odds ≤ 2.5||2,116||90%|
|2.51 < odds ≤ 5||5,432||89%|
|Greater than 5||509||85%|
In my book Fixed Odds Sports Betting: Statistical Forecasting and Risk Management I reviewed the evidence of the bias across European league football more generally, in which a sample of nearly 12,000 matches was analysed across the 2000/01 and 2001/02 seasons. Backing every home and away team with the bookmaker William Hill would have returned, on average, £0.87 for every £1 staked. By contrast, backing all home and away prices greater than 3.00 would have returned only £0.82 for every unit stake, whereas betting on all teams with an odds-on price would have returned as much as £0.93. Backing teams shorter than 1.50 would have lost only 4 pence for every £1. Repeating the analysis with a much larger data set taken from the 2005/06 to 2010/11 seasons using average prices recorded by the online odds comparison service betbrain.com has provided further confirmation for a demonstrable market inefficiency. This time, draw prices have also been included with home and away wins, creating a total betting odds sample size of 150,018. The results are tabulated below, first for average market prices, which reveal a strong bias, and second for best available market prices, which ingtriguingly are still not fully efficient which one might have reasonably expected (i.e 100% returns across the full ranges of odds).
|Average betting price||Number of bets||Return|
|Odds ≤ 1.5||6,235||97.74%|
|1.5 < odds ≤ 2||19,243||93.90%|
|2 < odds ≤ 2.5||22,164||94.21%|
|2.5 < odds ≤ 3.25||44,470||90.03%|
|3.25 < odds ≤ 5||45,153||87.50%|
|Odds > 5||12,753||79.55%|
|Maximum betting price||Number of bets||Return|
|Odds ≤ 1.5||4,962||101.25%|
|1.5 < odds ≤ 2||16,341||99.13%|
|2 < odds ≤ 2.5||20,552||100.25%|
|2.5 < odds ≤ 3.25||29,886||97.73%|
|3.25 < odds ≤ 5||60,796||95.84%|
|Odds > 5||17,481||95.70%|
None of this, of course, helps us identify a meaningful profitable edge over the bookmaker with which to generate a profit. But it does at least show where uninformed punters are likely to make the biggest losses.
Next time: Evidence of the favourite-longshot bias in a tennis betting market.