How consistently do football teams cover their Asian Handicap?
Posted 8th September 2016
Pinnacle.com recently published an article looking at the use of historical Asian handicap data as a predictive tool to find betting value. For example, since Mark Hughes took over at Stoke, the team has covered their Asian handicap more than 50% of the time every season. Teams with an average percentage of over 50% have essentially beaten the bookmakers in the Asain handicap betting betting. Liverpool, by contrast, has on average failed to cover the spread every season during the last six seasons. What this would have meant for bettors is that betting on Liverpool's opponents to cover their spread throughout the season might well have proven to be profitable. But are these historical patterns evidence of something real and consistent or just lucky flukes? This article looks at a larger data set of teams to search for answer.
Taking the whole of last season's Asian handicap and full time scores from this website, I have calculated an average goal excess for every team in my database managed to achieve over the course of its games. For example, if a team's handicap was -1 goal and it won its game 3-1, this equates to a +1 goal excess for that game. The size of the average excess therefore provides a measure of how well the team had performed during the 2015/16 season relative to what the betting market had expected. One problem that might arise with this approach would be where the betting market was consistently publishing a handicap that was not close to a 50-50 spread, i.e. odds of 2.00. Fortunately, this does not appear to have been the case for the 404 teams in this analysis. Average handicap price was 1.985 (best market prices), with a range from 1.87 to 2.10.
Average goal excess scores varied from a maximum of 0.836 for French Championship winners Paris St. Germain (implying it scored nearly 1 goal more than its opposition on average per game than the Asian handicap betting market thought it would), to a minimum of -0.76 for another French side Troyes (which managed to get relegated with such a performance). The average was zero, whilst the standard deviation was 0.265. 68% of scores were within plus or minus one standard deviation of average, whilst 95% were within two standard deviations. Such figures imply these scores are normally distributed. The chart below provides further confirmation of this feature. Any data that are normally distributed provide strong evidence that they arise as a consequence of randomness.
Average goal excess scores for each team were then calculated for the first and second half of the season. If such scores could be considered predictive of future performance, we might reasonably expect a high/low score in the first half of the season to more readily be followed by a high/low score in the second half of the season. The plot below reveals that for the 404 teams as a whole this is not the case. There is essentially no correlation between first and second half season scores, implying virtually complete randomness and no predictability by this measure.
Really it is unsurprising that such randomness is found. If a team is outperforming its market expectation, that market will readjust by increasing the size of handicaps until performance regresses towards the mean. This is really a restatement of the efficient market hypothesis: where mistakes in prices (or handicaps) exist, they will be exploited to the point where they no long exist. Of course, the elimination of such errors will not be instantaneous. Those bettors who spot the mistakes earliest are most likely the ones who will be able to exploit them for any possible profitability.
Learn more about the efficient market hypothesis and its implications for bettors in my new book: Squares & Sharps, Suckers & Sharks - the Science, Psychology and Philosophy of Gambling.