دقت پیش بینی ساختارهای مختلف بازار - مبادلات فروشندگان بلیط در برابر مبادلات شرط بندی

There is a well-established body of literature on separately testing the prediction power of different betting market settings. This paper provides an inter-market comparison of the forecasting accuracy of bookmakers and a major betting exchange. Employing a dataset covering all football matches played in the major leagues of the “Big Five” (England, France, Germany, Italy, Spain) during three seasons (5478 games in total), we find evidence that the betting exchange provides more accurate predictions of a given event than bookmakers. A simple betting strategy of selecting bets for which bookmakers offer lower probabilities (higher odds) than the betting exchange generates above average, and in some cases even positive returns.

Similarly to financial securities, betting markets trade contracts on future events. The price of a contract reflects the owner’s claim, which is tied to the event’s outcome. Therefore, the market price can be interpreted as a prediction of the future event. According to Vaughan Williams (1999), betting markets are particularly well suited to the investigation of forecasting accuracy because–in contrast to most financial markets–the contracts have a definite value that becomes observable after a clear termination point. The traditional form of gambling on sports events is bookmaker betting. In this market setting, the bookmaker acts as a dealer announcing the odds against which the bettor can place his bets. However, in recent years a different market structure has evolved: betting exchanges. Whereas the bookmaker defines the odds ex ante, the prices in the bet exchange are determined by a multitude of individuals trading the bets among themselves. This form of person-to-person betting has lately experienced rapid growth. Empirical research on the prediction accuracy of bookmaker odds is well established in the literature. While some papers document a good forecasting performance of bookmaker odds (e.g., Boulier and Stekler, 2003 and Forrest et al., 2005), other research provides evidence of biases in bookmaker predictions. However, these biases turn out to be rather small, and thus hardly provide opportunities to systematically beat the odds (e.g., Cain et al., 2000, Dixon and Pope, 2004 and Goddard and Asimakopoulos, 2004). Furthermore, there is a growing body of literature concerned with the predictive power of bet exchange markets. It is found that these markets exhibit high predictive accuracy, as they regularly outperform non-market forecasting methods (e.g., Berg et al., 2008, Forsythe et al., 1992, Snowberg et al., 2008, Spann and Skiera, 2003 and Wolfers and Leigh, 2002. The coexistence of different betting markets offering quotes on the very same event enables us to compare their predictive power. Surprisingly, examples of this kind of research are rare.1 To the best of our knowledge, this paper is the first to contrast the forecast accuracy of the bookmaker market with that of a major betting exchange. Using a dataset covering all football matches played in the major leagues of the “Big Five” (England, France, Germany, Italy, Spain) during three seasons (5478 matches in total), we compare the prediction accuracy of eight different bookmakers’ odds with the forecasting power of the corresponding odds traded at Betfair, a common bet exchange platform. Our results indicate that the prices of the bet exchange market exhibit higher power than the bookmaker odds. Furthermore, we develop a simple betting strategy in order to test the economic relevance of our findings. We show that a strategy of selecting bets for which the bookmaker announced lower probabilities (and thus, offered higher odds) than the person-to-person market, is capable of yielding above average, and in some cases even positive returns. This betting strategy is not restrictive in terms of betting opportunities. Our findings contribute to the ongoing discussion about the predictive properties of different market structures by providing empirical evidence of the superiority of exchange betting in delivering more accurate forecasts of the outcomes of sporting events.

A considerable amount of research has been conducted on separately testing the prediction accuracy of different betting market settings. This paper exploits the coexistence of different market structures offering odds on the same event in order to provide an inter-market comparison of the predictive power of bookmakers and a major betting exchange. We analyze a dataset covering 5478 matches of the major European football leagues and containing the odds of eight bookmakers, together with corresponding prices of the leading person-to-person betting platform Betfair. Our results reveal a clear superiority of the betting exchange over the bookmaker market. First, we estimate a univariate probit regression to explain the actual outcome of a certain bet with the implicit probabilities of the different markets. The goodness-of-fit measures indicate that the bet exchange prices predict the actual match results better. Second, we rerun this regression for all of the bookmakers and include a variable capturing the difference between the two different markets’ implicit probabilities. The estimated coefficient of this variable suggests that the bet exchange has additional explanatory power beyond the bookmakers’ odds. Finally, we assess the economic relevance of the previous results. A simple betting rule of selecting bookmaker bets for which the average bookmaker offers lower probabilities (higher odds) than the bet exchange is capable of generating abnormal, and in some cases even positive, returns. However, we are reluctant to interpret these findings as a failure of the bookmakers to process relevant information. The underlying reasons for the higher prediction accuracy of the bet exchange market are not clear a priori. Bettors with more accurate information and beliefs may self-select into the exchange market while less skilled bettors may place their bets in the bookmaker setting. Alternatively, our findings could be due to the different market structures dealing with similar but potentially biased demand. Bookmaker odds may reflect not only the dealer’s true prediction of the outcome but also his (profit-maximizing) response to the expected (biased) demand. As Levitt (2004), Forrest and Simmons (2008) and Franck et al. (2010) suggest, bookmakers actively shade prices in the presence of a partly irrational betting audience in order to increase their profits. With regard to our findings, the price impact of a biased demand may be less pronounced in the person-to-person situation than in the bookmaker market setting. Nevertheless, a proper examination of these suggestions lies beyond the scope of this paper, and needs further research.