ABSTRACT
Undoubtedly, the most popular object of research is the famous favourite-longshot bias (FLB). Initially, the bias was observed in horseracing, but subsequent research has documented its existence in a variety of sports betting markets (for a survey of studies see Sauer (1998) and Vaughan Williams (2005)). A FLB exists when the favourites’ (longshots’) winning probability, as implied by the odds, is underestimated (overestimated). However, the reverse phenomenon has also been documented (e.g. Woodland and Woodland, 1994) and is usually referred to as ‘reverse’ or ‘negative’ FLB. The literature concerning the FLB in the football betting market is now reviewed. Pope and Peel (1988) investigated the UK market (data derived from the 1981/1982 season) and concluded that favourites seem to be more profitable compared to longshots. Cain et al. (2000) analysed a data set of 2,855 games played in the United Kingdom during the 1991/1992 season and found some evidence that market odds underestimate the winning probability of heavy favourites (including the probability of frequent (low) exact scores). Similar evidence was presented by Malaric et al. (2008), who explored a data set of 12,218 games played in ten European leagues in the period 1999-2002. Deschamps (2008) also documented FLB associated with several European leagues in the 2005/2006 season, with more pronounced effects in second-division leagues. Finally, Vlastakis et al. (2009) observed that the market underestimates merely the winning probabilities of favourites playing away from home. The ‘draw’ outcome is more frequent in football compared to other sports and therefore it is interesting to investigate whether its existence influences the FLB in any way. Deschamps and Gergaud (2007) explore a data set of 8,377 football matches played between the 2002/2003 and 2005/2006 seasons in English leagues and observe ‘positive’ FLB concerning the odds of the home and away teams. However, for the odds on the draw outcome a reverse FLB is identified. Additionally, it is found that the probability for a draw is, in general, underestimated by market odds. Consequently, betting on ‘draw’ yields superior returns than betting on the home or the away team winning. Variation in the magnitude of FLB has also been associated with the level of transaction costs and with league-specific characteristics, including competitiveness. Paton and Vaughan Williams (1998) found that the fixed-odds football betting market, where transaction costs are higher, exhibits higher FLB
compared to the spread-betting market, where transaction costs are lower. Oikonomidis and Johnson (2008) suggested that if bookmakers fail to fully account for cross-league differences in competitiveness, heterogeneity in the magnitude of the FLB would be expected. This hypothesis was confirmed by their analysis of a sample of over 56,000 football matches played in 22 European leagues over the last decade; the level of league competitiveness almost completely determined the degree of FLB in each league, with relatively competitive leagues exhibiting significantly higher bias. Conclusively, it can be stated that significant ‘positive’ FLB exists in the football betting market. The fact that it has been documented across different samples and is shown to be persistent across years points to the fact that it is a structural idiosyncrasy of the market.
