On the longshot bias in tennis betting markets: the CaSco normalization

Abstract

This study focuses on investigating bookmakers’ behavior in the tennis gambling market in presence of a clear underdog. The aim of this paper is threefold. First, it investigates the distance (bias) between the true but unobserved probability of a given sport outcome and the published odd by a bookmaker. Second, it tests the predictive skills of the most widespread normalization methods when a player is clearly favourite on another. Third, it proposes a new normalization method (called CaSco normalization), which takes into account the positive relationship between the bias and the distance between the odds. The empirical analysis relies on sample odds provided by Bet365 about over 27,000 matches from 2005 to 2015. Our findings show that. First, when there is a clear underdog, the bookmaker minimises the losses in case of unexpected outcomes by increasing the bias in the public available odds. Second, the normalization methods which take into account the bias generally perform better than the other alternatives. Third, in-sample forecasts based on CaSco normalization always outperform the other methods and more importantly, the proposed technique always guaranties unbiased normalized probabilities.