Abstract:
In the sport of cricket, player ability is generally assessed using traditional statistics, such as batting and bowling averages. However, such measures fail to account for variations in ability that may occur over the short-term, during a match, and over the long-term, between matches. As a result, batting and bowling averages are unable to distinguish between players whose abilities are declining, and those still yet to reach peak performance. This is a major shortcoming of many proposed measures of cricketing ability; coaches and selectors often cite recent performances, or form, as a reason for dropping or selecting certain players, but have no means of quantifying how such factors may impact a player’s true, underlying ability.
In order to detect and quantify temporal variations in ability that may be observed over the course of a playing career, a set of Bayesian parametric models are derived to measure and predict the career trajectories of professional cricket players. Career trajectories are modelled using a Gaussian process and aim to estimate a player’s past, present, and future abilities, accounting for recent form, and a number of contextual variables that are frequently ignored by alternative measures. A simulation-based method of predicting the outcome of upcoming matches is then proposed. The match-simulation algorithm takes predictions of ability obtained from the estimated batting and bowling career trajectories as inputs and attempts to quantify the likely performance and contribution of individual players in a given match.
Generally speaking, the results suggest that underlying batting and bowling ability does not fluctuate significantly in the short-term as a result of recent form. Instead, ability appears to improve and deteriorate slowly over time, likely as a result of players gaining experience in a variety of match conditions; participating in specialised coaching programmes; and due to changes in physical attributes, such as fitness and eyesight. These findings may have practical implications in the likes of player comparison, talent identification, and team selection policy, as coaches and selectors are able to better quantify player ability and understand the individual-specific risks and rewards of selecting certain players over others.