Match statistics that discriminate between winning and losing teams in ODI and T20I cricket

Abstract

Background Cricket players and teams have a different strategy for batting for the different formats of cricket, namely Twenty-Twenty International (T20I) and One Day International (ODI). Different application of skills is required for each format of cricket can clearly be seen as mostly a different team is selected for each format of the game in professional cricket. Analysis of performance variables such as boundaries hit by batsmen and runs scored during the power play can be used to predict future success or failure of a cricket team based on the match outcome. This study will provide batting statistics that discriminate between winning and losing teams in ODI and T20I cricket. Furthermore, the study will reveal which variables correlate the highest with successful performance within the different formats of the game. Aims The aim of this study was twofold, firstly to analyse batting data in ODI cricket that discriminate between winning and losing teams. Secondly to analyse batting data in T20I cricket that discriminate between winning and losing teams. Method Sample Ten international teams were selected for the purpose of this study. The ten teams were selected because they all participate in all three formats of cricket namely ODI, T20I, and test cricket. Six matches from each team’s records were randomly selected and observed (3 batting first, 3 batting second). The first aim consisted of conducting analysis of a total of 60 professional ODI cricket matches resulting in 120 records (innings) (both teams involved per match). The second aim consisted of conducting analysis of a total of 60 professional T20I cricket matches resulting in 120 records (both teams involved per match). Drawn matches, and those which employed the Duckworth-Lewis method, were excluded from the study. Measuring instruments Retrospective data from the 2014 and 2015 international cricket season was collected from ESPN Cricinfo website. Data analysis In this research, a strong and reliable data source is needed which was found in Statsguru. Statsguru is ESPN Cricinfo's cricket statistics maintenance database. The data was then analyzed using the SAS statistical software (SAS, 2013). Because of the fundamentally different match situation faced by the team batting first and second, respectively, the data were analysed separately for the team batting first and for the team batting second. The outcome of the match is a binary variable (win/lose) since drawn matches were excluded from the analysis. The association of the potential predictor variables with the match outcome was analyzed using univariate logistic regression, fitting each predictor variable, one at a time. The statistical significance of each predictor variable was tested using an exact test (exact conditional logistic regression); the exact P-value is reported. The analysis was carried out using SAS procedure LOGISTIC (see SAS, 2013). Results For aim 1 the significant predictors of winning an ODI cricket match when batting first were: runs scored in the first 20 overs (p=0.0019), runs scored in the last 12 overs (p=0.0004), sixes scored (p=0.0017), and the number of runs scored among the top four batsmen (p=0.0015); For aim 1 the significant predictors of winning an ODI cricket match when batting second were: fours scored (p=0.0024), sixes scored (p=0.00277), runs scored between the top order batsmen (p=0.0197), and runs scored between the lower order batsmen (p=0.0222). Variables that predict success in ODI cricket differed for teams batting first and second, respectively. For aim 2 significant predictors of winning a T20I cricket match when batting first were: runs scored in the first 5 overs (p=0.0035), runs scored in the last 7 overs (p=<0.0001),