Accounting for Missing, Censored, and Truncated Data in Cricket Batting Performance Analysis

Abstract

Cricket performance data often contain missing values, truncated observations, and censored outcomes that distort statistical evaluations of batsman ability. This study applies an integrated quantitative framework to examine the effects of incomplete data on batting performance and to propose robust analytical solutions. Using a dataset of 2,882 innings from 223 simulated batsmen, the research identifies systematic missingness patterns—primarily Missing at Random (MAR)—that lead to inflated performance estimates when complete-case methods are used. Multiple imputation effectively restored unbiased mean scores and preserved distributional properties. Right censoring from not-out innings was shown to significantly affect performance metrics, highlighting the limitations of traditional batting averages and the value of survival analysis techniques such as Kaplan–Meier estimators and Cox proportional hazards models. Truncation, particularly the exclusion of players with short careers, introduced survivorship bias that artificially elevated overall batting statistics. Truncated regression and selection models corrected for this bias, enabling fairer comparisons across players. The study demonstrates that addressing missing, truncated, and censored data is essential for valid performance analysis in cricket and provides a comprehensive framework for producing more accurate and reliable estimates of batting ability. These findings have practical implications for sports analytics, talent identification, and evidence-based decision-making in cricket.