A data-driven performance index for center forwards in the English Premier League

Abstract

Abstract Despite the growing availability of performance data in professional soccer, existing player rating systems lack positional specificity and fail to capture the multidimensional demands of center forwards. Building on composite index and regularized regression approaches in sports analytics, this study develops a Soccer Performance Index (SPI) tailored to center forwards in the English Premier League (EPL) across the 2021–2024 seasons. Data from 194 player-season observations were analysed using 109 Wyscout performance metrics as predictors. Three SPI versions were constructed using Lasso and Ridge regression: one based on market value, one on the InStat Index, and a hybrid combining both. The hybrid model, weighting market value at 70% and the InStat Index at 30%, achieved the strongest explanatory fit (R 2 = 0.676, RMSE = 0.669), accounting for approximately 68% of variance in player valuation — a result consistent with the complexity inherent in behavioural and performance modelling contexts. xG per 90, shots, and key passes per 90 emerged as the strongest predictors. Methodological considerations include the use of a minimum participation threshold (≥ 20 matches), which may introduce survivorship bias; the treatment of player-season observations as independent units, which does not account for repeated-measures dependence; and the reliance on internal validation only. The SPI demonstrates ecological relevance by integrating both financial and on-field performance indicators, offering a structured framework for talent identification and recruitment support in applied professional contexts.