Author: Nikolay Shmakov
Title of host publication: Proceedings - 2024 International Sports Analytics Conference and Exhibition, ISACE, Part of the book series: Lecture Notes in Computer Science, 2024
Abstract: Hockey clubs are currently subject to strict budgetary and regulatory restrictions, which increase competition and struggle for the best players, according to the ratio between expected salaries and final utility to the team. These restrictions force clubs to evaluate the effectiveness of player contracts and continuously benchmark them with the transfer market to find reinforcements. This article aims to create a framework for evaluating any hockey player in any professional league. The built framework is based on the Expected Goals model (XG). The XG model calculates the probability of each shot being a goal by solving a binary probabilistic classification task. The final rating is calculated using the explainable and auditable Expected Goals on Target model (XGoT), which complements the classical Expected Goals model by adding information about points on the gate plane for the shots on goal. This introduces the realization factor into the model. This article expands the discussion of the XG methodology for evaluating player efficiency and puts forward a new way of player evaluation in ice hockey. The traditional approach of rating creation in hockey has been strengthened by the implementation of the opposition level that balances the game in equal strengths, power play, and penalty kill to assemble an overall rating. The resulting fully interpretable framework facilitates stakeholders to make transfer decisions based on reliable analytics.
Title of host publication: Proceedings - 2024 International Sports Analytics Conference and Exhibition, ISACE, Part of the book series: Lecture Notes in Computer Science, 2024
Abstract: Hockey clubs are currently subject to strict budgetary and regulatory restrictions, which increase competition and struggle for the best players, according to the ratio between expected salaries and final utility to the team. These restrictions force clubs to evaluate the effectiveness of player contracts and continuously benchmark them with the transfer market to find reinforcements. This article aims to create a framework for evaluating any hockey player in any professional league. The built framework is based on the Expected Goals model (XG). The XG model calculates the probability of each shot being a goal by solving a binary probabilistic classification task. The final rating is calculated using the explainable and auditable Expected Goals on Target model (XGoT), which complements the classical Expected Goals model by adding information about points on the gate plane for the shots on goal. This introduces the realization factor into the model. This article expands the discussion of the XG methodology for evaluating player efficiency and puts forward a new way of player evaluation in ice hockey. The traditional approach of rating creation in hockey has been strengthened by the implementation of the opposition level that balances the game in equal strengths, power play, and penalty kill to assemble an overall rating. The resulting fully interpretable framework facilitates stakeholders to make transfer decisions based on reliable analytics.