A variety of machine learning models have been proposed to assess the
performance of players in professional sports. However, they have only a
limited ability to model how player performance depends on the game context.
This paper proposes a new approach to capturing game context: we apply Deep
Reinforcement Learning (DRL) to learn an action-value Q function from 3M
play-by-play events in the National Hockey League (NHL). The neural network
representation integrates both continuous context signals and game history,
using a possession-based LSTM. The learned Q-function is used to value players'
actions under different game contexts. To assess a player's overall
performance, we introduce a novel Game Impact Metric (GIM) that aggregates the
values of the player's actions. Empirical Evaluation shows GIM is consistent
throughout a play season, and correlates highly with standard success measures
and future salary.

本论文提出了一种基于深度强化学习（DRL）的新方法来捕捉游戏背景，该方法使用 3M NBA 比赛的 play-by-play 事件来学习一个行动价值 Q 函数，并引入一种新的游戏影响度量（GIM）来评估玩家的整体表现。评估结果表明，GIM 在整个赛季中保持一致，并与标准成功指标和未来薪资高度相关。

冰球中基于上下文的球员评估的深度强化学习

Deep Reinforcement Learning in Ice Hockey for Context-Aware Player  Evaluation

Despite growing interest in quantifying and modeling the scoring dynamics
within professional sports games, relative little is known about what patterns
or principles, if any, cut across different sports. Using a comprehensive data
set of scoring events in nearly a dozen consecutive seasons of college and
professional (American) football, professional hockey, and professional
basketball, we identify several common patterns in scoring dynamics. Across
these sports, scoring tempo---when scoring events occur---closely follows a
common Poisson process, with a sport-specific rate. Similarly, scoring
balance---how often a team wins an event---follows a common Bernoulli process,
with a parameter that effectively varies with the size of the lead. Combining
these processes within a generative model of gameplay, we find they both
reproduce the observed dynamics in all four sports and accurately predict game
outcomes. These results demonstrate common dynamical patterns underlying
within-game scoring dynamics across professional team sports, and suggest
specific mechanisms for driving them. We close with a brief discussion of the
implications of our results for several popular hypotheses about sports
dynamics.

本研究旨在探讨四项职业团体运动比赛的计分动态，分析了得分节奏与得分平衡的共同模式，并利用生成游戏模型预测比赛结果以及驱动这些模式的机制。