Oliver Schulte and Guiliang Liu, Simon Fraser University: What i
s the value of an action in ice hockey? Deep Reinforcement Learning for Co
ntext-Aware Player Evaluation
Date
–
Location
ICCSX836
Title: What is the value of an action in ice hockey? Deep Reinf
orcement Learning for Context-Aware Player Evaluation
Abstract:
A fundamental goal of sports analytics is to rank player performance. A c
ommon approach is to assign a value to each player action and rank a playe
r by his or her aggregate action value. For measuring the value of an acti
on, a recent AI-based approach, successful in a variety of team sports,estimates its expected impact on team success (e.g., the team’s chance o
f scoring the next goal). We introduce a high-resolution neural network re
presentation of the expected action value, which integrates both continuo
us context signals and the recent match history. Deep Reinforcement Learn
ing is used to learn an action-value Q function from 3M play-by-play event
s in the National Hockey League (NHL). Empirical evaluation shows that theresulting player ranking is consistent throughout a play season, and cor
relates highly with standard success measures and future salary. A full ve
rsion of the paper is available at https://www.ijcai.org/proceedings/2018/
478.
Bio: Oliver Schulte is a Professor in the School of Computin
g Science at Simon Fraser University, Vancouver, Canada. He received hisPh.D. from Carnegie Mellon University in 1997. Current research focuses o
n machine learning for structured data, such as events, networks, and r
elational databases. He has published papers in leading AI and machine lea
rning venues on a variety of topics, including sports analytics, learnin
g Bayesian networks, learning theory, game theory, and scientific disco
very. While he has won some nice awards, his biggest claim to fame may bea draw against chess world champion Gary Kasparov.
Guiliang Liu is aPh.D. candidate at Simon Fraser University working in reinforcement learn
ing.
Find more undergrad events on our internal portal at https://my.cs.ubc.ca.
This event's address: https://my.cs.ubc.ca/event/2019/03/caida-applications-seminar