Brendan Kumagai

Brendan Kumagai

Data Scientist

Zelus Analytics


I am a Data Scientist on the Hockey Research & Development team at Zelus Analytics where we develop advanced statistical and machine learning models using cutting-edge data to improve our understanding of the game of hockey and provide clients with objective, data-informed insights. I specialize in applications of statistical methods to spatiotemporal tracking data in hockey. I hold a M.Sc. in Statistics from Simon Fraser University where I studied under the supervision of Dr. Tim Swartz with research focused on the NHL draft.

Previously, I graduated with a B.Sc. in Mathematics and Statistics from McMaster University in April 2020 and completed Data Analyst and Data Scientist internships with the McMaster Visual Neuroscience Lab, the Canadian Tire Sports Analytics team and Stathletes.

My career objective is to apply data science methods to develop innovative and original research in a field that I am passionate about such as neuroscience, biomechanics, conservation and sustainability, and especially sports with hopes of becoming a leader at the intersection of data science and hockey in the future.


  • Sports Analytics and Operations
  • Machine Learning
  • Spatiotemporal Statistics
  • Bayesian Statistics


  • B.Sc. in Mathematics and Statistics, Sept 2016-Apr 2020

    McMaster University

  • M.Sc. in Statistics, Sept 2021 - Dec 2022

    Simon Fraser University



Data Scientist

Zelus Analytics

May 2022 – Present Remote

Data Scientist Intern


Dec 2021 – Apr 2022 St. Catharines, ON, Canada

Data Analyst Intern

Canadian Tire Financial Services, Sports Analytics

Jan 2021 – Aug 2021 Toronto, ON, Canada

Data Analyst

McMaster University, Visual Neuroscience Lab

May 2020 – Dec 2020 Hamilton, ON, Canada

Projects & Talks

Dynamic Prediction of the NHL Draft with Rank-Ordered Logit Models

This project began as joint work with Ryker Moreau and Kimberly Kroetch and developed into my thesis project for my M.Sc. in Statistics under the supervision of Dr. Tim Swartz. We adapt the rank-ordered logit model - a sequence of conditional multinomial logit models - to create a novel way to forecast the outcome of the NHL draft using draft rankings from hockey experts and previous NHL draft results.

Bayesian Velocity Models for Horse Race Simulation

This project was joint work with Tyrel Stokes, Kimberly Kroetch, Gurashish Bagga, and Liam Welsh for the 2022 NYRA/NYTHA Big Data Derby competition. We leverage the horse racing tracking data provided in the competition to develop Bayesian B-spline regression models to forecast the probability that each horse will finish in each possible finishing placement. Our team was awarded 1st place out of 108 submissions.

Punt Returns: Using the Math to Find the Path

This project was joint work along with Ryker Moreau, Elijah Cavan, and Robyn Ritchie for the 2022 NFL Big Data Bowl. Our work involved designing and implementing an algorithm to find the ball carrier’s optimal path on punt return plays. We were awarded 1st place out of 268 total submissions.

The Shot Chain

Joint project with Tyrel Stokes, Mikael Nahabedian, and Thibaud Châtel presented at WHKYHAC 2021. Our work here goes beyond the traditional expected goals model by building out a chain of Bayesian logistic models to analyze shooter talent and the context of shots in women’s hockey.

Bayesian Space-Time Models for Expected Possession Added Value

This project was joint work along with Tyrel Stokes, Mikael Nahabedian, and Thibaud Châtel for the 2021 Stathletes’ Big Data Cup. Our work involved quantifying the value of each event in offensive entry-to-exit sequences through the use of various Bayesian spatiotemporal models and a Markov Decision Process. Our team was awarded 1st place out of 71 submissions.

Clustering and Analyzing 5v5 NHL Shot Location Data

A poster presenting my work on clustering 5v5 NHL shot maps at the 2020 Carnegie Mellon Sports Analytics Conference poster competition. I find 10 clusters of players based off of their shooting patterns and analyzing the composition of these clusters.

Point Trend Values

In this project, I create a new metric called Point Trend Values to quantify trends in a player’s cumulative point production and explore the predictive power of this metric on OHL and WHL forwards. I was awarded a tie for 1st place at the 2020 U-Conn Sports Analytics Symposium (UCSAS) Poster Competition.