ABSTRACT: In the 2026 offseason, the Buffalo Sabres hired a Data Scientist and a Data Engineer. We received 1000+ applications for these positions, but only about half of those were from human beings. We asked participants in our evaluation process to complete a detailed technical assessment of their data science skills, but only about 2/3 of candidates who passed that technical assessment adequately answered follow-up questions relating to what they did. In this talk, we discuss our strategies when hiring for sports analytics in the age of LLMs and autonomous agents, and we share our thoughts on best practices for both employers and applicants moving forward. Candidates were rated on the degree to which they relied on LLM assistance in their technical assessments. Those that passed the assessment were subsequently evaluated on a variety of soft and hard skills throughout the interview process. We find that candidates who relied heavily on AI in their technical assessments underperformed expectations during the technical interviews. We also find that candidates who scored highly on subjectively-assessed "soft skills" (e.g. communication, idea generation, capacity to learn) more frequently met or exceeded the expectations set by their performance on the technical assessment. We close with a discussion of our specific technical interview questions for sports analytics, with accompanying strategies for both interviewers and interviewees.

ABSTRACT: How many times has the sports analytics world changed in the last 20 years? What were the driving forces behind the evolution of the quantitative role in the front office and the subsequent adoption by more traditionally minded leadership? And how much has the work been driven by technology that threatens to outpace the analyst? I will review the long-running relationships among the quantitative analysts, the traditionally minded executive class and the recent boom in communications driven by agentic coding and large language models in the context of history, reflect on the current state of the industry, propose means for the next class of analysts to position themselves for success in the next wave.

ABSTRACT: This talk will provide an inside look at the role of a Sports Analytics Engineer at ESPN and the work of the teams that support analytics across a wide range of sports. It will explore how sports data is collected, analyzed, interpreted, and used to power both fan-facing experiences and behind-the-scenes decision-making. The presentation will also touch onLauren's career path to ESPN, offering context on the skills, experiences, and opportunities that led to a role in multi-sport analytics. Students will gain a real-world perspective on the day-to-day work, challenges, and rewards of working in sports analytics within a media company.

ABSTRACT: Imagine you're watching the World Cup. The French striker ended the tournament with 10 goals, while the one from Cabo Verde scored 4. Who is actually better? Raw statistics can't answer that, because they mix a player's own ability with the team around them. Answering it properly means building a model that can imagine a world that never happened -- one where both players wore the same shirt. In this session, we'll build a model that can, live in PyMC, from scratch. We'll start with a simple model of goal-scoring and uncertainty, add Gaussian processes so that form can rise and fall across a career, and then add team factors that strip out context to leave individual skill behind. We'll finish with the question that matters most: does it actually work? We'll look at out-of-sample validation, and at a cautionary tale from our own model search -- a variant that gained points in-sample but lost under cross-validation. No Bayesian background needed. If you've fitted a regression, you'll be able to follow. Notebooks provided, no setup required. Paper: Unveiling True Talent: The Soccer Factor Model for Skill Evaluation (Andorra & Göbel, 2024), arXiv:2412.05911 Live platform: https://www.soccerfactormodel.com By registering for and participating in this workshop, you acknowledge that this workshop will be audio and video recorded for use in the Learning Bayesian Statistics podcast without compensation or further approval.

ABSTRACT: In 2011, a college student walked into the MIT Sloan Sports Analytics Conference and watched two research presentations that would foreshadow the future of basketball analytics; Kirk Goldsberry's pioneering shot charts and a small USC spinoff called Second Spectrum analyzing rebounding, both powered by a new technology: SportVU's 2D optical player tracking. That moment marked the beginning of a data revolution in basketball - and crystalized the path for that student's career. This talk traces the evolution of motion tracking in the NBA, from those early 2D coordinates to the introduction of 3D skeletal tracking, which captures the full movement of every player's body in real time. With this leap in granularity has come an explosion of both data and possibility. We'll briefly explore the underlying tracking technology, data pipelines, and data structures before diving into some of the applications this data now enables - from new statistical measures and enhanced decision-making at the NBA Replay Center to a 3D game recreation engine that allows fans to experience the game like never before. Whether you're a researcher, an aspiring analyst, or simply a fan of the game, this talk will show how the richest dataset in sports is reshaping basketball from the court to the couch.

ABSTRACT: If professional teams can accurately predict the order of their league’s draft, they would have a competitive advantage when using or trading their draft picks. Many experts and enthusiasts publish mock drafts: forecasts of the order players will be drafted into professional sports leagues. Using a novel dataset of mock drafts for the National Basketball Association (NBA), we explore mock drafts’ ability to forecast the actual draft. We analyze authors’ mock draft accuracy over time and ask how we can reasonably aggregate information from multiple authors. For both tasks, mock drafts are usually analyzed as ranked lists, and under this assumption, we propose ways to improve on these methods. We propose that rank-biased distance is the appropriate error metric for measuring accuracy of mock drafts as ranked lists. To best combine information from multiple mock drafts into a single consensus mock draft, we also propose a combination method based on the ideas of ranked-choice voting. We show that this method provides improved forecasts over the standard Borda count combination method used for most similar analyses in sports, and that either combination method provides a more accurate forecast across seasons than any single author. To conclude, we will discuss ongoing approaches to relaxing the ranked-list assumption.

ABSTRACT: The rise of pitcher injuries in Major league Baseball the past decade has become an epidemic. The defining questions around pitcher injuries have turned from prevention to treatment and management. Determining the factors that are impacting pitcher health is essential in ensuring the longevity of the sport and the well-being of its players. One of the hypothesized factors behind these concerning trends is pitching while fatigued. Starting pitchers often throw at high velocities for several innings and fatigue leads to changes in performance. We implement a novel hierarchical Bayesian change-point model to identify trends in the changes of a pitcher’s pitch characteristics and biomechanics over the course of multiple games throughout the season. Each player is unique in the types of pitches they throw, their biomechanics, and how they fatigue. Our hierarchical framework allows us to capture these differences between pitchers in terms of locations of when their mechanics change and the slopes of the trends before and after the change-point during an outing. The model sets a baseline for players and allows teams to identify abnormal changes. We examine velocity, horizontal and vertical movement, throwing shoulder abduction and external rotation, and torque applied to trunk and upper arm. Our model allows us to develop a robust understanding of physical changes in a player’s mechanics and pitch characteristics throughout a game.