A Bayesian circular mixed-effects model for explaining variability in directional movement in American football

Quang Nguyen, Ronald Yurko · 2025

Read A Bayesian circular mixed-effects model for explaining variability in directional movement in American football by Quang Nguyen, Ronald Yurko (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-06122.txt.

Introduction

Quang Nguyen and Ronald Yurko model how football players change direction in a 2025 paper. Sharp turns are a core skill for ball carriers, yet coaches lack a fair way to grade that skill from game data. The pair use tracking data from the first nine weeks of the 2022 NFL season.

Their model reads the turn angle between each pair of steps a carrier takes. It lets both the mean turn and its spread vary with game context and with the player. Wide receivers after a catch and running backs after a handoff are the focus. The player effects single out the shiftiest carriers, those whose turns vary the most.

Catalog record

CREATORQuang Nguyen, Ronald Yurko
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE37 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-06122

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A Bayesian circular mixed-effects model for explaining variability in directional movement in American football Quang Nguyen, Ronald Yurko Abstract Change of direction is a key element of player movement in American football, yet there remains a lack of objective approaches for in-game performance evaluation of this athletic trait. Using tracking data, we propose a Bayesian mixed-effects model with heterogeneous variances for assessing a player’s ability to make variable directional adjustments while moving on the field. We model the turn angle (i.e., angle between successive displacement vectors) for NFL ball carriers on both passing and rushing plays, focusing on receivers after the catch and running backs after the handoff. In particular, we consider a von Mises distribution for the frame-level turn angle and explicitly model both the mean and concentration parameters with relevant spatiotemporal and contextual covariates. Of primary interest, we include player random effects that allow the turn angle concentration to vary by ball carrier nested within position groups. This offers practical insight into player evaluation, as it reveals the shiftiest ball carriers with great variability in turning behavior. We illustrate our approach with results from the first nine weeks of the 2022 NFL regular season and explore player-specific and positional differences in turn angle variability. Keywords: Bayesian statistics, circular statistics, heterogeneity, mixed-effects model, National Football League, player tracking data 1 Introduction 1.1 Background: change of direction in American football American football is a dynamic and intricate sport due to the complexity of many athletes interacting over the course of every play. Player movement in football consists of a variety of elements that together contribute to athletic performance on the field. One vital component of player movement is change of direction, which demonstrates the ability of a player to rapidly and efficiently alter their trajectory. On offense, players use this movement skill to create separation and evade tackles—whether it is a running back cutting through traffic, a receiver breaking off a route, or a quarterback escaping pressure. On defense, change of direction enables defenders to close space quickly and maintain leverage against the opposing players on offense. In the National Football League (NFL), the most commonly-known way to assess change of direction is through timed physical drills at the NFL Scouting Combine, namely the three-cone and short shuttle. In a three-cone drill (or L-drill), a player is asked to run, stop, and loop around three cones arranged in a right angle “L” shape. The short shuttle, also known as the 20-yard shuttle or 5–10–5 drill, is a test that requires an individual to first sprint five yards to one side, reverse and sprint ten yards the opposite way, then run five yards back to the starting point. Both of these tests are designed to evaluate agility, quickness, and lateral movement of NFL prospects. These traditional drills, while providing some baseline for evaluating change of direction, have apparent limitations. Specifically, they fail to capture how players respond in live, real-time scenarios where movement is influenced by spatial constraints and different moving pieces on the field. As a side note, there are other timed drills in sports science related to the aforementioned NFL Combine evaluations. These include the T-test, 505, Illinois, and Arrowhead agility tests, to name a few. As an example, the 505 test focuses on 180-degree turning performance and involves a 15-meter sprint, followed by a 180-degree turn and a 5-meter sprint back. Based on this test, one can accordingly calculate the so-called change of direction deficit by computing the difference between the 505 time and the time it takes for a straight-line sprint of an equivalent distance. We also note a substantial number of articles in the sports science literature that analyze change of direction from a performance training perspective (see Brughelli et al., 2008 for a thorough review). Within this body of work, the goal is usually to link change of direction with factors like strength, conditioning, fatigue, or injury risk. Recently, Giles et al., 2019 and Giles et al., 2024 use camera tracking data to detect and classify directional movement in tennis. Since all of the above contributions do not attempt to model in-game change of direction performance of athletes, they are of limited relevance to our paper. Despite its importance, there are currently no widely-adopted public approaches for evaluating directional movement in American football using in-game data, to the best of our knowledge 1 1 1 We are aware that NFL teams and vendors do have proprietary methods for evaluating in-game change of direction. For example, Teamworks Intelligence uses ideas from the aforementioned change of direction deficit concept in sports science to develop their own metric ( Kassam, 2025 ) . . However, with the increasing availability of player tracking data, there are ample opportunities to quantify movement traits like change of direction directly from observed on-field information. Below, we provide a brief summary of recent developments in football analytics with tracking data. 1.2 Background: player tracking data in American football The advent of player tracking data has transformed the field of sports analytics in recent years ( Baumer et al., 2023 ; Kovalchik, 2023 ) . In the NFL, tracking data are collected via the Next Gen Stats system with radio frequency identification (RFID) chips placed inside the football and shoulder pads of players. This outputs spatial location and trajectory information for all 22 players on the field during any given play at a rate of 10 Hz (i.e., 10 frames per second). The granularity and continuous nature of tracking data enable in-depth exploration of various topics in American football that were not previously possible with traditional box-score statistics or play-level data. In an attempt to crowdsource public insight into player tracking data, the NFL organizes an annual contest called the Big Data Bowl ( Lopez, 2020 ) . Every year since 2019, the NFL releases a sample of player tracking data to accompany a football-related Big Data Bowl theme such as pass coverage, special teams, linemen, tackling, and pre-snap motion. The competition has consequently established the academic literature on NFL tracking data, promoting data-driven innovations in football analytics ( Lopez, 2020 ) . Recent developments in this area include models for hypothetical completion probability ( Deshpande and Evans, 2020 ) , receiver route detection ( Chu et al., 2020 ) , pass coverage identification ( Dutta et al., 2020 ) , and within-play expected points ( Yurko et al., 2020 ) ; as well as novel performance metrics for pass rush pressure ( Nguyen et al., 2024 ) , tackling ( Nguyen et al., 2025 ) , and snap timing ( Nguyen and Yurko, 2025 ) . It is important to note that some of the aforementioned contributions offer insights into previously understudied areas of American football like defensive performance and pre-snap actions. 1.3 Our contribution In this paper, we introduce a modeling framework to assess in-game change of direction ability of players in American football. We first characterize change of direction at the frame level using the concept of turn angle, borrowing from the animal movement literature. This simply measures the angular deviation between consecutive movement steps of NFL players during any given play. Our primary quantity of interest is the variability in

Read the complete text