Pace and Power: Removing Unconscious Bias from Soccer Broadcasts

Sep 13, 2026

Upload by Kyle

This research challenges common stereotypes in professional soccer related to race and gender by using computer-vision match recreations. Player location and body-pose information are used to generate broadcast-like views with player skeletons, allowing audiences to analyze games without visual cues about players’ race or gender. Viewers are asked about segments from men’s and women’s games with different racial backgrounds. Results show that identifying player race increases attribution of athletic or physical traits to Black players, and suggest women’s games may be judged as lower quality when viewers can recognize them.

LanguageEnglish

CategoryPresentations

FormatPDF

Pages20

Reading Time7 min

AccessFree to download

Introduction

Motivation and background

“Pace and power” as a stereotype

Methodology

Computer vision to generate body pose renders

Camera calibration and template replacement

Pose estimation and skeleton generation

Pace and Power: Removing Unconscious Bias from Soccer Broadcasts Sam Gregory, Victoria University & Inter Miami CF Devin Pleuler, Toronto FC Daniel Daly-Grafstein, University ofBritish Columbia Yang Liu, Sportlogiq Paul Marchwica, Sportlogiq Abstract This research challenges common stereotypes in professional soccer based on race and gender by employing computer vision based match recreations. Using player location and body pose information we recreate match broadcasts with player skeletons so that the viewer can analyze a match without any visual information about the players’ race or gender. We ask a series of questions about a video segment involving teams of different racial backgrounds and video segments from a men’s game and a women’s game. We use the broadcast recreations as a control group and show a second group the original broadcasts. We find that when viewers are able to identify player race they are much more likely to attribute athletic or physical characteristics to Black players. We also find some evidence that when viewers can tell they are watching a women’s game they may identify it as of a lower quality than when they are watching therecreation. 1. Introduction Racial and gender biases pervade every walk of life and sports are no different. From broadcast commentary (Eastman and Billings 2001) to scouting reports (Boylan, McMahon and Munroe 2017) to refereeing (Parsons et al 2007), almost every aspect of sports has been affected by biases based on gender and race. This paper seeks to provide a framework to highlight and challenge some of these biases in how we analyze professional soccer. When analyzing a team’s style of play or individual player characteristics in soccer, many factors, both conscious and unconscious, come into play. Perception of a player may be biased by the player ’s gender, race, or other physical characteristics that have nothing to do with the player ’s underlying ability. Decisions based on these analyses can have serious detrimental effects including but not limited to: unfair decisions made by front offices and referees; perpetuation of racist or sexist stereotypes by commentators or pundits; and inefficient allocation of media coverage or resources. Wider societal effects that reach beyond the sport itself are also exacerbated by these biased analyses. This paper derives its name from one of the more widespread of these biases: the term “pace and power” has been particularly criticized as a lazy stereotype for describing Black players (Madu 2018) . A study found that in the most popular Dutch soccer television program RTL Voetbal, 22% of all comments were about players of Surinamese descent whereas 46% of the comments made about physicality were in reference to this group of players (van Sterkenburg, Knoppersand de Leeuw 2012) . Using state of the art computer vision techniques, we collect player location and body pose information in order to recreate match broadcasts with player skeletons. This approach allows us to remove any biases viewers may have based on the appearance of the players or the production quality of the broadcast. We then survey a group of sports fans, asking them to identify trendsand style of play characteristics from either the original broadcast or the body pose renders with player skeletons. By comparing the responses of those who saw the original broadcast to those who saw the virtual render of the game showing only the skeletons ofthe players, we were able to identify areas where gender or race likely influenced the analysis. 2. Methodology To challenge and highlight these biases we use two basic frameworks: a pipeline of computer vision techniques to recreate skeletal broadcasts using 2D body pose and a survey to analyze whether viewers can identify the same basic trends from broadcasts and the recreations while highlighting those which may be influenced by racial or gender bias. 2.1. Computer vision to generate body pose rend

FAQs

How does the study remove race and gender cues from soccer broadcasts?

It recreates match broadcasts using computer vision to generate player skeleton views based on player location and 2D body pose, so viewers cannot rely on players’ appearance or broadcast visuals tied to race or gender.

What comparison groups does the study use to measure bias?

It uses broadcast recreations as a control group and compares responses from viewers who saw the original broadcasts against those who saw skeleton-based recreations.

What effect does identifying player race have on viewers’ judgments?

When viewers can identify player race, they are more likely to attribute athletic or physical characteristics to Black players.

What evidence does the study find about judgments of women’s games?

The study finds some evidence that when viewers can tell they are watching a women’s game, they may rate it as lower quality than when they are watching the recreation.

Sep 13, 2026

Upload by Kyle