This is a data science, machine learning computer-vision focused project that aims to create a dataset of labeled punches from professional boxing matches that can be used to video AI models for for sports (boxing) analytics. In associated with Boxing Undefeated, it strives to help democratize the world of boxing data.
For decades, boxing analytics have been locked behind corporate walls, creating a two-tier system that benefits the wealthy and powerful while leaving fighters, trainers, and fans in the dark.
CompuBox is the monopoly for punch statistics, used by major networks like HBO, NBC, and ESPN. But here's the shocking reality: CompuBox offers no API, no public database, and no way for anyone outside their exclusive corporate partnerships to access the data they've been collecting for 40 years.
Meanwhile BoxRec(boxing's primary record keeper with over 2.1 million bouts) refuses to offer any API and actively blocks researchers with aggressive anti-scraping measures. As their own users admit: "BoxRec would be the best place to do this, but their data isn't public" and "it's outright hostile to anyone trying to get their data".
Independent developers acknowledge that "boxing data is very hard to come by, there is no single source of truth" and are forced to build complex scraping systems just to access basic information.
This creates a massive inequality: elite promoters like Top Rank, Golden Boy, and Matchroom Boxing have access to sophisticated analytics and million-dollar insights, while amateur boxers, independent trainers, and developing fighters are left to guess.
We're fixing this.
SERP AI is teaming up with Boxing Undefeated to help change the world of boxing.
This project aims to create the first truly open, comprehensive boxing dataset - starting with punch recognition and classification that we will use to break down through the boxing world's gatekeepers and democratize the analytics that should belong to everyone who makes this sport possible: the fighters, trainers, and passionate fans who keep boxing alive.
Our goal is to create a large enough data set that we can use to train artificial intelligence and machine learning models to create an open source, real-time, AI-powered punch tracking and classification software which can be used to help gyms, fightes, analytics, or anyone else with fight analytics, stats and insights -- or for building other types of boxing applications.
Ready to help change boxing?
We are looking for people to help volunteer for this project. Click here to read more about the specifics of contributing.