NBA processes 16 gigabytes of player data per game on Azure
The NBA has developed a Microsoft Azure-based AI solution using Azure Kubernetes Service (AKS) to process real-time player movement data, including speed, dunk height, and injury risk, gathered through computer vision. This system processes up to 16 gigabytes of raw data per game to provide insights for player performance improvement and future plans for live statistic overlays on viewer screens. The NBA aims to continue building new learning models and support statistic overlays for fans using Azure AI infrastructure.
Key Takeaways
- The NBA says its Azure Kubernetes Service setup handles about 15 to 16 gigabytes of raw data per game, not including RGB video signals.
- Computer vision tracks 29 points on each player’s body at 60 hertz, giving the league subsecond player-movement data.
- Azure Cosmos DB stores the metadata, while a Redpanda cluster runs on Azure virtual machines for NBA API and social-media delivery.
- Charlie Rohlf said the new system is already helping coaches spot patterns in player strengths and weaknesses.
- Caroline McKee said the NBA is planning live stat overlays, including shot-release height and defender proximity, for viewer screens.
Why It Matters
The immediate impact is operational: the NBA has moved from manual box-score collection to a deployed, real-time pipeline that ingests millions of tracking points per game. That matters for the broader media and sports tech stack because the league is using Azure Kubernetes Service, Cosmos DB, and Azure virtual machines to run both internal performance analysis and fan-facing data delivery. The next signal to watch is whether the NBA expands beyond current shot and movement metrics into the more sophisticated Azure AI models and GPU-accelerated virtual machines McKee says are under consideration.
Read full article at microsoft.com