Stats Perform’s Playbook for Scaling Sports Intelligence

General-purpose large language models can describe sports video at a high level, but they cannot reliably answer the questions that matter most in sport. They may identify that a goal has been scored, but they cannot determine expected goal value, explain which team is controlling momentum, identify tactical formations, calculate player speed, or assess how a moment changes a championship race.

Stats Perform takes a different approach by mapping sport to objective measurements grounded in space, time, and history. Event data captures what happened and where. Tracking data follows the movement of players and the ball at 25 frames per second. Historical data provides context for whether a moment is routine, rare, or record setting.

Together, these layers create a specialized language of sport. This is where Stats Perform’s proprietary advantage becomes critical. Opta is not simply a data source; it is one of the most trusted and widely adopted languages of sport, built over decades of expert collection, validation, historical depth, and domain-specific modeling. By combining this proprietary sports data layer with AI models trained specifically for sport, Stats Perform can deliver outputs that are more accurate, contextual, and explainable than generic AI systems trained only on open web content. NVIDIA accelerates the infrastructure needed to process video and model outputs at scale; Stats Perform provides the sports intelligence layer that makes those outputs meaningful, trusted, and commercially valuable.

This allows precise answers to tactical, fitness, historical, and predictive questions. Whether evaluating the expected threat of a pass, the likelihood of winning after a goal, the impact of runs off the ball, or whether defenders were positioned correctly during a decisive play, the system delivers insight with depth and reliability.

Scaling this intelligence to every match is a significant challenge. High-quality tracking has traditionally relied on in-venue camera systems, operational crews, and controlled capture environments. While precise, this approach cannot easily cover the full volume of games played globally.

Stats Perform instead uses remote feeds, the same video fans watch at home. These feeds introduce difficult AI problems. Cameras move constantly. Players are occluded. Lighting varies by venue and competition. Jerseys, motion blur, pixelation, compression artifacts, low camera angles, close-ups, and missing field of view all reduce consistency.

On average, only about half of the players may be visible in a given frame. Close-up shots can remove nearly all usable tracking context at exactly the moments when insight is most valuable.

To overcome this, Stats Perform developed foundation models that reconstruct complete tracking data from partial remote inputs. By combining computer vision, human-annotated event data, and historical sports context, the system infers where players and the ball are, even when they are not visible.

This unlocks the ability to digitize every pass, run, shot, possession, and tactical shape from remote video at global scale.

With structured data and foundation models in place, Stats Perform is moving toward agentic AI experiences that allow users to interact with sport in more natural and powerful ways.

That value is already being delivered across the sports ecosystem. For broadcasters, it supports richer live analysis and faster access to trusted in-game context. For media and digital platforms, it enables automated storytelling, scalable highlight creation through products such as Opta Pulse, and more personalized fan experiences. For betting operators, leagues, and rights holders, it supports faster, more immersive live experiences through solutions such as Bet LiveStreams and Content Player Pro. Across these use cases, Stats Perform’s AI helps customers improve fan engagement, increase automation, scale content production, deliver real-time insight, improve operational efficiency, and maintain the trust and accuracy associated with Opta.

A recent example of this in practice was Stats Perform’s work with BBC Sport during UEFA Euro 2024, where Opta-powered live data, AI-driven insights, and automated storytelling tools helped support richer real-time match coverage and fan engagement at tournament scale. By combining trusted Opta data with AI-powered workflows, broadcasters were able to surface deeper context, faster analysis, and more engaging live experiences for audiences throughout the competition.

Opta Live already delivers real-time metrics, visuals, and pushed insights to broadcasters during matches. The next step is conversational interaction through experiences such as Ask Opta AI, where users can query live or historical data and receive trusted, context-aware answers.

A fan watching a match can ask where a team ranks historically, which tournaments produced the most goals from outside the box, or how a red card changes match outcomes. A broadcaster can request an instant tactical breakdown. A team analyst can evaluate whether a player made the right pass, how likely a possession was to create a shot, or where defenders should have been positioned. By combining proprietary sports data, live video understanding, foundation models, and NVIDIA-accelerated AI, Stats Perform is making those answers accessible.

This represents a shift from business intelligence to decision intelligence. Analysts and coaches no longer need to assemble dashboards or manually search for examples. They can explore decisions, simulations, causal questions, and performance scenarios directly.

Looking ahead, the same foundation models will power visual search, predictive simulations, tactical ghosting, automated highlights, second-screen experiences, and professional analysis workflows. The goal is not simply to collect more data, but to turn the global record of sport into an intelligent platform that fans, teams, broadcasters, and leagues can build on, scaling elite insight to audiences everywhere through continued work with NVIDIA.