Google DeepMind is expanding its sports-focused AI work with a system that can model soccer movement and forecast what may happen next on the pitch. The project, called TacticAI, was built to help analyze corner kicks and predict player trajectories using broadcast-style video data.
The company says one of TacticAI’s most notable capabilities is its ability to estimate where players will move up to eight seconds into the future. The model also generates tactical recommendations, including alternate player placements that could improve a team’s chances during set-piece situations.
Unlike general-purpose AI tools, TacticAI was designed specifically for football tactics. It uses geometric deep learning to study player positions and interactions during corner kicks, then produces predictions about likely outcomes such as who will receive the ball or whether a shot will follow.
Google says the system was not evaluated only through automated tests. Researchers worked with football experts at Liverpool FC, who reviewed the AI’s suggestions against real match scenarios. In that qualitative assessment, experts reportedly preferred TacticAI’s recommended setups 90% of the time over the original configurations from matches.
That result suggests the tool may have practical value for coaching staffs looking for a faster way to review tactics, not just a model that performs well on benchmarks.
The published research also found that TacticAI outperformed earlier baseline systems in predicting both the likely recipient of a corner and whether the play would end in a shot. It also produced alternative player arrangements that closely matched realistic professional match situations.
DeepMind has now announced a partnership with Brazilian club Palmeiras, which the company describes as the first football team to meaningfully build on TacticAI. The club will use the system to simulate field scenarios and forecast open-play dynamics up to eight seconds ahead.
The move suggests Google wants to move the technology beyond a research setting and into practical use with a professional team. If the collaboration proves useful, the system could become a more active part of match preparation and in-game analysis.
For now, the focus remains on helping coaches understand how a play might develop before it happens. That kind of insight could be especially useful in moments where small changes in positioning can alter the outcome of a match.
Although TacticAI is aimed at football, the underlying approach may have uses outside sports. Predictive models that track how multiple agents move and interact could potentially support autonomous robotics, traffic systems, logistics planning, and other environments where coordination matters.
That broader potential is part of what makes the project stand out. On the surface, it is a soccer analytics tool. Underneath, it reflects a wider effort to build AI systems that can understand complex motion and anticipate events before they unfold.
DeepMind’s latest announcement shows how that research is beginning to move from controlled experiments into real-world settings. Whether on the training ground or in other fields, the question now is how far these predictive systems can go once they are asked to interpret dynamic situations in real time.