Google developed an AI capable of predicting football plays before they occur.

      DeepMind's latest research forecasts player movement for up to eight seconds ahead.

      Football managers dedicate extensive hours to scrutinizing corners, free kicks, and player positions in pursuit of small competitive edges. Google DeepMind is convinced that artificial intelligence can accelerate this process substantially, and its newest initiative, TacticAI, has been developed for this purpose. TacticAI is a football-focused AI assistant that can model player movement, anticipate future play dynamics, and even suggest tactical modifications for corner kicks. One of its remarkable features is its ability to predict player movements up to eight seconds into the future based solely on broadcast-type visual data.

      TacticAI was created in collaboration with Liverpool FC and validated by football specialists.

      In contrast to general AI models, TacticAI is specifically tailored for football tactics. Utilizing geometric deep learning, the system examines player positions and interactions during corner kicks and then generates forecasts about possible future events, recommending alternative player arrangements that could enhance the outcomes.

      The model's tactical adjustments were illustrated in TacticAI's findings.

      Crucially, the model was not limited to laboratory testing. Google asserts that its effectiveness was assessed through a qualitative study involving football experts from Liverpool FC, who compared the AI’s suggestions against actual match situations. According to the published findings, experts favored TacticAI’s proposed tactical setups 90 percent of the time over the initial match configurations, demonstrating the system’s practical effectiveness beyond mere statistical achievements.

      The benchmarking results are equally remarkable. TacticAI surpassed existing baseline models in predicting both the probable receiver of a corner kick and whether a shot would follow, while also producing realistic alternative player formations that closely mirrored authentic professional match scenarios.

      This could extend well beyond football.

      The research is already advancing beyond the laboratory setting. Google DeepMind has announced a partnership with Brazilian football club Palmeiras, marking it as the first team to significantly utilize TacticAI to simulate on-field scenarios and forecast open-play dynamics up to eight seconds in advance. If this proves successful, it could signal the onset of AI functioning as a true tactical assistant on the sidelines rather than merely serving as another analytics tool in the background.

      Our collaboration with @Palmeiras establishes the first football club to meaningfully integrate TacticAI: our AI system that assists in simulating field scenarios and predicting open play dynamics up to eight seconds in advance. ⚽ pic.twitter.com/M3Krejk9Er— Google DeepMind (@GoogleDeepMind) June 11, 2026

      Moreover, the foundational technology has potential applications far beyond the realm of sports. Similar predictive models could someday aid autonomous robots, traffic systems, logistics planning, or any settings where comprehending and forecasting coordinated movement is essential. Perhaps the most intriguing aspect of TacticAI is that, while it appears to be designed to assist coaches in winning football matches, it quietly may be laying the framework for machines that can understand and anticipate complex real-world interactions before they occur.