Revolutionize Basketball With AI

The system was presented at AWS Summit Japan 2026, a two-day event held at Makuhari Messe in June that brought together some of the latest developments in artificial intelligence. Organized by Amazon Web Services, the Amazon group company that provides cloud computing and information technology infrastructure, the event is one of Japan's largest gatherings for business leaders and engineers interested in emerging technology.

Among the demonstrations was an AI-powered analyst support system developed by B.League club San-en NeoPhoenix in cooperation with AWS.

San-en NeoPhoenix, founded in 1965 and operated by Phoenix Co., has introduced the technology to help its analysts process the rapidly expanding volume of information generated during games and across an entire season.

Team analysts are responsible for studying opponents, evaluating their own players' performances and preparing reports for coaches. Producing those reports requires them to identify the factors influencing victory or defeat from close to 30,000 data points in a single game and translate the findings into tactics for the next matchup.

As more seasons are played, the accumulated data becomes increasingly vast and complex, making it difficult for analysts to examine every possible pattern manually.

The new system first feeds information such as individual playing tendencies and successful shots into a machine-learning program. It then extracts the factors judged most important to the outcome of a game.

Generative AI developed with AWS analyzes those findings in greater depth, converts the machine-learning results into understandable language and produces a strategy report for the team's human analyst. By allowing AI to organize the initial information, the analyst can devote more time to reviewing video, testing hypotheses and preparing for the next game.

The system can also be adjusted to reflect the analyst's own thinking and the club's tactical philosophy. Prompts are used to guide the AI toward the types of information the analyst or team considers significant, effectively teaching it which data deserves particular attention.

Kimura, an analyst with San-en NeoPhoenix, said the technology has helped reveal perspectives he had previously overlooked.

An analyst may decide subjectively that a particular opposing player or type of play represents the main threat. That judgment may prove correct, but another player could instead emerge as the decisive factor. The AI acts as a second analyst by suggesting alternative players, movements or patterns that may also warrant attention.

Kimura can then return to the data and video, determine whether the AI's suggestion is meaningful and pass the information to the coaching staff. He described the system as an adviser capable of examining the full range of available information more objectively.

The benefits became more visible during the second half of the season. Kimura said the system initially struggled because AI performs poorly when there is insufficient data, but its analysis improved as more games were added and the team's winning percentage increased later in the season.

The technology has also forced analysts to reconsider when to trust an algorithm.

Kimura recalled receiving an AI finding that he believed was wrong and therefore did not share with the coach. In the actual game, the player and play identified by the system became decisive.

The experience persuaded him to give greater consideration to the AI's conclusions. However, he said there had also been occasions when the system emphasized a factor that ultimately proved less important than expected.

The challenge, he said, is neither to reject the technology nor to accept its recommendations automatically, but to improve the human judgment used to interpret them.

That judgment remains essential because information has no value unless coaches and players believe it and can put it into practice. San-en NeoPhoenix's head coach has an established tactical philosophy, and recommendations are discussed extensively before the team decides how to proceed, regardless of whether the original idea came from a person or the AI.

The use of artificial intelligence could also create a new tactical contest between teams. A club might deliberately use different strategies in several less important games to mislead an opponent's analysis before introducing its true approach in a crucial matchup.

Human analysts can generally examine only a limited number of recent games in detail. An AI system operating in the cloud, however, can continuously process 50, 100 or more games and may eventually detect that an opponent has temporarily departed from its usual style.

If more teams adopt similar systems, professional sports could develop into a contest of AI-supported tactics, with human coaches and analysts supplying the creativity and individuality that distinguish one club from another.

Paradoxically, teams that rely little on data may become especially difficult for AI-oriented opponents to predict. A club that changes its approach from game to game or allows players to act largely on instinct may produce fewer repeatable patterns, reducing the reliability of algorithmic forecasts.

The same applies to star players. Elite athletes often become exceptional precisely because they depart from expected patterns. Analysts must therefore determine whether a player is ignoring the team's structure because of poor decision-making or because the player possesses the ability to succeed outside conventional tactics.

Data may also be used to identify players who should be granted greater freedom. A team could instruct four players to follow an AI-informed plan while allowing one unusually creative player to operate independently, making its movements harder for an opponent to model.

Despite the growing influence of technology, Kimura said complete dependence on AI was unlikely because much of the appeal of sport comes from events that cannot be predicted. Human intervention can improve the quality of forecasts and decide which information truly matters.

San-en NeoPhoenix is considering further uses for the system, including broader data management across the B.League and the prevention of player injuries.

The club also hopes to teach the AI about the habits and philosophies of opposing coaches and analysts.

The B.League draws head coaches from Australia, the United States, Europe and other parts of the world, creating wide differences in playing style. Feeding information about a coach's background and philosophy into the system could improve its ability to interpret data and predict tactics.

A coach influenced by American basketball may rely heavily on data, while an Australian coach may favor a more physical approach. Incorporating such tendencies could allow the AI to produce analysis better suited to each opponent.

The basketball project forms part of a wider effort to apply AI to Japan's economic and social challenges.

The AWS Japan president said physical AI, including robots and other intelligent systems operating in the real world, represents a major opportunity for Japan.

Previous advances in internet and smartphone technology largely remained inside virtual software environments. Physical AI can move objects, deliver goods and assist people directly, making it particularly important for a country facing an aging population and severe labor shortages.

Japan also has globally competitive companies in robotics, factory automation, sensors, actuators and motors. Those strengths could become more valuable as software companies seek to connect virtual intelligence with machinery in the physical world.

The expansion of physical AI could therefore help solve domestic labor problems while creating new international opportunities for Japanese manufacturing.

Another critical advantage is data generated by Japan's highly developed workplaces and operating systems.

Factories produce goods to exacting standards, trains operate with minimal delays, and convenience stores maintain full shelves through frequent deliveries. These operations depend on exceptionally high-quality processes, but much of the knowledge behind them remains tacit rather than digitally recorded.

AI performance depends not only on the quality of the model but also on the information used to train and optimize it. A system cannot reproduce Japan's operating standards unless the skills and decisions behind those standards are converted into usable data.

In countries where trains arrive only every 10 minutes, for example, the available information may be insufficient to optimize a system designed for the precision of Japanese rail operations. Japan's detailed and reliable real-world processes could become a major asset if they are successfully digitized.

The greatest challenge is capturing the tacit knowledge of experienced craftspeople before an aging workforce retires.

One example can already be seen in the restaurant industry. Many restaurants can serve consistently high-quality meals without employing a specialist chef at every location because experts have analyzed and standardized the original chef's techniques. Central kitchens reproduce sauces and other components, while part-time workers complete and serve the dishes according to carefully designed procedures.

A similar method could be applied to manufacturing and other fields. By identifying why highly skilled workers produce superior results, companies could convert their knowledge into repeatable systems and eventually export solutions based on Japan's operating expertise.

The AWS Japan president said Japan must move quickly because many of the craftspeople holding this knowledge are growing older. The country's weakness has often been the slow pace at which it turns existing strengths into scalable products and services.

Rapid advances in semiconductors will also play a central role in the expansion of AI.

The two main measures of semiconductor performance are processing power and the electricity required to achieve it. As chips become more capable, they can generate greater output using the same amount of power, sharply improving efficiency.

AWS both procures chips through partnerships and invests in developing its own semiconductors. Different tasks require different levels of performance, much as some journeys require a Formula One car while others are better served by a vehicle capable of transporting many people efficiently.

Providing a range of semiconductor options allows companies to balance computing performance and cost as AI use expands.

Demand is expected to rise further as users move from generative AI toward AI agents.

Generative AI responds to questions, summarizes information and produces text or images. AI agents are designed to pursue a stated objective and carry out complex tasks autonomously.

A conventional system may suggest several travel plans. An AI agent could instead receive a budget, the number of travelers and other conditions, present suitable options and, after receiving approval, book flights and hotels and complete payment.

Future systems may also learn from a person's previous decisions and understand the context behind requests, enabling them to act more like a human assistant familiar with the user's preferences.

Services once available only through personal concierges could therefore become accessible to a much broader public. As these agents handle increasingly complicated tasks, the volume of data and computing power required will continue to grow.

AWS said organizations must retain multiple technological options while building AI in secure environments with effective governance and advanced cybersecurity.

Japan still has significant opportunities in the AI era, particularly if it can combine its expertise in manufacturing, robotics and high-quality operations with cloud computing, semiconductors and artificial intelligence. The country now faces a race to convert the knowledge embedded in its workplaces into digital systems before that expertise is lost.

Source: テレ東BIZ