Not the whole match

The final Grand Slam tournament of the year is currently underway in New York. Over the past few years, we’ve repeatedly observed how new technologies are transforming the sport of tennis. At Wimbledon 2023, the focus was on the line judges, who were gradually being replaced by the electronic eye. At Roland Garros in 2025, the focus was on the question of whom to believe when humans and machines reach different conclusions. Now it’s the US Open—and the story takes another twist.

There’s a simple reason why we keep taking a closer look at tennis, of all things: we enjoy watching it ourselves. At the same time, this sport offers a particularly vivid illustration of the potential of AI—even beyond the generation of text and images. On a clearly defined court with fixed rules, it becomes apparent how technical systems are gradually taking on new tasks.

At the U.S. Open, it’s no longer just about whether a ball was in or out. Technology is playing an increasingly significant role in what spectators actually perceive during a match, which moments seem significant to them, and what stories are told.

This makes all the difference, especially in the early days of a Grand Slam tournament. Matches are being played simultaneously on numerous courts. At Arthur Ashe Stadium, experts comment on every serve, every tactical shift, and every visible reaction from the players. At the same time, the more interesting match might be taking place on an outdoor court, yet in the end, hardly more than a single line in the scoreboard remains of it. That doesn’t make the match any less significant. It’s just that no editorial team can observe, analyze, and report on every match equally.

It is precisely this gap that has been a concern for IBM and the United States Tennis Association for several years now. Kirsten Corio, who was head of the USTA at the time, described it to VentureBeat in 2023 as follows: While many matches had statistics and results, they lacked commentary. Their stories remained “untold”—unshared.

From Commentary to Interpretation

As early as 2023, the U.S. Open introduced automated commentary. Based on match data, the system generated audio commentary and subtitles for the highlight clips of the singles matches. At that time, 56 data points were typically recorded per point played, and each clip generated from this data then underwent human quality control.

In this way, even matches on the outer courts received a commentary-filled recap, for which there had previously been no editorial capacity. While they didn’t receive the same attention as a match in the largest stadium, they at least gained more visibility than a simple list of scores. IBM described the technical foundation; VentureBeat documented the statement about the “untold stories” and the human review.

The announcement for 2026 continues this trend. A new live overview allows users to select their favorite players and directs them more quickly to the corresponding matches, highlights, and stories. The enhanced Match Chat answers questions using current and historical data and can also include images and videos.

This new dimension is particularly evident in the ” Serve Quality” metric. For all 254 singles matches, 21 body and racket positions are tracked fifty times per second. This generates approximately 1.2 billion data points during the tournament. The system breaks down the serve into individual movement phases and evaluates, among other things, efficiency, precision, consistency, and ball toss.

Key Moments marks the next step. The existing probability of winning is supplemented by automatically generated indicators of turning points and shifts in momentum. The system thus no longer merely indicates who is likely to win a match; it also provides an interpretation of when and how the match changed course.

When Systems Organize Attention

That is where the truly interesting development lies. Serve Quality doesn’t just measure more accurately. The metric translates the quality of a serve into selected, measurable variables, thereby providing a specific idea of what constitutes a “good” serve. Key Moments, too, doesn’t simply present turning points in a neutral way. Based on its criteria, the system determines which changes during the course of a match appear to be significant. The personalized live overview, in turn, prioritizes which match and which moment users see first.

More data does not, therefore, automatically result in a more complete picture. What becomes apparent at first is what the data fields, metrics, and models are designed to capture. Anything that cannot be captured or is not given weight in the system tends to fade into the background.

A long rally, an ace, or a saved break point can be counted. What’s harder to capture is an inconspicuous tactical shift that doesn’t take effect until five games later. The same applies to a player’s growing uncertainty, a deliberate slowing of the pace, or a point that looks ordinary statistically but changes the match psychologically.

Until now, it has primarily been up to editors and commentators to shape such observations into a story. They, too, selected certain elements, interpreted them, and ignored others. Now, parts of this judgment are being shifted to data models and technical processes. So the selection process isn’t disappearing, but the criteria by which it is carried out are becoming less visible.

Even human quality control at the end of the process does little to change this at first. It can verify whether a generated post is plausible and suitable for publication. However, it does not answer the underlying question of why the system highlighted this particular moment and overlooked another.

The external locations of each organization

Organizations, too, have their “front lines.” That’s where difficult conversations with customers take place, where unusual decisions are made in support, where projects are saved at the last minute, or where solutions are improvised that aren’t found in any process manual. Such situations often go undocumented. That doesn’t mean they’re unimportant; it’s just that there isn’t enough time to record them and categorize them later.

AI can actually add value in these areas. It can analyze conversations, connect events, and generate coherent documentation from scattered clues. This creates knowledge where previously there were only isolated memories, files, or metrics.

With this new ability to tell stories, however, the same question arises as with tennis: Based on what criteria does the system determine which deviation is noteworthy, which decision is typical, and which nearly failed project is worth telling a story about?

Those who do not determine these criteria for themselves—or at least cannot understand and evaluate them—gain narrative capacity, but relinquish some of the power of interpretation.

Over the past few years, a shift has thus become apparent. In 2023, the question was: What human jobs will disappear when technical systems take over their tasks? In 2025, the focus was on whom we should believe when humans and machines reach different conclusions. In 2026, a third question is added: What will actually become visible when technical systems not only measure, but also select, weigh, and narrate?

People never tell the whole story either. Whether a machine could do that is of lesser importance. What matters is who decides which parts become part of the story—and who notices what’s missing from it.

P.S.: In Germany, the U.S. Open can be watched on Sky and WOW, as well as on SPORTEUROPE.TV without a subscription. The AI features described here can be accessed directly on USOpen.org and in the official U.S. Open app.

From Commentary to Interpretation

Neue Beiträge direkt im Postfach.

When Systems Organize Attention

The external locations of each organization

External Sources

Sharing is caring