When Algorithms Mislabel Sports: The Sports Writer in the Age of Machines

3 AM in Miami. I opened my email and found an attachment titled: "Stage-2 Deep Professional Analysis — Domain Label: tennis." I opened it. Inside was a nine-section analysis about a Houthi drone attack on the Madinah power station in Saudi Arabia, about Pakistan's condemnation, about defence commitments between Islamabad and Riyadh. Not a single tennis player. Not a single set. Not a single serve, forehand, or break point. Just an algorithm that had labeled a geopolitical article as "tennis," and then another system that tried to analyze it through the tennis framework — and failed spectacularly, forced to write "N/A – insufficient information" eighteen times in the same document. I sat there, cold coffee in hand, and thought about Rai Benjamin. I met that kid on the NCAA track, before the world knew his name. In 2026, in Eugene, I abandoned my assigned story to run down to the mixed zone and ask him forty-five minutes of questions about 400-meter hurdle technique. He ran in lane 8. Nobody noticed. No algorithm labeled him a "future star" that day. That is why I trust human eyes. That is why I distrust automated labeling systems. And that is why the story of an article mislabeled as a sport is not merely a technical story — it is a story about how we understand sports, and about whom we are entrusting that understanding to. Over the past decade, the global sports media industry has witnessed a silent revolution. Major newsrooms like Sports Illustrated, ESPN, and hundreds of regional sports sites have introduced automated classification systems — algorithms trained to read thousands of articles a day, tag topics, and route content to the right editorial desk. The idea sounds reasonable. An article about Wimbledon goes to the tennis desk. An article about Premier League transfers goes to the football desk. An article about doping goes to the investigations desk. Machines read faster than humans, never sleep, never ask for a raise. But machines do not understand sports. Machines understand word probabilities. When an article contains the words "Pakistan," "Saudi Arabia," "Houthi," "drone," "Madinah" — a poorly trained classifier can latch onto a single keyword, or be influenced by an error in training data, or simply encounter a case the model has never seen. The result is a "tennis" label slapped on an article with no tennis player in it. I have followed the development of these systems as an insider. I once worked with data engineers at my former newsroom who tried to build a system that could automatically classify Olympic news by sport. They had a dilemma: swimming, athletics, and gymnastics share a great deal of vocabulary — "competition," "athlete," "medal," "record." An athletics article could easily be labeled "swimming" if the algorithm relied only on keyword frequency. Their solution was to add context. But context, in sports, often lies in details machines cannot read: a name, a place, a moment. When I read that Stage-2 analysis, what stopped me was not the error. It was how the system handled the error. In the first section — "Technical & Tactical Analysis" — the system wrote: "No player, coach, match, or stroke element is identified among the entities." It listed Pakistan, Saudi Arabia, Houthi militia, the Madinah power station, the Prophet's Mosque, Sajjad Haider Khan, Khawaja Asif. And it concluded: "This dimension is not applicable to the supplied article." That was a correct action. But it was also a frightening moment. Because in twenty-two years in this profession, I have never seen an automated system with enough courage to say "I don't know." The systems I have worked with always try to fill the gap. They predict. They infer. They fabricate a plausible-sounding answer rather than admit ignorance. But this system — at least the version I was reading — chose otherwise. It wrote "N/A – insufficient information" in every cell. It wrote "Any attempt to fill this template would be pure fabrication." It admitted that it could not analyze a geopolitical article through the tennis framework without fabricating. And I asked myself: is this progress, or just luck? I have spent years following data systems in sports. I once watched a model predict World Cup results based on social media data, and it got the Croatia-England semifinal of 2026 completely wrong — the match I stayed in Moscow three extra days to interview assistant coaches and write an analysis of Croatia's flexible 4-2-3-1. Data could not predict that. But the human eye could. Because the human eye saw Luka Modric run 12.2 km in a match while maintaining perfect ball-control rhythm. The human eye saw that and understood it meant something. The algorithm only saw a number. Amid endless data, I always look for a human being still breathing. In the case of this mislabeled article, the human being still breathing is some journalist — perhaps a geopolitical reporter in Islamabad, or an editor in Dubai, or a writer anywhere — who wrote an article about a drone attack, and had no idea their work was being fed into a tennis analysis system. That is a violation. Not legal, but professional. Look at what the analysis actually found in the original article. It found a Houthi drone attack on a power station near Madinah. It found one transformer out of service. It found Pakistan's condemnation — issued by Sajjad Haider Khan and Khawaja Asif — calling it "a grave provocation." It found defence commitments between Pakistan and Saudi Arabia. It found concerns about Red Sea shipping disruption. These are important geopolitical events. They belong to an entirely different field of analysis — international politics, regional security, energy infrastructure. But the Stage-2 analysis, bound by the "tennis" label, could do nothing with them. It was forced to write "N/A" across all nine sections. And in doing so, it inadvertently revealed something about the nature of sports. Sports, as I understand it, is a language. It is how humans tell stories about their own limits. An athlete running the 400-meter hurdles in 48.33 seconds is telling a story about what the human body can do. A tennis player saving five match points is telling a story about mental endurance. But sports is not the only language. Politics is also a language. War is also a language. And when you try to translate one language into another by machine, you lose meaning. The article about the drone attack is not a sports article. It does not need to be. It has its own value, in its own field. The problem is not that it was "mislabeled" in the sense that it belongs to a different sport. The problem is that it was mislabeled in the sense that it does not belong to sports at all. And that made me think about how we are defining sports in the digital age. There is one thing that analysis did right, and I want to spend time on it. In most AI systems I have seen, when data is missing, they fabricate. They infer from what is available, producing a plausible-sounding but baseless answer. In sports, this leads to meaningless predictions, distorted rankings, stories conjured from nothing. But this analysis did not do that. It said "insufficient information" twenty-two times. It refused to fill in the blanks. It refused to manufacture a sports story from a geopolitical article. The golden trophy is not at the finish line, but at the turns we never planned for. In this case, the unplanned turn was admitting that some gaps do not need filling. Some questions do not need answers. Some articles do not belong to any sport at all. This runs counter to the instinct of modern sports media. We are trained to find stories in everything. We are taught that every event has a sports angle — a lesson about teamwork, an example of perseverance, a metaphor for victory and defeat. But it isn't so. A drone attack on a power station teaches us nothing about tennis. It teaches us nothing about football. It teaches us nothing about athletics. It teaches us about geopolitics, and we should leave it there. I remember an old colleague at Sports Illustrated who once told me: "The hardest thing in sports journalism is not finding the story. It is knowing when not to write." She said that in 2026, when I was new to the profession. I did not understand. Now I do. So what do we learn from this specific case? First, automated sports labeling systems need regular auditing. An error like this — labeling a geopolitical article as "tennis" — is not an isolated incident. It is a symptom. If the system can make this error once, it can make it thousands of times. Second, analysis systems need the ability to refuse analysis. The ability to say "I don't know" is not a weakness — it is a strength. It prevents the spread of misinformation. Third, humans need to be in the loop. Not to fix errors — though that matters — but to ask questions. A human editor, seeing an article about Houthis and Madinah labeled "tennis," would immediately sense something is wrong. An algorithm would not. I have spent years following matches and analyzing data, and I have learned that data is only valuable when placed in human context. A serve-speed number means nothing if you do not know the player is competing with a shoulder injury. A win-rate statistic means nothing if you do not know the athlete just lost her mother. In this case, the context is: this is not sports. This is war, diplomacy, and energy. And any system that cannot recognize that is failing at a fundamental level. I write this from Miami, but I think of Vietnam. In recent years, Vietnam's sports media industry has grown rapidly. Online sports sites have mushroomed. Social media platforms are flooded with sports content. And more and more newsrooms are seeking to automate their editorial workflows. This makes sense. It helps process a massive volume of information. It helps report faster. It helps expand coverage. But it also carries risks. The risk of mislabeling. The risk of misanalysis. The risk of manufacturing stories that are not true. I once saw a Vietnamese sports site report on a "Vietnamese tennis player" who was actually a table tennis athlete. I once saw an article about a "national tennis championship" that was actually a badminton tournament. These errors seem small, but they accumulate. They erode reader trust. And in a market where readers increasingly have choices, trust is everything. When the stands are empty, the most honest voice comes from an old phone. During the 2026 pandemic, when every stadium was closed, I learned that sports news does not need stadiums. It needs truth. It needs people. It needs a writer willing to call a track coach in Kenya and spend two hours listening to footsteps on rain-soaked earth. No algorithm can do that. If I were running a sports newsroom today, I would do three things. First, I would establish a topic-check gate. Any article automatically labeled "sports" by the system but containing no sports entity (athlete name, tournament name, team name) would be flagged for manual review. This would prevent cases like the Houthi article being labeled "tennis." Second, I would train my team on the limitations of AI. Not to fear it, but to understand it. They need to know that a language model can produce fluent text on any topic — including topics it knows nothing about. They need to know that AI's confidence does not correlate with its accuracy. Third, I would invest in people. Not to replace AI, but to complement it. Reporters who can see what the algorithm misses. Editors who can ask the questions machines do not know to ask. I have worked in this industry for twenty-two years. I have witnessed the transition from typewriters to computers, from print to digital, from local newsrooms to global networks. I have learned that technology changes how we work, but it does not change why we work. And the reason we work — the reason I still sit before a screen at 3 AM — is to tell the stories of people. Of those who run, jump, swim, compete. Of those who win and lose. Of those who try. No algorithm can tell that story. But a well-designed algorithm can help us find it. When I closed that analysis file and looked out the window, the Miami sky was shifting from black to gray. A new day was beginning. I thought about the original article — the one about the drone attack. Somewhere in the world, a journalist wrote it. They took time to verify details, to contact sources, to ensure they were telling the story accurately. They deserve to be read by people who understand their subject. Instead, their work was fed into a tennis analysis system, where it was deemed "insufficient information" across nine different categories. It is a small injustice, but it is part of a larger problem. When we automate the classification and analysis of information, we risk losing understanding. We risk creating a world where everything is labeled, but nothing is understood. The stadium is silent, but I hear the heartbeat of an entire generation. That heartbeat — the heartbeat of those who write sports, those who read sports, those who live for sports — is still beating. It beats in every carefully written article, every honestly conducted analysis, every empathetically told story. And it will keep beating, as long as we remember that sports, in the end, is about people. Not data. Not algorithms. Not labels. Just people.