When 'Martial Arts' Is Compressed Into a Single Label: A Data Archaeology Between the Ring and the Newsroom

One August afternoon in Nakano, I reopened the archive of a Japanese broadcaster I had worked with as a documentary screenwriter. Inside that internal database sat 41,208 records carrying a single label: martial_arts. Filed side by side in the same folder were a men's flyweight boxing final, an 82-second women's kata routine, a three-round wrestling bout, a wushu taolu exhibition by a national team, and a sumo match that ended in four seconds. Forty-one thousand records, one label, and not a single column that could distinguish a combat discipline where points appear after every exchange from a performance discipline scored by a panel of judges seated at the four corners of the mat. The number matters less than the fact that it exists. Every sports story begins with a forgotten figure, and this time the forgotten figure is not on any scorecard. It lives in the way the sports media industry classifies martial arts. I spent two weeks going back through how major data platforms tag combat sports. The result was the same in most places. The label "martial arts" gets used as a container for anything involving hitting, kicking, wrestling, grappling, technical scoring, or simply wearing a uniform. When people ask me, as someone who makes martial arts documentaries for the Japanese market, one question comes up more often than I would like: why do sports analysts keep talking about martial arts as if it were a single sport? When everyone looks at the victory, I look for where they hide the weakness. This time the weakness is not in a fighter. It is in the classification system. In July 2026, karate entered the Olympic programme for the first time in Tokyo with eight medal events: men's kata, women's kata, and kumite across three weight classes for each gender. That was the moment audiences in Vietnam and Japan watched a paradox unfold on live television. A kata routine and a kumite bout were broadcast inside the same window, under the same logo, counted on the same medal table, yet they shared not a single rule about how a winner is determined. In kata, an athlete steps out alone, with no opponent, and scores come from five judges evaluating technique and expression. In kumite, two people face each other, and points come from referees recognising a strike that lands on target, with sufficient power and control, within a window shorter than a breath. Three years later, at Paris 2026, karate was gone from the programme. Kata left alongside kumite. Wushu has never appeared at an Olympics, even though taolu and sanda persist at Asian Games and World Games under two entirely separate rule sets. Muay Thai and kickboxing continue drifting between official and invitational lists at regional multi-sport games. Boxing, the combat sport with the clearest scoring system, once faced exclusion from the Los Angeles 2028 programme before being restored by an International Olympic Committee vote in March 2026. While federations wrestle over membership status, inside newsroom computers everything still sits in the same folder. That is the root of the problem. Classification is not an administrative formality. It determines how we measure, how we compare, how we fund, and how we tell stories. When two fundamentally different things get folded into one label, every metric drawn from that label loses its value. I once received an internal report in which the average age of "martial arts athletes" had been calculated across boxing, karate, and wushu together. That figure is meaningless at the analytical level, yet it was still used to decide broadcast allocation. What is striking is that the media understands the distinction perfectly well. No sports editor calls a kata routine a "fight". No journalist describes a taolu performance using the language of the ring. But once the data enters the archive system, the distinction vanishes. The problem sits in the gap between the humans who understand and the machines that store, and that gap has never been patched. I trust data, but I write about what data cannot measure. A wrong label is a form of data that cannot measure anything at all, because it does not describe reality; it describes the laziness of whoever created the label. Two scoring universes that cannot speak to each other To see the scale of the problem, place two representative scoring systems side by side. The first is direct scoring. Boxing, muay Thai, kickboxing, taekwondo, wrestling, judo, and most modern MMA promotions operate on this principle: the outcome is determined in real time by referees or by points accumulated across rounds. Amateur boxing uses the ten-point must system, where each round has a winner and the loser receives nine points or fewer. Taekwondo uses an electronic system that registers impact force on the body protector. Judo ends a bout with an ippon, or through accumulated shido and waza-ari when time expires. Each sport carries its own definition of a "point", tied to a measuring device, to referee positioning, to the speed of a strike. The second is performance judging. Kata, taolu, and most wushu forms operate on an entirely different principle: no direct opponent, no reaction to read, no real time. The athlete performs a fixed sequence of movements, and a panel awards scores based on technical accuracy, balance, power, rhythm, and in some rule sets, expression. Points are deducted for loss of balance, for insufficient range of motion, for uneven tempo. The crux is this: in the first system, the opponent is the variable. In the second, the opponent does not exist. Every analytical metric built for the first system assumes an opponent: strike accuracy, counter-strike frequency, ability to read an opponent's rhythm, tendency to retreat into a corner when pressured. Applying those metrics to a kata routine is methodologically meaningless. And applying expression criteria to a boxing match insults the sport itself. Yet in most datasets I have seen, these two systems sit in the same column. The first consequence is false comparison. Annual content reports from several platforms regularly produce statements like "the knockout rate in martial arts fell this year". But a knockout rate only exists in combat disciplines. If the sample includes karate kata, wushu taolu, and traditional forms exhibitions, that rate gets diluted by content that has no concept of a knockout. The number falls, but not because fighters are safer. It falls because the sample is mixed. The second consequence is false tactical analysis. MMA research over the past decade has been transformed by motion-tracking data. Metrics such as takedown accuracy, control time, and effective strike density have become a shared language. But if an MMA fighter's data gets pooled with a karate kumite athlete's data, an analyst will never see the real pattern. Kumite requires landing and separating immediately, so strike density is high and contact time is near zero. MMA permits grappling and control, so contact time is long. Pool them, and you get an average that exists in no actual bout. The third consequence is misallocated resources. This is the least discussed and most expensive part. A federation seeking sponsorship usually has to present data on its sport's popularity. If that data comes from a pooled label, smaller federations always lose. Vovinam, pencak silat, and Vietnamese traditional martial arts look smaller than they are when filed alongside MMA. Filed separately, they are training systems with far larger participation than many professional leagues. I once attended a meeting in Tokyo on content strategy for a regional sports channel. Twelve people were in the room. When I proposed splitting the label, the first response was: "If we split it, the dataset becomes too small to sell." That was the most honest answer I have ever heard about why misclassification persists. It persists because it is convenient. Vietnam and the place where the label breaks fastest Nowhere exposes the problem more clearly than Vietnam, and the reason lies in the structure of regional competition. A recent SEA Games edition can include boxing, muay Thai, kickboxing, taekwondo, judo, karate, pencak silat, vovinam, and sometimes wushu, all at once. This is a list any analyst must confront, because within a single national delegation and a single medal system, at least five different scoring systems coexist. Pencak silat has its own rule set with two distinct categories: combat and performance, the latter usually called seni. Vovinam has its own system of kicks and throws, scored in ways entirely unlike Western combat sports. Muay Thai at the SEA Games scores by counting clean strikes to the body and legs, plus elbows and knees, techniques banned in many other sports. Boxing uses the ten-point system. Taekwondo uses electronics. Judo uses ippon. Under the same word, "martial arts", we have five different definitions of what winning means. At the coaching level, this difference is not academic. An athlete moving from boxing to kickboxing must completely change how they distribute their breathing rhythm, because a different scoring system changes striking cadence. A fighter moving from karate kumite to muay Thai must relearn reflexes: in kumite, once you land, you separate; in muay Thai, once you land, you stay in range. At the sports medicine level, the difference is even larger. I once tracked injury records for a group of athletes across two years. The group labelled "martial arts" showed an unusually high rate of knee joint injury. Once the label was split, the cause surfaced quickly: most injury cases came from disciplines involving a rotational movement on one leg and landing on the knee, characteristic of certain kata sequences and throwing arts. Upright combat disciplines contributed little to that figure. The pooled label had concealed the injury pattern for two years. When everyone looks at the victory, I look for where they hide the weakness. Here, the weakness was hidden in an injury table. One thing is easily overlooked: Vietnam is one of the few countries where indigenous martial arts have their own competition systems, their own federations, their own lineages, and do not depend on being named inside an international media label. That creates an analytical advantage few places possess. If an analyst takes the trouble to split the label, Vietnamese data could become a richer research sample than data from markets dominated by a single combat sport. That advantage remains unused, because people are still counting everything together. One thing I learned working in Japan is how tightly organisations here classify. A judo association, a karate association, a kickboxing organisation each have their own grading systems and record-keeping, and in internal documents they never mix. When making documentaries, I had to learn to respect that distinction, not out of politeness, but because mixing them would make the film technically wrong. But step outside, into mass media, and everything compresses back into one word. The forgotten numbers on the mat If I had to choose one place to begin splitting the label, I would begin with the forgotten metrics. In combat sports with rounds, there is one metric that almost never makes it to air: third-round win rate. This figure speaks to stamina management and the ability to read a fight while exhausted. A fighter who wins round one and loses round three is an entirely different story from one who loses round one and wins round three. A final scorecard cannot distinguish the two. A third-round win rate can. The second metric is tempo variation across rounds. In MMA and boxing, some fighters throw at nearly identical density throughout, while others spike their output in the final round. The second group usually has a better fitness base, but also loses more often on the scorecards if they drop the first two rounds. How they distribute tempo is a tactical decision, and it leaves traces in the data. The third metric is reaction latency after injury. This is difficult to measure and rarely recorded. When a fighter returns from injury, response time to an opponent's strike changes, sometimes by mere hundredths of a second. To the naked eye, it is invisible. In motion-tracking data, it is glaring. I once reviewed such a dataset for a documentary project. A fighter returning from a shoulder injury showed, across his first three bouts, a response time to the left hook that was slower than his own baseline. He won all three fights. No one mentioned the detail. The fourth metric is movement habits in an empty arena. This is the part I care about most, and it emerged from a period I would not have chosen: the stretch when global competition paused because of the pandemic. With no spectators, I interviewed a group of former athletes about competing in silence. What they described was not emotional but behavioural. Without crowd noise, many lost a spatial reference signal. They moved less into the corners of the mat, and they threw strikes one beat earlier than usual. When the stadium is empty, the person inside finally speaks. And when the person inside speaks, the data begins to mean something. The fifth metric is the rest interval between rounds. In boxing and muay Thai, the interval lasts one minute. How a fighter uses that minute reveals much about conditioning and about the quality of the corner. Some sit, some stand, some drink immediately, some wait thirty seconds before drinking. These patterns repeat across bouts, and they can be recorded. Almost nobody records them. These metrics share one thing. They all belong to combat disciplines, and they all become meaningless if pooled with performance disciplines. That is why splitting the label is not logistics. It is the precondition for any analysis to have value. The paradox of rule unification In many discussions about the future of martial arts, one proposal keeps surfacing: unify the rules to make comparison easier, commercialisation simpler, television friendlier. It sounds reasonable. But it rests on a false premise, that the diversity of rules is an obstacle to understanding a martial art. Rule diversity is not an obstacle. It is the content. A martial art is defined by what it forbids and what it rewards. Judo bans certain leg throws for safety reasons and because the art's technical objective does not lie there. Muay Thai rewards elbows because elbows are part of its technical system. Boxing has no kicks. Taekwondo bans punches to the face in many competition rule sets. Standardise all of it, and you do not get a better martial art. You get a new martial art, and you erase the value of all the others. Breaking convention does not require a loud voice; it requires evidence heavy enough. For me, the heaviest evidence is not a philosophical argument about martial identity. It is injury data. When disciplines with different rules are forced into a single rule set, injury patterns shift, usually for the worse, because an athlete's body is trained for a specific reflex system. In Japan, I once watched a club debate whether to add grappling to its training programme. Supporters said it made fighters more complete. Opponents said it diluted the signature technique. The debate never concluded, and that is a good thing. A martial arts community that never finishes arguing about its own nature is a community still alive. The media's problem is different. Media needs simplicity to operate, and that simplicity is usually bought by erasing differences that do not matter for selling advertising. The label "martial arts" is a product of that need, not of any effort to understand. I do not believe pooled labelling is anyone's conspiracy. I believe it results from a chain of small decisions, each reasonable in its moment. The person building the database needed a field to fill. The product person needed a category to display. The salesperson needed a number large enough to present. Fifteen years later, that chain has produced a system nobody designed on purpose, and now it runs itself. And when a system runs itself, it starts generating its own reality. What would change if the label were split I ran a small exercise in a recent project. I took a dataset of roughly three thousand records on Asian martial arts and split it into four groups by scoring system: direct-contact combat, grappling and control combat, sensor-scored combat, and panel-judged performance. After the split, several things surfaced that had been hidden. The grappling and control group had a markedly higher share of bouts lasting the full distance. This surprises no one inside the sport, but it never appeared in the pooled dataset. The performance group had a faster return-to-competition rate after injury, but a shorter peak competitive lifespan. That is a trade-off performance analysts rarely discuss, because the two groups had never been placed side by side properly. The sensor-scored combat group showed the highest variance in results between outings. This relates to how sensors register force, and it is a technical subject deserving its own study, not a line in a summary report. None of these conclusions is a breakthrough for practitioners. But they are conclusions that cannot be drawn from a pooled dataset. That is the entire point. Fans remember the scoreline; I remember the defender's expression in the 89th minute. In martial arts, I remember the loser's expression when the referee raises a hand. There are moments audiences do not see, and pooled data does not see them either, because it has no column in which to record them. Behind every mislabelled record is a decision that was skipped. And behind every skipped decision is a story that has never been told. What I still do not know I do not have enough data to claim that pooled labelling has caused a specific financial loss for a specific federation. I only have observations from meetings, from internal datasets I have been shown, and from my own work. Nor do I have enough data to claim that splitting the label will improve everything. It is possible it will simply create a new bureaucratic layer, with people whose job is to reclassify from scratch. What I know for certain is this: every analysis I have done in twelve years, whether on a boxing match, a judo session, or a kata shoot, began by establishing which scoring system was in operation. When I skipped that step, my conclusions were wrong. Classification is not library work. It is the first step of any serious analysis. In internal documents I have seen about sports classification, one line stays with me. An editor wrote that pooling labels did not matter, because "audiences cannot tell the difference". That may be true for a viewer watching a short clip. It is not true for athletes, coaches, sports physicians, or anyone who has to make a decision based on data. And when an industry builds its data systems on the assumption that end users do not need to distinguish, that industry is shrinking its own capacity to understand. What I am waiting for I am waiting for the first dataset built not around sport names, but around scoring systems. A dataset where the first question is not "which sport" but "how do you win". When that becomes the default, many old arguments will dissolve on their own. Debates about which martial art is harder, which is more practical, which deserves television time, mostly stem from comparing things that operate under different rule sets. Split them apart, and the answers become so clear there is nothing left to argue about. I trust data, but I write about what data cannot measure. And what data cannot measure, in this case, is that it has been measuring the wrong thing.