Three Scoring Systems Under One Label: The Martial Arts Classification Error in Combat Sports Data

A document moved through an entire content-analysis pipeline and stopped at a single line: martial_arts. No title. No source. No athlete name. No publication date. The only surviving token was typed with an underscore, deviating from the slash format the classification system requires.

For anyone working with data, that was the most notable fact of the day. An empty classification field forces every table behind it to be filled by inference. In sport, inference rarely appears in the form of a question. It usually appears in the form of a conclusion that has already been closed.

I pulled two older documents from my personal archive. One was a professional boxing scorecard. One was a taolu scoresheet from a continental wushu championship. The two sheets differ in everything, from the number of judges to the unit of measurement. In the same classification field of the software, both were recorded as “martial arts”.

The arena did nothing wrong. The dataset did.

Context: a broad label read as a narrow instrument

Classification looks like administrative paperwork. It decides which toolset is permitted, and the toolset decides which conclusions may be drawn. Change one label, and the entire chain of reasoning behind it changes with it.

Over roughly fifteen years, the volume of combat-sports content has grown faster than the pace of data-standard building. Content platforms bundle boxing, MMA, kickboxing, Muay Thai, wrestling, jiu-jitsu, sanda and taolu into a single category because it makes advertising easier to sell. Ranking sites, statistics portals and part of the sports database industry followed. Three entirely different scoring systems now sit under one name.

The first group is modern professional combat sport: MMA, boxing, kickboxing, Muay Thai, grappling. Its basic unit of measurement is the win-loss result and the method of finish.

The second group is traditional and performance martial arts. Taolu is scored on movement difficulty, execution quality and the overall quality of the routine. The concept of a knockout does not exist in that system.

The third group is sanda, a hybrid discipline permitting punches, kicks and throws, contested over three rounds under its own ruleset, and not interchangeable with Muay Thai or kickboxing.

These three groups share no common definition of victory. That is why a technical analysis can be methodologically sound and completely wrong about its subject, purely because a classification field was read carelessly.

The consequences do not stop at academic debate. They reach contracts, purses, entry slots and the way audiences price an athlete.

A taolu scoresheet has never heard of a knockout

A taolu scoresheet has never heard of a knockout. That is why I trust it in its own place — and why I never place it beside a boxing scorecard.

A professional boxing scorecard runs on the ten-point must system. Three judges score independently, the winner of a round receives ten points, the loser nine or fewer. A twelve-round fight produces three separate number sequences, and the final result is the sum of those sequences. Reading a boxing match properly means reading all three sequences, not the aggregate.

A taolu scoresheet operates on a completely different logic. There is no direct opponent. There are no rounds. There is no loser. A panel of judges is divided into groups scoring movement difficulty, technical quality and overall quality of the performance. The athlete competes against their own routine and against the scale, not against the person standing next to them.

Sanda sits in the middle and belongs to neither side. Three rounds, shorter than professional boxing, with points awarded for effective strikes, successful throws and knockdowns. A clean throw in sanda carries a value it simply does not have on a boxing scorecard, and conversely, the count of clean strikes does not carry the same weight.

Grappling is a fourth system entirely. Events such as ADCC or the IBJJF framework score specific control actions: takedowns, guard passes, positional dominance, back control. Each action carries its own value, and winning on points is a fundamentally different outcome from winning by submission.

Even inside the professional combat group, the data is already fragmented. Professional boxing exists with four major title organisations: WBA, WBC, IBF and WBO. A data row reading “world champion” without the organisation name is a row that cannot be verified. MMA largely operates under the unified rules adopted by state athletic commissions, and metrics such as significant strikes landed or absorbed per minute only mean something inside that statistical framework. Carrying such a metric into another discipline manufactures a number that does not exist.

Based on my experience tracking fights as an investigative reporter, the most serious error does not come from assigning the wrong label. It comes from reading a broad label as though it were a narrow instrument.

A correct classification label can still lead to a wrong conclusion. The label “martial arts” is not linguistically wrong. It is simply so broad that it becomes useless the moment it is placed in a column requiring a specific unit of measurement.

I once reconstructed a data chain in which a taolu athlete appeared in a statistics table with a “finish rate” column holding a value of zero. That number was technically accurate as stored data and meaningless as sport. But when the ranking algorithm ran over it, it read that zero as an athlete who had never finished an opponent, and pushed them below a puncher whose record had been padded against weak opposition. Nobody in that chain lied. An empty classification field was simply filled with a default value.

Three scoring systems sit under one name. That name is never the data.

The same pattern appears in health and safety data. Dehydration risk from weight cutting, cumulative brain-injury risk and weigh-in collapse risk are specific to the weight-class combat group. The performance group carries a different risk profile: joint injuries from movement difficulty, lower-back injuries from repeated twisting. Merging the two into a single “martial arts risk” column renders both risk profiles invisible.

The reasonable part of bundling

Saying that bundling is entirely wrong would also be wrong. There are legitimate reasons platforms keep doing it. Audiences consume these disciplines together: the same account that watches professional boxing also watches kickboxing and MMA, and commercially, bundling serves that demand accurately. The boundaries between disciplines are also genuinely blurring at certain points — crossover bouts between athletes from different disciplines, events applying hybrid rulesets, and the overlap between sanda and kickboxing are real phenomena, not marketing artefacts. Over-classifying carries its own cost: if each discipline is sealed inside its own box, analysts stop seeing the crossover cases, which is where most structural change in the industry actually happens.

The blind spot lies elsewhere. The label “martial arts” existing is normal. What matters is that the label is being used in place of another mandatory field. A compliant record needs two separate fields: discipline, and competition ruleset version. Discipline answers what this is. Ruleset version answers what this number means.

A second blind spot belongs to the analyst. Demanding perfect classification can become a pretext for publishing nothing at all. Holding an investigation for years because an exact label has not been determined is a failure, not a standard. Publishing a null result with a stated reason is the honest handling, and in this case, the null result itself is the information.

What needs to happen

The fix is far cheaper than the consequences. Make two data fields mandatory — discipline and ruleset version — before any metric is stored. Refuse to publish any ranking that merges the combat group with the performance group. And state the level of confidence explicitly when the evidence chain is incomplete, rather than filling the gap with a default number.

The sports industry has learned how to measure a great many things. What remains is learning to state clearly what is being measured, before measuring it.