Referees, Replay and the Analytical Framework Trap: Why 'Martial Arts' Is Too Broad a Label to Score a Fight

That night, in a small arena in District 1, the third round was four minutes deep. The fighter in the red shorts, a man from central Vietnam, had been breathing hard since the end of round two and was beginning to drop his hands. His opponent threw a downward elbow on a diagonal. It caught the cheekbone and slid behind the ear. I was standing less than three metres away. The mat made a dry, hollow thud. No whistle sounded, because martial arts have no whistle. There was only me, two fighters' hands, and roughly 1.2 seconds to decide where that contact sat within the three layers of criteria that define a round.

That is why I keep telling the young analysts in my group that the hardest part of officiating is not seeing a foul. It is choosing the right framework in which to read it.

Three layers of rules the crowd only ever sees as one

When spectators watch an MMA bout, they see two people fighting. When an analyst sits down at the desk, they must see at least three layers of rules stacked together. The first is the ruleset: what is permitted, what is banned. The second is the scoring rules: which legal acts earn points, how many, in what order of priority. The third, and the one almost nobody discusses, is the interpretation layer: when two criteria collide, which one wins.

These three layers are not identical across disciplines. A single analytical framework applied to all martial arts therefore fails on the first line, no matter how careful the writer is.

The clearest example: the Unified Rules of Mixed Martial Arts, drafted by the New Jersey State Athletic Control Board and adopted in 2026, later revised by the Association of Boxing Commissions in 2026 to push the concept of damage toward the centre. The current order of criteria runs effective striking and grappling, then effective aggression, then cage or ring control. The scoring scale inherited from boxing is the 10-9 system.

Professional boxing uses the 10-point must system under a different logic: the round winner must receive 10 even in a razor-thin round, and judges may not award 10-10. Traditional Muay Thai at Lumpinee and Rajadamnern scores round by round with unusual weights attached to the teep, to balance, and to a clean sweep, sometimes producing a winner who absorbed more strikes. Sanda scores legal strikes to legal targets and credits throws, with no interest in cumulative damage as MMA understands it. Grappling under IBJJF or ADCC scores points, advantages and penalties, where a four-second hold can decide a match. And wushu taolu, under International Wushu Federation standards, scores three groups: five points for technical quality, three for performance level, two for difficulty.

I list these not to show off. I list them to prove one thing: apply win-loss logic and finish rate to a taolu routine and you have produced an analysis that is meaningless but sounds highly professional. That is the most dangerous kind of error.

Why an over-broad label is a systems failure

In content classification work, I have received datasets tagged with a single label: martial_arts. That label throws a professional bout containing twelve elbows into the same bucket as a three-minute forms routine, a youth sanda match, and an unsanctioned challenge with no official referee.

For content people, this is a silent disaster. Before analysing anything, you must answer a classification question: which class does the subject belong to? Modern combat sports, meaning MMA, boxing, kickboxing, muay thai, sanda. Traditional and performance martial arts, meaning taolu, forms, choreographed sparring. Or the mixed class with no clear ruleset.

Each of the three needs a different analytical framework. Modern combat sports need strike accuracy, takedown success rate, and finish-defence data. Performance disciplines need difficulty values, balance losses, and technical deductions. The no-rules class can barely be analysed numerically at all and must be read through cultural context and social relationships.

I see what others do not, and I have to live with it. What I see here is a bad data label producing a bad analytical framework, and a bad framework producing wrong conclusions delivered in a confident voice. Nobody audits the data layer, because the data layer hides behind the presentation layer.

This explains why so much online martial arts analysis reads smoothly, cites numbers, uses jargon, and still teaches the reader nothing. It was written in the framework of a different sport.

The subjective gap inside combat sports replay

In football, VAR is a sealed room with monitors. In combat sports there is no such room. But there is a logical equivalent: the referee's right to review footage.

Some United States commissions, including Nevada and New Jersey, permit referees to review incidents within defined limits. Not everything. Not to overturn a winner. Usually to establish a specific violation, an eye poke, a strike to a banned area, or to determine whether a bout should continue after an injury.

Here is where I want readers to pause. The threshold for opening the replay is written in near-identical language everywhere: clear error, obvious mistake, an incident with a direct bearing on the result. It sounds rigorous. Ask instead: clear to whom, and clear in which framework?

An elbow sliding from cheekbone to back of the head, from three metres away, is one continuous motion. On a 60-frames-per-second camera at the opposite angle, it is three separate frames. On a 240-frames-per-second camera, it is twelve frames, one of which shows the point of the elbow sitting exactly on the boundary between legal and illegal territory. Whether that frame exists depends on where the camera stood and how fast it rolled. Physical reality is singular. The reality humans are permitted to see comes in several versions.

That is the subjective gap I always raise. It does not lie in whether referees cheat. It lies in the fact that the threshold for calling something obvious is itself a subjective decision, frozen into a clause.

The twenty minutes I betrayed myself

I still remember the Euro 2026 semi-final night, when I was twenty and my analysis group had reached twelve thousand members. In the 104th minute a collision in the penalty area blew the group apart. I reviewed the footage, saw only light contact, and concluded there was insufficient evidence for a penalty.

Twenty minutes later hundreds of objections poured in. I panicked and edited the post, changing the conclusion to a possible penalty. When the governing body confirmed the original decision was correct, some members mocked me with the very nickname I was proudest of: a VAR eye that follows the trend.

That semi-final night, I betrayed my own conviction to embrace the heart.

The lesson was not to never change my mind. The lesson was to separate two things and label each clearly. One section states what the rules say, grounded in statutes and footage. One section records fan sentiment, noting social reaction. These two are allowed to conflict. The writer is not allowed to let the second quietly rewrite the first.

Apply that lesson back to the labelling problem and the structure repeats exactly. The emotional layer always tends to overwrite the classification layer. Spectators see two people fighting, call it martial arts, and load onto it every expectation of a professional sport with rankings and contracts. Wrong from the root.

Asymmetric criteria and the illusion of fairness

There is something I call asymmetric criteria: two fighters in the same bout measured with different rulers, not because the referee is biased, but because the structure of the rules cannot avoid producing it.

A system like ONE Championship's scores the whole fight with weight placed on damage inflicted rather than scoring each round on the 10-9 system. It sounds fairer, rewarding whoever produced real consequences. It also creates a new asymmetry: a fighter who lands little but lands heavy can overtake one who lands often but lightly, and audiences are not trained to read that logic, so it feels absurd.

The 10-9 system has its own asymmetry. A round deemed very close must still have a winner, and forcing a winner in a balanced round creates a gap in the total, though the two fighters were near-identical. Three such rounds become 30-27 for one man while the arena felt a dead-even fight.

Taolu is asymmetric in a completely different way. Technical deductions are visible and marked by eye, but performance scores in group B are almost purely felt. A routine can lose points on a detail the crowd never notices, with nobody able to explain it to them. That is why disputes at performance events linger longer than disputes at combat events, even though media pays them far less attention.

What the main camera misses

From years standing close to the mat, I learned there is information that exists only within three metres. Breathing. The way a fighter shifts weight into the rear foot after taking a body shot. The very short silence before someone commits to a strike.

Cameras capture image but not impact feedback. When a kick to the body lands cleanly, the receiving body reacts in a specific sequence: the chest drops, the shoulders close, the hips push back. That sequence takes about a third of a second and cannot be convincingly faked. If I see the sequence, I know the kick had weight. If I only see a grimace followed by immediate recovery, I start to doubt.

This is the kind of detail I want in my analysis, because it is the one thing readers cannot look up. Anyone can open a stats sheet and read significant strikes. Nobody can feel for me what a contact at three metres feels like.

When I re-examine footage of disputed incidents, a fairly consistent pattern emerges: most controversies erupt at moments when the main camera sits at an oblique angle and nobody has a head-on view. The first whistle is blown by the camera, and I hear the whole world holding its breath, because the world is holding its breath over a missing frame.

Counter-intuitive: the fault is in the data layer, not in the people

After years inside debate groups, I reached a conclusion that initially unsettled me. Most officiating errors that communities call bad refereeing are actually failures of the data and classification layers sitting above the referee.

When a promoter does not publish specific scoring criteria per weight class and discipline, the referee must invent a private scale in his head. When an event blends disciplines without stating which ruleset governs judging, officials must swap frameworks between bouts. When a data platform attaches a single label to all martial arts, every downstream analysis is built on sand.

This is the counter-intuitive point, and it runs against the instinct of the majority. The instinct is to find someone to blame. If the referee erred, replace the referee. If the referee was right and the crowd still booed, then the crowd is wrong. That thinking ignores a structural fact: a system that allows multiple analytical frameworks to coexist without labels will always generate controversy, no matter who occupies the chair.

Five thousand members, one old monitor, and trust split into five thousand pieces. I have sat among those five thousand pieces many times. What I learned was not how to persuade them. It was how to let them see the framework I was using, so they could check whether that framework matched the discipline being fought.

The referee blows the whistle, but the fans are the ones who deliver the final score. And fans only score correctly when they are handed the right ruler.

Four things to fix before we discuss reform

First, and cheapest: lock the classification before analysis. Every analysis should open with one line stating whether the subject belongs to modern combat sports, performance disciplines, or the mixed no-rules class. That single line removes most downstream error.

Second, publish scoring criteria per event in language a non-professional can read, with video examples for each score level. Without examples, every reader invents a private version.

Third, keep a judging rationale log. It need not be published immediately. Simply requiring one sentence of reasoning per round changes the threshold of attention. I did this for two seasons and found I missed far less.

Fourth, standardise minimum camera coverage for reviewable incidents. If the rules allow replay, the rules must state how many angles are required, at what frame rate, and which angle governs when angles conflict. Without that clause, replay rights merely add another subjective layer.

What I want to leave behind

A fight is not scored by one number. It is scored by a chain of small decisions, each dependent on whether the judge chose the right framework. That framework sits in the data layer, before the fight begins, and is usually invisible.

If you are a spectator, next time a decision angers you, try asking a different question instead of whether the referee has eyes: what rules governed this bout, and were they stated from the start? The answer to the second question usually explains most of your anger.

For analysts like me, what must be preserved is not the fame of a nickname. It is the habit of correctly labelling what we are looking at, before we open our mouths to say anything about it.