The 'martial arts' label is too broad: why combat sports data needs a new taxonomy
**Core answer**: Nhãn "võ thuật" gộp ba hệ thống thi đấu khác nhau — đối kháng hiện đại (MMA, quyền Anh), quyền thuật biểu diễn (wushu), và tán thủ — vào một ô dữ liệu duy nhất, khiến mọi phân tích so sánh trở nên vô nghĩa. Cần tách nhãn bộ môn trước khi phân tích. **Key facts**: - "Võ thuật" bao trùm ba logic: thắng-thua (MMA, quyền Anh), chấm điểm biểu diễn (wushu), và lai (tán thủ). - Hệ thống 10 điểm (10-9) dùng cho MMA và quyền Anh, không áp dụng cho quyền thuật biểu diễn. - Chỉ số nỗ lực (số đòn/phút, quãng đường) chỉ có nghĩa khi gắn với một luật thi đấu xác định. - Dữ liệu rỗng không đồng nghĩa "không có rủi ro"; đó là báo động về chất lượng đầu vào. - Ba trường dữ liệu tối thiểu cần có: nhóm môn, luật thi đấu cụ thể, và thang đo sử dụng. **Source attribution**: Báo cáo phân tích chuyên sâu Stage-2 (lĩnh vực võ thuật, nhãn miền martial_arts) | Cross-checked: VuaBong.vn **Related Q&A**: Q: Vì sao không thể so sánh trực tiếp một võ sĩ MMA với một võ sĩ wushu? A: Vì hai hệ thống dùng thang đo khác nhau — thắng-thua so với chấm điểm biểu diễn. Q: Một bộ dữ liệu trống có nghĩa là không có rủi ro không? A: Không; đầu vào rỗng là lỗi pipeline cần khắc phục, không phải xác nhận an toàn. Q: Cần tối thiểu những trường dữ liệu nào để phân tích võ thuật đáng tin? A: Nhóm môn, luật thi đấu cụ thể, và thang đo sử dụng.
In the Asian combat sports database I once helped process, one column made every calculation meaningless: the "discipline" column. It recorded just two words — martial arts. The first time I tried to filter for bouts with the highest early-finish rates, the results mixed a professional MMA fight, a scored forms routine, and a sanda bout. Three things sat in one cell, yet they are not measured with the same ruler.
I remember a morning in Beijing, when the analysis team produced a "martial arts performance" ranking for a regional event. The score column looked so clean that nobody wanted to question it. Only when we checked it against the judges' records did we see that half the rows were scored on a forms scale and the other half by win-loss. That ranking, technically speaking, compared nothing at all. Every record has two pages: the published page and the hidden page, and here the hidden page was the definition of the discipline itself.

Martial arts, at the level of language, is an umbrella term. At the level of data, it is a trap. Such a label merges three competition systems with entirely different logics into a single cell, and when three logics are locked together, the result brings noise rather than richness.
The first system is modern combat sports — MMA, boxing, kickboxing, Muay Thai, grappling. They operate on win-loss logic: there is an opponent, there are rounds, and there is a finish by knockout, submission, or judges' decision. The data that matters here is finish rate, opponent quality, significant strikes per minute, durability, and takedown ability.
The second system is performance forms, such as tai chi or the routines of wushu. There is no direct opponent. Points come from judges scoring difficulty, precision, and artistry. A competitor who "wins" here may never have fought once in an entire competitive career, and within their system that is entirely normal.
The third system is sanda — a combat discipline allowing punches, kicks, and throws, sitting between tradition and modernity. It has an opponent, but its rules and scoring differ sharply from MMA or boxing. A strong sanda fighter may not necessarily transition well to MMA, because the weighting between strikes and throws is not the same.
These three systems require three different analytical toolkits. Merge them, and we do not have data — we have a pile of meaningless numbers presented neatly.
When I re-examined the structure of a large dataset, the problem surfaced at the very input layer. A domain label that is too broad collapses the entire analysis chain behind it. Without a specific athlete's name, without an event, without an organization, every metric hangs in the air with nothing to anchor to. The breadth of a label determines the maximum depth of every analysis that follows it.
This sounds abstract, but the consequences are concrete. Take scoring. In boxing and MMA, the 10-point system requires the winner of a round to take it by at least 10-9. In performance forms, scores typically cluster around a 10-point scale with weights for technical difficulty. If someone dumps both kinds of scores into one column and averages them, that average says nothing about anyone.
The same applies to effort metrics. Strikes per minute, distance covered, sprint counts — they only mean something against a defined ruleset. A fighter who kicks continuously for three rounds may top a "workload" table, but most of those kicks hit air. Based on my experience watching fights, attractive effort metrics are often used to praise a performance that was, in substance, ineffective running. Distance covered and sprint counts get packaged as effort indicators, yet ineffective running also produces beautiful numbers.
At the organizational level, the problem is even clearer. A professional boxing body, an MMA promotion, and a wushu federation do not share ranking systems, contracts, or weight standards. Merge them into an "Asian martial arts map" without layering, and you will draw false transmission paths — for instance, crediting the rise of an MMA movement to the achievements of a traditional federation, when the two are barely related. In many markets, exclusive contracts also prevent athletes from moving between organizations, making cross-organizational head-to-head data almost nonexistent.
In Vietnam, the picture is even more complex. On one hand, traditional martial arts are preserved as cultural heritage, with their own rank systems and forms. On the other, professional boxing and MMA are growing, with domestic events and athletes competing on the regional stage. When the media calls all of it "Vietnamese martial arts," readers struggle to tell which is a genuine combat result and which is a performance or lineage activity. These two kinds of achievement require two different ways of reporting, and blending them harms both.
Trustworthy martial arts data must begin by separating labels. That is the least glamorous step, but it is the decisive one. A 90-minute match is only a moment; a 300-day cycle is the truth, and that cycle can only be measured once we know which discipline we are measuring.
There is a temptation analysts often fall into: when data is missing, they still want to reach a conclusion. An empty report is often misread as "no risk detected" — as if silence were safety.
The reality is the opposite. An empty input is an alarm, certainly not a green checkmark. When there is no athlete's name, no event, no verifiable number, every judgment is speculation. An honest writer must say plainly: there is not enough data here to conclude. Saying "not enough" is far harder than inventing a conclusion that sounds reasonable, because it requires accepting that one does not yet know.
I once paid for haste. In an old project, I published a cycle model before I had enough control samples, and part of the prediction was wrong. Since then, I have learned to state risk variables and data gaps within the analysis itself. A model that admits its limits is more trustworthy than one that claims perfection. The same holds for a martial arts ranking: better to write "unclassified" than to merge carelessly just to get a clean number.
A fix is not especially complex. Each data row should carry at least three fields: discipline group (modern combat, performance, or hybrid), the specific ruleset, and the scale used. Once these three fields are separated, comparisons become feasible. When they are merged, every number is just decoration.
With martial arts, the first question was never "who is strongest." The first question is: which discipline are we talking about, under which rules, measured by which ruler. When the track runs long, initial speed is only an illusion — and when a label is too broad, every ranking is only an illusion presented neatly. Data does not need fans; it only needs patient readers.
