Vietnamese Martial Arts and the Data Gap: 41 of 47 Records Left Blank

At 2 a.m., a 47-row spreadsheet ran through the analysis system and came back with exactly one surviving column: the domain label "martial arts." Title blank. Source blank. Type blank. One-sentence summary blank. Author stance blank. Article purpose blank. And the list of information points, completely blank, not a single row. An outsider would call that a broken system. I read it differently. It is the most honest snapshot of the fight records I work with, after more than thirty years spent between gyms and monitors. The only thing that survived a night of processing was the one thing that needs no content: the domain tag. Everything requiring a concrete fact disappeared. In my trade, an information point is the smallest claim that cannot be argued with. The fighter rested eleven days after a left ankle injury. Tuesday's session ran 96 minutes, load up 14 percent. In the third round against the Thai opponent, strike frequency dropped noticeably. Each sentence like that is a brick. No bricks, no wall, only a blueprint hung on the wall. The system returned seven empty fields at once. Those seven fields are not independent. They are a domino chain, and the first piece to fall is the list of information points. With no information points there is no entity to identify, so the instruction "identify entities from the information points above" hangs in the air, because above it is nothing. With no entities there is no stance to evaluate. With no stance there is no purpose. And with no purpose, every summary is invention. What caught my attention is how those seven fields collapsed. They did not collapse because the algorithm concluded the article had nothing worth saying. They collapsed because the extraction layer in front of it never fed in a single word. There is a wide gap between "analysis finished and found emptiness" and "there was never anything to analyze." My spreadsheet belongs to the second kind. To a rehabilitation specialist, that gap resembles the difference between a clean scan and a blurred scan caused by a patient moving. Both produce a sheet nobody can read. But the first says the body is fine. The second says the machine is broken, and if you draw conclusions from it you are diagnosing the machine, not the person. In Vietnam, boxing, muay, vovinam and pencak silat have stepped onto the continental stage behind names such as Truong Dinh Hoang and Nguyen Thi Tam. Clubs have started buying wearables, collecting GPS data in training, logging rounds on the heavy bag. The collection layer exists. The conversion layer does not. A coach writes by hand in a notebook: "left leg sore, rest two sessions." That sentence enters the system and becomes a blank cell, because the system cannot read handwriting on a page photographed with a phone. There are five failure modes I have seen often enough to spot instantly in a results table. First, unsupported input format: images, scanned PDFs, javascript-rendered pages. Second, an extraction error swallowed silently, the system returning defaults instead of raising an alarm. Third, a template bug, where the information-points section never renders and therefore never gets filled. Fourth, language encoding. Fifth, truncation of overlong text, leaving the sent chunk empty. The fourth is the one I meet most often here, and it is tied directly to Vietnamese. "Chan thuong" with diacritics is one string. Without diacritics it is another. An extractor trained mainly on English meets those two strings as two different words, or worse, as two noisy strings. When a coach types fast on a phone mid-session, the diacritics drop. When an assistant re-enters it on a computer, the diacritics return. Same injury, two spellings, and the system counts two cases. The third mode is more dangerous because it is invisible. A broken template means the most important part of the report is never generated. The operator looks at the screen, sees other fields populated, and believes the system is running. They do not know that the information-points list, the only part that actually carries information, is an empty cell sitting inside a report that looks complete. Four steps I once applied to three small clubs, and all three showed results within a month. One: always keep the raw text beside the processed output, so a blank cell still has something to be checked against. Two: manually inspect the first twenty records of every intake batch, with human eyes, not the machine. Three: standardise injury names, with and without diacritics, into a single code. Four: every blank cell must carry a reason for being blank, even if that reason is one word: blur. The domain label survived. It survived because it needs no content, only a keyword. This is where I want to linger, because it is the biggest lesson of that night. The "martial arts" tag may come from a genuine technique breakdown, or from a glove advertisement with "martial arts" in the headline, or from a political piece that mentions martial arts once in its final paragraph. Without information points, nobody can tell those three apart. In a decision system, those three cases lead to three different decisions. A technique breakdown leads to adjusting the training plan. An advertisement leads to skipping it. A political piece leads to doing nothing. And if all three carry the same tag, the system treats them identically. I have fallen into exactly this trap myself, in the opposite direction. In July 2026, assessing Brazilian forward Alan Carvalho for a club in Guangzhou, I reviewed 47 matches across 18 months and found his sprint output dropped 15 percent on artificial turf. I advised the club against a long-term contract. Six weeks later he suffered a hamstring injury against Shanghai SIPG. The data saved a contract, but only because the raw data was still intact to be re-read. Injury data never lies, only the reader lacks patience. But that sentence only holds when the data exists. A blank cell does not lie; it simply says nothing. In 2026, when the Chinese top flight was suspended and the stands held nobody, I lost every commentary contract. I contacted 23 young Guangzhou players, collected sensor data from their at-home sessions by phone, and spent eight months building a load-recovery model. When the league returned in June, the team recorded only four injuries in its first ten matches, 30 percent below the average of the previous two seasons. But the model sat scattered across twelve separate spreadsheets and was never widely adopted, because long-term planning is not my strength. The 2026 spreadsheet taught me that the body never rests, it only needs a patient enough algorithm. It also taught me that a correct model nobody uses is just a beautiful model. Kazan taught me that public opinion is noise and numbers are signal. In July 2026, during the Brazil-Belgium quarter-final, while the world still believed Neymar would shine after his foot injury, I presented data from twelve matches: change-of-direction ability down 12 percent in the second half, left thigh response 0.3 seconds slower. Brazil lost 1-2. Listenership for the programme rose 300 percent overnight. I retell this to make one point: if that night I had only a "football" label and an empty information-points list, I would have had nothing to say, and someone would have said Neymar was fine. There is a way to read this whole story backwards, and I want to put it on the table before closing. People usually blame the machine. But in most gyms I have entered, nobody is paid to keep clean records. Coaches write reports to protect themselves, to explain why a student lost, to ask for more budget. That report is a political document written in medical language. It is not designed to provide information; it is designed to exert influence. A blank cell inside such a system is not necessarily a technical fault. It is a behavioural signal. And there is one more angle rarely mentioned: a blank cell is more honest than a fabricated one. If the extractor cannot read the coach's handwriting, returning "undetermined" is correct behaviour. What is wrong is that nobody reads that notice. A system willing to say "I do not know" is worth more than one that guesses and presents the guess as data. A body reader like me knows: every ache is an answer. But the body only answers when someone asks the right question and records the answer the right way. In Vietnam, the club layer is ready to spend on hardware. The question left is who takes responsibility for the conversion layer in between, the person who turns a smudged notebook into a readable data row. If the answer is still "nobody," then ten years from now we will have more sensors, prettier dashboards, and exactly the same forty-one blank cells as tonight.