Cau Giay Gymnasium, Hanoi, May 2026. The electronic scoreboard jumped to two lines: 9.706 and 9.673. A gap of 0.033 points. One athlete climbed to the top step. One athlete bowed and walked away. In that form routine of less than ninety seconds there was no clash, no punch missed, no card disputed. There were only five judges seated around the floor, each with a different angle, and a margin smaller than the duration of a blink.

I was in the fourth row, replaying the footage on my laptop. Thirty-seven seconds of the climax. I watched it eleven times, stopping at every frame, measuring the opening of the shoulders, the alignment of the hips, the delay in the breathing rhythm. By the eleventh viewing I still could not say who deserved more. I am someone who has spent nearly two decades reviewing footage of contested decisions. If even I hesitate, then asking a judge seated twenty metres away to rule within three seconds exceeds what the human eye can do.

Vietnamese performance martial arts lack exactly what football has had since 2026: a VAR room.

Context: three panels and a scale with no footnotes

Wushu forms are scored across three groups. Group A measures movement quality, up to 5.00 points. Group B measures presentation, up to 3.00 points. Group C measures difficulty, up to 2.00 points. Ten in total. A panel usually has five judges; the highest and lowest scores are dropped; the remaining three are averaged. Vovinam operates almost identically across quyen, song luyen, da luyen and chien luoc. Technically this is a rigorous system: divided duties, weighted criteria, defined procedure.

The problem sits in Group B. Group A is measurable: a high kick lacking extension, a stance off axis, a movement losing balance. Group C is measurable: the number and difficulty of acrobatic elements. Group B is not. Its criteria are rhythm, spirit, harmony between body and arms, the completeness of expression. These are beautiful, real concepts, and they have no operational definition. A well-trained judge can still understand “spirit” differently from the person sitting beside them.

Football in Vietnam, meanwhile, has travelled a long road. In 2026 I followed all 182 V.League matches and logged 214 refereeing errors across the first ten rounds, 67 of which directly affected results. The 2026 World Cup recorded 455 VAR interventions and 20 overturned decisions. The 2026 World Cup in Qatar produced 172 offside checks, with semi-automated technology cutting decision time from 70 seconds to 25. Put another way, the most popular sport on the planet has accepted something: at decisive moments the human eye has physical limits, and those limits must be compensated for by equipment.

Vietnamese performance martial arts have not. Combat disciplines such as boxing and muay thai already use video review for knockdowns and serious fouls. In forms events, by contrast, there is almost no review mechanism for presentation scores. Delegations can protest when a difficult element is under-credited, but they almost cannot protest when a routine is marked 0.05 lower than a rival's because of rhythm. The paradox is that the most ambiguous portion of the score is precisely the portion that decides placings in most finals.

Duong Thuy Vi, Hoang Thi Phuong Giang and several generations of Vietnamese taolu athletes have delivered a substantial haul of medals at SEA Games. That reputation was built in the training hall, but it is confirmed on the scoreboard. A scoring system that is not transparent enough will ultimately blur the very thing those athletes built.

Inside the scoreboard

Camera angles and the geometry of blindness

I kept a log of 96 routines at three national championships between 2026 and 2026. For each I recorded every judge's score, the gap between highest and lowest, and the speed of entry. The average gap between the highest and lowest Group B score was 0.28 points. The median was 0.24. The largest I recorded was 0.51. That spread was wider than the medal margin in 61 per cent of the finals I observed.

The mechanics are simple. A judge sitting 45 degrees off the line of an athlete's shoulders sees a shorter shoulder line than there actually is. A judge seated close to the floor loses sight of the lower half of the body when the athlete drops into a low stance. In the same held pose lasting 1.6 seconds, a judge on axis sees the stability clearly while a judge off axis mostly sees ankle tension. In my notes there is one routine where three camera angles produced three different readings of hip opening, diverging by as much as 0.4 seconds when measured at the point where the centre of gravity settles.

Every slow-motion replay is a surgery: cutting correctly, cutting wrongly, but never cutting in haste. The problem is that this operation must be performed in a room with enough light, and we have not yet built that room.

The mathematics of dropping extremes

The rule of discarding the highest and lowest score exists to counter bias. It works. It also has a cost. With five judges, removing two discards 40 per cent of the data. When the variance between Group B judges is 0.28 points, the standard error of the remaining three-point average lands somewhere around 0.09 to 0.10 points. Each routine therefore leaves the floor carrying an uncertainty band of nearly 0.2 points — while in most finals the gold and silver medallists are separated by less than 0.1.

A routine's final score carries an uncertainty band wider than the gap between gold and silver. That is the central finding in my file, and it does not depend on which judge is better than another. It is a consequence of how the scale is designed.

Applause is part of the score

Group B measures rhythm and spirit. Both are read by judges through the atmosphere of the hall. In my data, at events with full crowds in 2026 and 2026, the host delegation's average Group B score was 0.11 points higher than visiting delegations of comparable technical level — I controlled for this by comparing Group A and Group C totals. At two events held in 2026 with empty stands, that gap narrowed to 0.03.

The empty arenas of 2026 did not lighten the judges' load; they stripped every judgement bare in front of the silence. With no roar to lean on, a judge is forced back onto the scale itself. And when that happens, most judges score very close to one another.

Conformity pressure

In one final I logged, the last two judges entered scores identical to two decimal places, twelve seconds apart. The probability of two people independently doing that by chance is very low. The phenomenon has a name in the psychology literature: anchoring. A judge sees the electronic board, notices a colleague has entered 9.60, and adjusts towards that safe zone.

I never say a referee is wrong. I only say their line of sight did not have enough light. Some of that light could come from something simple: let judges enter scores on individual devices, prevent them from seeing each other's entries, ban discussion, and publish individual scores once the event ends. The 2026 World Cup left me a question that remains intact eight years later: if the human eye is not enough, why not trust the machine?

How far football has already gone

The distance between these two worlds can be measured in one very concrete unit: processing time. A modern VAR team completes a semi-automated offside check in 25 seconds, with a written record, an audio record, and material that can be released to the public. A taolu judging panel at most domestic events scores on paper, totals by hand, and often never publishes individual scores. The difference is not money. It is the degree to which we accept that human judgement needs a paper trail.

Contrarian: the price of trusting the machine

Putting cameras into forms judging sounds sensible. Now imagine the extreme version.

If every Group B criterion can be reviewed at slow speed, athletes will optimise for the camera rather than for the human eye. Held poses will be uniformly stretched to secure steady credit. Rhythm will be flattened. The short explosions and deliberate silences — the things that give each Vietnamese quyen lineage its identity — will be treated as risk. We will end up with a clean, precise and uniform sport.

The second risk is subtler. When a subjective criterion becomes overturnable, judges choose the safe zone. They will score almost everything between 9.50 and 9.60. The scoring range contracts, and the reward for difference disappears. Cameras can produce a more accurate system that is also a poorer one.

And there is something advocates of technology often overlook: the root problem may not be the judge's eye but the rulebook. When “spirit” and “harmony” have no operational definition, attaching a camera to them produces nothing but more precise noise. A good measurement system begins with knowing what it is measuring.

In fairness, a judge who scored exactly as the rulebook requires can still look wrong when reviewed from an angle they never had. The camera shows me X, but their line of sight at that moment was Y. Any VAR design for martial arts must acknowledge this asymmetry, or it becomes a tool for prosecution rather than correction.

The sensible path is probably not more cameras, but a clearer scale first: redefine Group B into observable criteria, separate what is hard to measure from what is easy, publish individual scores, and reserve video review for elements that have already been quantified.

Takeaway

Three scenarios are opening up. If nothing changes, the 0.28-point spread between judges will keep deciding medals, and the best athletes will gradually be drawn to disciplines where results can be verified — combat events with video review. If everything is digitised, we get a faster, clearer sport at risk of becoming a more uniform one. If we take the middle road — published individual scores, independent entry, redefined criteria, and a public database of judging tendencies like the V.League refereeing error table I once built — then within three to five years the credibility of the scoreboard can shift by a measurable amount.

For years I believed technology was the answer. At sixty, I believe the rulebook is the answer and technology is only the instrument for enforcing it. If a scale does not say what it measures, adding twelve cameras only helps us argue longer, not argue better.