Three Judges, One Scorecard: The Fracture Inside Taolu Scoring

The scoring file sat on an old hard drive belonging to the event's secretariat. Last modified: 23:41, the night before competition day. The file was named final_v3. In the same folder were two others — final and final_v2 — saved at 21:12 and 22:55. Three versions of the scores for one event, on one evening, none of them accompanied by a written justification. That is the starting point. It is not the conclusion. On 14 November 2026, an international wushu event in East Asia staged the men's taolu discipline with 23 athletes and five judges seated at the row of tables in the middle of the arena. Scores appeared on the big screen less than four minutes after each routine. None of the 1,400 spectators that night saw the file final_v3. I received the folder eleven weeks later, from a person who had worked in the technical secretariat. That person asked not to be named. I accepted the condition, then spent seven weeks cross-checking every number against the published scores on the organiser's official results page. Taolu scoring runs on three panels. Panel A assesses movement quality, on a scale of 5.00. Panel B assesses power and speed, on a scale of 3.00. Panel C assesses technical difficulty, on a scale of 2.00. Maximum total is 10.00. Panel C is scored by the technical committee independently of Panels A and B. The three panels are barred from communicating during competition hours. That rule exists for a technical reason. Difficulty is a countable variable — number of rotations, landing angle, jump height. Movement quality is a perceived variable. Blending two variable types into one table lets systematic error spread across both. Federations separated Panel C from Panels A and B in 2026, after three consecutive events with disputed results and formal protests. In the file final_v3, the Panel C score of the fourth-placed athlete was edited from 1.85 down to 1.65. A reduction of 0.20. The gap between fourth and third place was 0.14 points. That subtraction was enough to change the ranking, and to change the qualifying slot for the continental championship held the following April. I built a comparison table of the three file versions against the published results. The finding: nine of the 23 athletes carried a Panel C score in the file that differed from the score printed on the public board. Total absolute discrepancy was 1.42 points. There was no discrepancy at all in Panel A or Panel B. A scattered error across nine athletes in a single scoring panel has a different structure from random error. Random error distributes evenly across all three panels. Here, two panels were perfectly clean and one was skewed in a direction. The direction deserves attention. In seven of the nine cases, the Panel C score was adjusted downward. Two were adjusted upward. Of the seven reductions, five belonged to athletes from the smallest-budget delegations at the event — delegations with no head coach in the technical area, delegations paying their own travel and accommodation. Based on my experience covering fights and martial arts championships over twelve years, judge-to-judge variance in Panel A typically runs between 0.1 and 0.3 points. Variance in Panel C rarely exceeds 0.1 points, because Panel C counts rather than judges by eye. A 0.20-point drift in Panel C sits outside the normal observation band. I contacted three of the judges who served on the technical committee that night. The first replied by email, stating that Panel C scores were locked before leaving the floor and signed onto a paper record. The second did not respond after four attempts. The third said by phone that the paper record had been submitted to the secretariat and that this record was the authoritative version. The secretariat declined to provide the paper record. The organiser replied with a two-paragraph letter stating that all score changes followed the protest procedure and carried the signature of the technical committee chairman. The letter enclosed no copy of any record. The arena was spotless. The secretariat room was not. In parallel, I cross-checked the event's doping control files. The first two samples from one athlete in the group whose scores were adjusted sat inside the permitted threshold. The third sample, collected fourteen hours after the final routine, showed a markedly different reading from the first two. The laboratory does not know the athlete's name. That is why I trust them. The third urine sample said what the first two did not dare to say. There is another reading, and it holds up technically. A Panel C score may legitimately be recalculated when the technical committee identifies an error in the original record — a movement logged at the wrong difficulty tier, or an athlete performing a movement absent from the pre-registered list. That procedure exists, has precedent, and was applied at two continental events between 2026 and 2026. Second, my data covers three file versions only. Four other versions in the save chain sit beyond my reach, and they sit with the organiser. Third, the secretariat's refusal to hand over the paper record is not itself evidence of wrongdoing. An event governing body holds the right to keep internal files confidential while a protest is under review. I do not infer intent from a single refusal. What I keep is the structure of the data. Nine athletes with divergent scores. Nobody divergent in Panel A or Panel B. Seven divergences running against the least-resourced delegations. And a file modification timestamp falling after the moment the paper record was signed, according to one judge's account. Taolu scoring does not need more judges. It needs an audit trail — every score edit leaving a timestamped, digitally signed, written reason behind it. Football learned this lesson after introducing VAR, and is still wrestling with the definition of a clear and obvious error. Performance martial arts faces the same choice, except here the numbers are smaller and the consequences land on the people with the least voice. A scorecard without a revision log is not a scorecard. It is a draft nobody has agreed to sign.