Data Extraction Failure Behind Table Tennis Analysis: When Stage-2 Speaks Without Evidence

I once spent 90 days watching 156 women's table tennis matches, most of them played in near-empty halls, only to discover that the report I had built stood on nothing at all. That experience taught me a discipline I now hold firmly: an untraceable number is more dangerous than the absence of a number.

The Stage-2 analysis report that reached my desk this week is the cleanest example of that disease I have seen. Every Stage-1 field was empty. The article title read N/A. The source read N/A. The article type was classified as "Unclassified". The Information Points list was empty. Core Viewpoints was empty. Entities Involved contained no name at all, only an instruction: "identify from the information points above" — when there were no information points above. Time Sensitivity was never assessed. Source Quality was never assessed.

What is striking is not that the report was empty, but that it was still written across nine dimensions with tables, a risk matrix, and a technical glossary — all filled with N/A. That is where I stopped and exhaled, because I recognised the pattern. This is not an analytical failure. It is a pipeline failure wearing the costume of analysis.

I have three times mispronounced Megan Rapinoe's name on live television and endured 1,200 re-shares within 24 hours, with 87 percent of comments negative. From that I learned that an honestly admitted error is far cheaper than an error concealed behind confident language. This Stage-2 report chose the second path. It did not invent players. It did not invent rankings. It did not invent head-to-head records. But it also did not stop.

Consider what actually happened in each dimension. The first, covering technique, tactics, and equipment, returned N/A on advancement, execution effectiveness, physical fit, and key data. No playing-style system, no serve or receive element, no rally data. The second, covering player data and head-to-head records, returned N/A on world ranking, points-defence pressure, foreign-match win rate, and clutch-point performance. The third, covering the event system and points rules, returned N/A on event positioning, champion's points, prize money, and draw analysis.

The fourth, covering the competitive landscape and the China-versus-world frame, returned N/A on world top-ten seats, titles across the last five editions of the three majors, and U21 generational depth. The fifth, covering rules and governance, returned N/A on competition-rule reform, event-system rules, selection rules, and disciplinary sanctions. The sixth, covering coaching staff and the talent pipeline, returned N/A on head-coach capability, staff stability, main-tier age structure, and youth conversion efficiency.

The seventh dimension is the only one that actually contains something. Its risk matrix records six competitive risk categories, all N/A, then adds a seventh row: analytical-integrity risk, rated High, with likelihood High if analysis proceeds, impact High on reputation and decision quality, and mitigation in the form of blocking publication until Stage-1 is populated. That is the only defensible finding in the entire document: the pipeline is blocked at the extraction stage and must be re-run before Stage-2 can produce any value.

The eighth dimension, covering public narrative and expectations, returned N/A on narrative label, heat-cycle position, fundamental support, and expectation-gap analysis. The ninth, covering table tennis industry transmission, returned N/A on the equipment market, the grassroots training base, the event commercial ecosystem, player commercial value, policy and capital, and the international ecosystem.

I want to be honest about something that may be uncomfortable for colleagues in the newsroom. I spent 33 years at the Daily Mail, and much of my early career was fact-checking. That work was tedious, invisible on screen, and never made it into a clip shared 1,200 times. But that work is precisely what stopped us from publishing reports that sounded authoritative while containing nothing.

There is one small detail in that document that I consider the most important and the easiest to miss. The "Additional Notes" field contains instructions, not findings. Instructions to derive entities from an empty point list. Instructions to judge source quality from an empty source field. When a template is emitted without ever being populated, what we are witnessing is not an article without content, but a system that failed to retrieve content and still reported as though everything were normal.

From a sociological standpoint, this is a phenomenon I know well. Institutions tend to preserve the form of their rituals long after the meaning has gone. Meetings are still held, forms are still filled, reports are still filed, and not one person asks whether anything is inside them. Nine dimensions with tidy tables are that form of ritual. They give the reader a sense of safety that the work has been done.

I will not pretend this is about table tennis. It is about how we treat sports information generally, and women's sports in particular. Over many years I have observed that coverage of women's sport is the most vulnerable to filling voids with narrative. When data is unavailable, when statistics are not collected, when matches are not recorded, that empty space gets filled with story. Stories about struggle, about tears, about inspiration. All of it sounds warm and none of it can be verified.

That is why I refuse to fill the void in this report with model memory or background knowledge. If I wrote a player's world ranking from memory, that figure might already be stale against the WTT rolling 52-week calendar. If I wrote a head-to-head record from general knowledge, I would be selling background as analysis. Readers deserve to know the difference between what I know and what I checked.

The recommendations arising from the document itself I consider correct and I endorse them fully. First, block publication and re-run Stage-1 extraction against the source document. Second, verify at least three populated information points and a non-null source field before Stage-2 is invoked again. Third, add a hard validation gate that rejects any Stage-1 payload whose Information Points list is empty or whose title and source are N/A. Fourth, forbid model-memory substitution for quantitative fields, or at minimum require a "data pending verification" label.

There is one note I would add myself. If re-extraction again returns empty, the source article most likely was never a substantive competitive article to begin with. It could be a photo gallery, a fixture list, or a JavaScript-rendered page that failed to fetch. In that case, the correct Stage-2 output is not a nine-dimension report but a single short paragraph stating that no assessable competitive content exists.

I close with something I have believed since my first video drew only 212 views in three days when I was 51. The value of an analysis is not determined by its length, by the number of its tables, or by the confidence of its tone. It is determined by whether someone can trace every claim back to its source. 212 viewers taught me that better than 30 years inside a studio. And a nine-dimension report entirely filled with N/A taught me the same lesson once more, from the opposite direction.