The Report With Nine Sections and No Data: The Silent Failure Eroding Sports Analytics

The nine-section report went out at two in the morning. Everything was in place: section headers, comparison tables, benchmark columns, a risk-warning block, even a glossary at the end so the reader would not get lost. Only one detail was wrong. Every data cell read "insufficient information". No player was named. No tournament was identified. No date appeared anywhere. The document read like a building with walls, doors and room numbers, but nobody inside.

The person who wrote it was not drunk. He followed the process, the template, and everything he had been trained to do. In sports data analysis, a wrong conclusion still has a chance to be corrected once the ball rolls. What is more dangerous is a conclusion that is formally correct and hollow, presented well enough that nobody bothers to open the source and check.

The two-stage funnel

Every professional sports analytics system I have operated runs through two stages. Stage one is extraction: it strips a source text into atomic units of fact — a player, a scoreline, a timestamp, a ranking figure with a snapshot date. Stage two is analysis: it places those units into nine dimensions — technique, player and head-to-head data, event systems and points rules, the Asian landscape against the rest of the world, governance and regulations, coaching staff and talent pipelines, risk surface, public narrative, and industry transmission.

In table tennis, stage one is stricter than in football. The WTT ranking operates on a rolling 52-week mechanism: old points expire on their own, new points must replace them. A ranking without a snapshot date is not data — it is a sentence with no one accountable for it. The three majors are the Olympic Games, the World Championships and the World Cup, each with a different points coefficient and field strength, so pulling a player out of the calendar pulls him out of his own record.

Stage two only has value when stage one delivers goods. When stage one returns zero, stage two still runs. It still prints nine sections. It still bolds the conclusion. It still ranks risks by priority. That is the moment the funnel becomes a mirror: it reflects the exact shape of an analysis without reflecting anything at all about the match.

Anatomy of an empty case

Three hypotheses for one symptom. The extraction pipeline may have degraded silently: it received a scanned image page, a paywalled page, a video with no transcript, or an anti-scraping interstitial — and instead of raising an error it returned an empty frame. It may have been pointed at a document that does not exist. Or the source genuinely contained no table-tennis substance worth extracting: a photo caption, a listing page.

Those three hypotheses cannot be separated by looking at the output. But one trace points directly at a template defect rather than a source defect: the instruction "identify entities from the information points above", while the information-points array is empty. An instruction that refers to nothing. No journalist writes a sentence like that. Only a carelessly programmed template produces it.

What bothers me most is the part that was handled correctly. The report refused to fill the gaps with the industry's ready-made knowledge. It did not invent a player, did not assign an undated ranking, did not sketch an Asian landscape from memory. Disciplined behaviour — and behaviour the market does not pay for. The market pays for a filled table.

The lesson sits here: a wrong report gets corrected by reality, while an empty report never gets corrected, because nobody can see the emptiness.

I learned this by paying twice. In the 2026 season I listened to xG whisper, and I stopped trusting my eyes — then xG itself taught me the second lesson. Leipzig generated 2.8 expected goals against Bayern, I declared Leipzig would win, and they lost 0-2 after missing three clear chances. The numbers were not empty. They were missing one variable: the psychology of young players under away pressure. That gap became visible because there was a result to compare against.

The 2026 World Cup was different. My 57-variable model put Germany in the semi-finals. Germany lost 0-2 to South Korea and left at the group stage. Germany did not die of a lack of talent; they died believing the script was fate. I sat for four days re-watching all 64 matches, counting pressing sequences, timing transition phases. But to do that I needed footage, records, raw data. Had the extraction pipeline returned an empty frame and I had filed anyway, I would have had nothing to re-watch — and no idea that I needed to.

The 2026 ghost season gave me the mirror image. Across 112 matches without spectators in Germany, home teams won 27 percent instead of the familiar 42 percent, and home advantage evaporated by 38 percent. Those figures exist only because there were 112 rows to count. Without them I would have had a vague feeling that empty stadiums change things — and a vague feeling does not beat a bookmaker.

When the risk sits with the analyst

In that nine-dimension frame, the risk section normally lists six categories: competitive, selection, generational gap, governance and public opinion, systemic, opponent. All six were blank. But one row was fully populated, and it belonged to no match: analytical risk — an output generated from zero input could be mistaken for a substantive finding. Level: high. Likelihood: high if not flagged. Impact: high, because it spreads into an entire chain of decisions downstream.

This is where I have to speak plainly to people in the trade. We are used to checking models. We backtest, we recalibrate, we compare forecast error against results. We almost never check the pipeline feeding the model. A good model fed empty data will not raise an alarm — it will analyse the emptiness very seriously. And because it is serious, readers believe it.

Every betting line is a confession nobody hears. An empty analysis table is worse: a confession never spoken, with only a signature attached.

There is a technical consequence few notice. This kind of failure is rarely isolated. If one document in a batch returns an empty frame, sibling documents from the same run likely carry the same defect, because they pass through the same parser, the same character recognition, the same configuration. Check one and you check the batch; skip one and you skip the batch.

And there is a deeper layer, the one I consider most worrying in the long run. Sports analysis is shifting from humans reading matches to machines reading matches, and then humans reading machine reports. The distance between the pitch and the decision-maker keeps growing. Once that distance is wide enough, an empty frame can pass through five review layers with nobody opening the source. I do not believe in hunches. But I believe in numbers that cannot be explained — and I also believe that the absence of a number is more suspicious than any anomaly.

A match is a chapter, a season is a scripture, I only read and recite. Before reciting, I need to be sure I am holding a book, not an empty hard cover.

The gate must close before, not after

The minimum threshold for an analysis to be allowed to run is modest: a defined title, a named source, a publication date, a clear article type, and at least five atomic information points carrying entities and timestamps. If any is missing, the system must stop and shout, not print nine sections and quietly send them out.

The cost of such a gate is close to zero. The cost of not having it, I have already paid once on someone else's behalf: four days re-watching 64 matches to discover the problem sat with the assumption of the person holding the model, not with the model. In table tennis, where ranking points roll over every week and one gap in the calendar can distort an entire selection cycle, the price of an empty frame is higher still.

When the stands are empty, I hear the ball breathe. Data is at its most naked then. But hand me an empty stand and call it data, and I will hear nothing at all — and my silence does not mean the match is over.

The next competitive edge in this industry will not come from a smarter model. It will come from a switch willing to say "not enough data to speak" — and a culture brave enough to pay the person who presses it.