An Empty Dataset Is Not an Empty Truth: The Missing-Data Trap in Football Analysis

Last season I opened my laptop in a European press box and saw five small words in the corner of the screen: zero rows returned. Five minutes to kick-off. The stand was roaring, and the reporter one row over had already filed his preview stating, flatly, that this team presses in the league's third-best high block. My feed had not delivered a single number from that source. Three minutes later the line-ups went up on the big screen; my laptop still showed nothing. Silence comes in two kinds — one honest, one with room to invent. The system that said 'I received nothing' was honest. The system that found an empty cell and filled it itself was a forger. The real trap in football analysis is not a wrong number; it is the empty cell. A wrong number shouts and leaves a trail. An empty cell leaves nothing at all.

In twenty-nine years of watching this sport I have looked at it through two doors — the touchline door and the balance-sheet door. In August 2026, working a night shift in London, I argued that Neymar's €222m move to PSG was not inflation but the purchase of the last unclaimed global football brand. Three national podcasters mocked me on air that week. The strength of that piece was never my hunch; it was a balance sheet with a visible commercial-revenue line. Emotion gets written easily; a claim only survives on information another person can go and check. That is now my first condition. Four months later I quit my job as an analyst, because I had realised the scoreline was not my language — the financial pillar was.

Yet that condition has a blind spot I did not see early enough. Which is more dangerous: absent information, or bad information? Instinct says bad information. Experience says the opposite. Club analytics departments now buy xG models, subscribe to PPDA feeds, and scan FFP or Premier League PSR filings to unpick transfer amortisation and sell-on clauses. The media cycle has compressed — anything not published within six hours of full-time effectively does not exist. At that speed, the one thing systems see least clearly is feed failure. Suppose an article enters the pipeline with no headline, no source, no club or player named, no date, and an information-point count of zero. Whatever comes out the other end is not analysis. It is arranged guesswork.

The danger there is structural. An empty input does not stop a model; an empty input only forces it to fill the cells from memory. In football, that memory bank is dangerously rich — familiar clubs, familiar trends, familiar coaching vocabulary. So a system handed zero facts can write in exactly the register it would have used if the facts had been real. The difference is invisible to the reader and often to the editor, because the sentences are smooth. Smoothness is the disguise.

The easiest place to catch a first-order gap is where fields depend on each other. When the instruction reads 'identify entities from the information points above' while the information-point list is empty, the chain cannot stand on its own legs. That circular dependency is nothing new in football data. Every match report's source line should carry a club statement, a league filing, or the name of a reliable journalist; in many reports that line is the emptiest of all — and that is precisely where the biggest claims fly in from.

So the hierarchy of sources cannot be left blank either. A club's own statement and a league filing sit on the top shelf; FIFA or confederation decisions come next, then investigative journalists, then wire agencies, and at the very bottom the aggregator pages that copy headlines and manufacture numbers. When the source cell is empty, that hierarchy collapses — there is no way to weigh one claim against another. To a reader it looks tidy. To an analyst it is thick darkness.

The same error recurs in risk rating. When no club, player or contract data is available, writing 'low risk' is the easy move. Absence of information is not absence of risk — those are two entirely different sentences. If a player's injury history is missing from the feed, that does not mean he is permanently fit; it means we do not know. If an owner's signals of pulling investment out are not in the dataset, that is not evidence of stability; it is ignorance. Marking an empty cell as low risk is another form of false information, because the decision-maker then advances with a false sense of safety.

In the summer of 2026 I spent 26 days in Russia and watched eleven matches live. Afterwards I wrote that the World Cup had become a set-piece sport. In England's 6-1 win over Panama in Nizhny Novgorod, five of the six goals came from dead balls; England finished the tournament with twelve goals, nine from set plays. That claim survived because my notebook was full — eleven matches, twenty-second clips filmed from the stands, my own written record. Compare that with 2026. On 16 May the Bundesliga returned behind closed doors, and in the first full round home sides won only two of nine. Within 72 hours I filed a piece arguing home advantage was never about the crowd, that the effect was permanent, and that clubs should reprice season tickets. By early 2026 home win rates had reverted almost exactly to pre-pandemic levels. Where the sample is small, the sample is itself a kind of empty cell. My correction video outperformed the original claim. Since that day, every prediction I publish carries a timestamp and an expiry date, plus a short line on how it breaks.

Even so, the contagion from an empty input does not stop, because the damage is created at step one, not at the last step. Before asking what decision a club should take from a feed, we should ask how complete the feed is. I now write three simple tests on the first page of my notebook. One, a named entity: at least one club, player, coach or competition. Two, a concrete claim: a transfer, a result, a contract renewal, a sanction, or a rule change. Three, a fixed point in time: nothing is verifiable without a date. If two of those three are missing, starting the analysis means inventing a story. Installing that gate costs seconds of delay, and it closes the single channel through which false certainty reaches readers under the name of football.

One simple habit changes a great deal — dates. 'Last week' in journalism means verification is impossible. 16 May 2026, 11 June 2026, 22 November 2026: those dates sit in my archive not for decoration but so any reader can check when I said what. A prediction without an expiry date is not a prediction; it is a mood. That is the core failure of work produced from empty input: no date, no source, no accountability.

Why the language of certainty is so dangerous became clear on 22 November 2026. Argentina lost 2-1 to Saudi Arabia in Lusail — their first World Cup defeat to an Asian nation. Six hours after full-time I wrote that Argentina would still win the tournament, resting the argument on a 36-match unbeaten run before the World Cup and on Messi's managed minutes. On 18 December they lifted the trophy, beating France on penalties. The piece passed 600,000 views. Note that the prediction did not come out of an empty cell; it came out of two checkable facts. Showing courage in an empty cell is easy; showing courage there and being right is a different distance entirely.

Where could I be wrong? Suppose the zero information points are not a feed failure at all — sometimes the source article really is empty: a stub behind a paywall, a deleted post, a source blocked by geography. Then zero is not a failure; zero is the correct answer, and all my gates become theatre. The second objection is more serious, and it is aimed at me. My brand is built on speed — filing within hours of full-time, because analysis that arrives after the conversation has started is worth less. A strict completeness gate could slow that down, and slowing down costs me the race. The third objection cuts deepest: a model that has learned to fill empty cells can also fabricate one entity and one claim in two seconds to clear the gate. Which means a gate that checks presence is worthless; the gate must check verifiability — is there a source, is there a date, and can somebody else check it?

So here is my public prediction, with an expiry. By 31 December 2026, at least one major European or British sports desk will publicly admit that a published analysis rested on a failed automated feed, and within the following eighteen months at least three top-tier publishers will print a machine-readable provenance line with every analytical piece. If no such correction appears publicly by 30 June 2026, I will withdraw the claim — in writing, with a date, as has been my rule since 2026. The question I leave with the reader: when did your favourite match analysis last show its sources?