The Ledger Came Back Empty: Football Analytics and the Real Test of Information Integrity

The file opened at two in the morning. Nine analytical dimensions, each with a complete table, and every cell carrying the same sentence: insufficient information. Zero information points. No club identified, no player, no competition. Source quality left at 'not assessed', time sensitivity likewise. A football analysis document with no football inside it.

The first instinct is easy — fill the cells. Chelsea, Benfica, £106.8 million, a 4-2-3-1; those names sit within reach, and the empty tables are already built. I sat with my hand on the cursor for twenty minutes. Then I closed the file.

Returning a zero is also an output, and in football analysis it is the least practised skill. The error does not look like a failure — it looks like unfinished work. Nobody asks which information point a claim came from. Everybody asks why the table wasn't completed.

Between Stage-1 and Stage-2 there is a contract, and it is about evidence, not football. Stage-1 pulls information points out of the source article — each one discrete, quotable, citable. Every conclusion in Stage-2 then carries a mandatory citation: evidence, information point 3. Structurally this is a blockchain, not metaphorically. A block cannot stand without the hash of the block before it; a conclusion here cannot stand without the reference to its information point. With zero information points the chain breaks, and when the chain breaks a conclusion stops being a conclusion and becomes a guess.

In 2026 I hand-coded twenty-four matches — the run-in of the 2026-17 Premier League, 1,400 possession sequences from television feeds, one spreadsheet. Chelsea's 3-4-3 worked because of Cesc Fàbregas's lateral passing lanes, not because of N'Golo Kanté's ball-winning, and that claim survived only because every pass had a timestamp. The piece drew 90,000 reads. The real asset was the spreadsheet, not the read count. That file taught me that a claim weighs exactly as much as its citation, never more.

At Russia 2026 I wrote 64 tactical match reports in 32 days. The final was France 4-2 Croatia — most coverage praised Croatia's midfield while I mapped how France's 4-2-3-1 behaved out of possession, how Antoine Griezmann vacated the No. 10 channel so Paul Pogba and Blaise Matuidi could press Croatia's first line. I counted 14 French recoveries inside Croatia's half before the 60th minute. Sixty-four reports in thirty-two days taught me that a vacancy is also a structure: look at the shape before you explain it with names.

In May and June 2026 the Bundesliga returned to empty stadiums. The freelance budgets that had funded me collapsed inside three weeks, and one editor returned my study saying he needed 'a more authoritative voice'. I did not argue; I coded all 90 matches. Home win rate fell from 43.2 percent to 32.1 percent, away sides' high-press success rose six percentage points. The crowd was worth 0.3 goals, and the algorithm has never let me forget it — which is why every breakdown now carries a permanent Environment block: crowd noise, heat, altitude, pitch width. A camera feed that drops is environment data too.

Back to that file. An empty payload does not distinguish between two very different things, and the distinction is the whole point: did the source article contain no football, or did the pipeline lose the football? The artefacts lean towards the second. A genuine match report or tactical preview almost always yields at least one formation string at the extraction stage. Here there is not even that. The domain label reads 'football' and everything else sits at default. That is the shape of a payload, not the content of an article.

Four risks follow, and none of them is a football risk. First, if this document travels downstream to an editorial desk or a scouting report, zero information will be consumed as signal. Second, the pull to fill empty cells; the template itself is the temptation, because plausible names are easy to insert. Third, repeat failure on the same source types — paywalled pages, image-only PDFs, live-blog HTML. Fourth, with no source named, source-tier grading is impossible, and without source tier the rules and narrative dimensions cannot run at all.

The fix is unglamorous and gate-based. Title, outlet, publication date, at least three information points, at least one named entity — if those five conditions fail, Stage-2 should not run. My own press-trigger file has followed the same rule since 2026: when a feed drops, I mark the gap as a gap. I do not interpolate.

That file paid out at Qatar 2026, where I coded all seven Morocco matches. Walid Regragui's 4-1-4-1 collapsed into a 5-4-1 against Spain and Portugal; five games, one goal conceded, and it was an own goal. I recognised the shape early because the file existed, not afterwards. On 31 January 2026 I filed the £106.8 million Enzo Fernández piece from Benfica to Chelsea within nine hours, because the template on his seven Qatar matches was already written. A template is not a guess — a template is where verification is scheduled before the filling begins.

The contrarian angle is simple and slightly uncomfortable. This document's real message is not failure but success. The most dangerous output is not an empty table — it is a full table that looks credible, where every cell has a name and no cell has a citation. Football media publishes that table daily: momentum, passion, the DNA of the club. None of it can be tied to an information point, yet it survives because nobody counts the error. I do not watch football for beauty; I watch for the moment the system lies — and today's system is lying about its own inputs.

My own structural weakness is visible here too. For nine years I have been the entire pipeline: coder, diagrammer, editor, verifier. If a match goes missing from my archive, there is no second person to catch it. The document I opened that night belonged to a large organisation; my gaps are smaller, the structure is the same. Having the nerve to declare a null result and having a mechanism to detect one are different skills, and the second is the one I have not fully acquired.

So the verification chain has to leave my own hands: re-run Stage-1 on the same source, cross-check my coding against someone else's, carry source-tier metadata forward. These are not rules but tests — and the advantage of a test is that it can be proven wrong.

The next time an analysis document lands, my first question will not be tactical. It will be: how many information points does this contain, and which source is each one chained to? If the answer is zero, the second question is worse — is the zero in the source, or in my own pipeline? The day that question has a clean answer, football analysis will balance its books for the first time.