Zero Information Points: The Cricket Analysis Report That Came Back Empty

Late on Monday night I opened a file on my screen. The title read "Stage-2 Deep Analysis — Cricket Domain." Inside were eight sections, each with a table, each table with rows, each row with a cell. At first glance it looked like a complete report. Then I read the second column. Everywhere, the same sentence: "N/A — insufficient information."

Twenty-five years of logging the things in cricket that nobody logs, and I have seen plenty of blank sheets — but they were never actually blank. The 61 DRS referrals of the 2026-17 home season, each tagged for trigger point, projection margin and the umpire's-call threshold. The 20 on-field reviews of the 2026 World Cup, sorted into factual and subjective interventions. The 60 IPL matches of the 2026 UAE season, 78 player reviews, and the piped-in crowd noise that quietly rewrote the arithmetic of review success. In all of them there was at least one fact I could put a finger on.

This time there was no place to put the finger. And writing that emptiness down properly is the most useful thing I will do this week.

Context: in cricket, format is not a label — it is the unit of measurement

The report's own audit table says it plainly. Article title unavailable. Source unavailable. Article type "Unclassified." Domain label "cricket_world," which is non-standard; the specified label was "Cricket." The list of information points is empty. The entities field reads "identify from the information points above," and there are no information points above. Time sensitivity: "not assessed in Stage 1." Source quality: "judge from the source fields of the information points," and no source field exists.

This is not an empty article. It is an empty pipeline. A parse error, an empty payload, or a scraping block — one of the three, and the report itself rates that inference High confidence.

Format comes first in cricket because format is the unit of measurement. A Test average and a T20 strike rate do not sit on the same scale. Comparing an ODI economy rate with a Test economy rate is a category error. Until there is a format tag, not one claim about a player, a team or a league can stand, because the scale of the claim has not been fixed. The report did exactly this: it refused to assign a format, and called that refusal a refusal to guess.

There is a ledger nobody keeps, so I keep it. The ledger least kept in cricket analytics is the ledger of failure. Successful reports get published; empty reports get deleted. And the ones that get deleted break again in the same place next time.

Core: eight dimensions, zero evidence, and a chain that snapped at the first link

The report's architecture is a chain of custody. Under every conclusion there is a mandatory line: "→ Evidence: [information-point number]." Zero information points means every cell across all eight dimensions must stay N/A. That is what the report did — from match format to broadcast-rights value, from a player's average to the ICC rankings, everywhere "insufficient information."

The tables that look full are not findings; they are framework completeness. In the governance section, the 2026 Cronje affair, the 2026 Pakistan spot-fixing case, the 2026 IPL scandal — all those rows exist, and the report says explicitly that they are retained for reference only and are not claims about this article. That is the honesty cricket coverage lacks most: filling a template and filling a fact file are two different jobs.

The only surviving token is "cricket." That token is consistent with any cricket article — a match report, an auction story, board politics, anything. Its analytical weight is zero.

I went back to the tape expecting to be wrong. I expected the payload to contain something and my first suspicion to be mistaken. The payload contained nothing. And when I traced where the failure occurred, the answer was cleaner than before — the mistake is never the moment; it is the twenty seconds before it. Here the failure happened before entity recognition. The system did not even get far enough to say which team or which player was being discussed. The break was not in reading the article; it was in reaching it.

The report rates one risk High that has nothing to do with sport: hallucination contamination. An empty evidence base plus a fully populated template are precisely the two conditions that make fabricated facts most likely. The mechanism is simple: a downstream model sees the shape of a table, and shape demands that empty cells be filled. Cricket's own example is the scorecard — headers present, numbers absent, and commentary filling the gap with "approximately." Hence the recommendation to gate the output at STATUS: INSUFFICIENT_INPUT and stop it entering any further summarisation.

Five signals are set for tracking. Re-extraction success: at least 5 information points and at least one named entity. Source metadata recovery: outlet, publication date, byline. Format identification. Entity recovery. Article-type classification, meaning movement off "Unclassified." If not one of these triggers, all eight dimensions stay shut.

Four concrete steps are proposed for repair: raw-HTML or PDF fallback, handling JavaScript rendering, paywall detection, and per-stage extraction-count logging. The cheapest and most effective is the extraction-count assertion — a rule that rejects any Stage-1 output carrying fewer than 3 information points.

There is one rule-audit here and it has a visible consequence. The consequence is that an unverified cricket item must never circulate as fact. In transfer-window weather this matters more, because rumour volume peaks exactly when the underlying evidence is thinnest. The filter we apply to transfer rumours — rank by evidence, follow the money, read the contract and the agent's movement — applies to cricket data too. The difference is that here you follow the process, not the money.

Contrarian: the room does not print a blank table

The report's most uncomfortable truth is not that the pipeline broke. It is that a blank table cannot be published. Broadcast talk runs on intent-language: "that's not cricket" is always a faster sentence than quoting the clause. In the same way, "the pipeline had nothing" is always a slower sentence than "sources indicate." The rulebook said one thing; the room heard another. Here the rulebook was the evidence standard, and the room was the publishing cycle.

The room's pressure is understandable. Board decisions move the world calendar, so the person who can explain the board becomes relevant fastest. The person who can say "I don't know" becomes relevant slowest. That is proximity drift, and it is this beat's oldest trap.

But the inverse argument is just as true, and it is the real contrarian angle: the null result is the only category in sports analysis that improves with repetition and degrades with attention. Publishing it costs a day of relevance and buys a year of credibility. The ledger is quarterly and the credibility is annual — which is exactly why the incentive tilts the wrong way.

Frame by frame, the story I had been told — the story of a "data-driven cricket desk" — quietly fell apart. A desk that publishes through empty data is not data-driven; it merely speaks the vocabulary of data.

Takeaway: gate it next time, don't hope

The correct use of this report is to not forward it. Hold it at STATUS: INSUFFICIENT_INPUT and re-run the pipeline. What should ship in the next iteration is not a hope but an assertion — that the pipeline raises its own alarm when fewer than 3 information points come back.

The question is not what was lost this time. The question is: if a pipeline cannot announce its own failure, what else is it announcing quietly — and who is checking? The cost of an empty table is one day. The cost of a full, handsome, fabricated table is far higher, because nobody audits it.