Zero Information Points, Full Template: The Silent Failure Inside a Cricket Data Pipeline

The report arrived with eight sections. Each section had a table, each table had four or five rows, and every row ended in the same two characters — N/A. Title: N/A. Source: N/A. The information-points field was entirely blank. And the entity column carried a sentence: "identify from the information points above." There were no information points above. This was not a report; it was scaffolding dressed as one. The dangerous part was that it looked technically clean. Six risk categories, and not one risk recorded in any of them. Which means a subject that was never analysed got stamped "no risks found." In cricket we have a name for that — a match report claiming no wickets fell, because nobody could find the scorecard.

I logged 538 deliveries by hand before I learned to distrust totals. March 2026, Johannesburg: Australia made 359/2 in the World Cup final, Ricky Ponting unbeaten on 140, India folded for 234, a 125-run defeat. Nobody sold ball-by-ball data in India then, so I spent eleven nights typing every delivery into a spreadsheet on a borrowed laptop. The ledger showed India's middle order had played 71 dot balls between overs 12 and 30. My editor ran the number as a sidebar and it out-read the main report. Since then the rule is fixed: before the number there is a ball, and someone has to count it. The other half of that rule is less quoted — where no ball was counted, no conclusion follows either. That is exactly what happened here. We did not skip the counting. We performed it.

The first signal is the cleverest. The information-points list is empty, yet the template looks fully populated. All eight analytical dimensions — format, player technique, team landscape, league economics, governance, risk, public narrative, industry transmission — are present, each with tables, each answering N/A. This is a shaped-but-empty output: the frame is convincing enough that any downstream reader can mistake it for a substantive "nothing to flag" finding. That, not the blankness, is the real hazard.

The second signal is blunter. The entity field instructs the analyst to identify entities from information points that do not exist. The instruction cancels itself. When a template refers upward to a payload that is missing, the break sits at the extraction stage, not the analysis logic. That is a diagnostic signature, and it is localised — which is good news, because localised failures can be re-run.

The third signal comes from the taxonomy. The domain label reads "cricket_world" where "Cricket" was expected. It is a tag, not a fact. A mis-mapped label can never be promoted into team-level evidence — but it does say that something went wrong at the classification step, and that same something may well explain the empty payload. Treating a wrong label and an empty payload as two unrelated incidents would be a mistake.

The fourth signal is the most expensive. The format field is absent. In cricket analytics, format is a hard gate. Comparing a Test batting average to a T20 death-overs strike rate is like forecasting a marathon from a sprinter's 100-metre split. Without a format tag, every metric comparison is technically invalid — so writing N/A was the correct call, and it was made. That is honesty, but honesty is not a fix. A model that answers N/A to everything is not a model; it is a blank form.

Here is the counter-angle, filed against myself. The easy reaction is to blame the pipeline and move on. But absence of evidence is not evidence of compliance, and that cuts both ways. The moment we say "no risks found," we also admit no search was properly run. On the other hand, it is entirely possible the source was genuinely thin — a photo caption or a one-line item does not yield eight dimensions of analysis, and in that case the fault was in the expectation, not the fetch. Three hours into the tape, coincidence stops being a defensible position; three minutes in, it never was one. So before assigning blame, re-extract: rendered HTML, PDF parse, legitimate paywall access. If it comes back empty again, the pipeline is at fault. If it fills in, the impatience was ours.

The forward signal rests on three things. First, a preflight validator that blocks any Stage-1 payload carrying zero information points, an empty title, and an unexecuted entity instruction. Second, format as a mandatory field — Stage 2 should refuse to start without it. Third, a source-quality grade attached to every information point at extraction time, because an analysis of ungraded provenance is never clean however clean it looks. And above all, one label: INSUFFICIENT_INPUT, so this record cannot climb onto a dashboard as "low risk." Every scorecard is a summary of decisions made by people you never met — and an empty scorecard is still a scorecard. The question is not what it says. The question is who counted it.