Eight Columns, Every Cell N/A: Auditing Cricket Analytics' Empty Set

The template had eight columns — format, player, team, ranking, league, governance, risk, narrative. Every cell in every column carried the same entry: N/A — insufficient information, cannot assess. The file landed last week, mid-tournament, in the same week every channel was pushing a fresh "pressure index" every third minute with an empty data column sitting under each one. My first instinct was to send the file back. My second was to publish it exactly as it arrived — with a date, a timestamp, and the auditor's name on it.

An empty cell does not frighten an editor. A filled cell does — the one filled by someone who did not know what he was filling. This season I have counted at least four on-screen graphics whose "home-ground advantage" figure was drawn from a series in which two matches were washed out and nobody had separated them out with a DLS adjustment.

Every scorecard is a summary of decisions made by people you never met. March 2026, Johannesburg — Australia made 359/2, India folded for 234, a 125-run defeat. I spent the eleven nights after that match on a borrowed laptop, re-logging every ball, because no vendor in India sold ball-by-ball data then. After 538 deliveries counted by hand, one number surfaced that the scorecard did not carry: between overs 12 and 30, India's middle order played 71 dot balls. My editor ran it as a sidebar; the sidebar out-read the main report.

Since that day I have held one condition — before the number there is a ball, and someone has to count it. The ball precedes the number, never follows it.

Now apply that question to this empty file. What is this N/A? It is not a conclusion. It is a location. It tells you the precise point at which the data pipeline broke. Stage-1 extraction produced no headline, no source, no information points, no entities. Which means the fault is not in the Stage-2 analysis; it is in the Stage-1 reader. An empty set is also a result — it hands you the address of the fault, and all it asks in return is the nerve to print that address. The analyst who ignores the address and fills all eight columns is not analysing; he is servicing a template.

Last year, working on a franchise's auction simulator, I saw this disease on my own skin. The model was recommending retention of a batsman on "recent form." When I pulled the feed apart, that recent form was 41 balls across three innings — one run-out, one life off a no-ball. A form rating built on 41 balls had settled a retention value in crores. The problem was not the model; the problem was the empty cell somebody quietly filled.

There is a third lesson here, and it is the least spoken. A null needs a date. "Data unavailable" is an unfinished sentence; the complete one reads — "as of 13 August 2026, from this source, for this format, no data is available." Without a date the null gets refilled, because next week somebody else arrives and drops his memory into the same empty cell.

One more thing is visible in this season's noise, what I call transmission leakage. A null at Stage-1 reaches Stage-4 and becomes a graphic. I have watched this happen from inside: last season, in a Mumbai commentary box, the producer turned to me during a live over-break and said, "Sir, we can't find a number for this player, just say a line." I told him I had no number. He said, "No problem, speak from experience." That is the exact spot where data journalism dies — experience is quietly installed as a substitute for data in front of an empty cell.

I did not say the line. But that day I understood the problem is not the commentary box alone. The data reaching the viewer's second screen — the Hindi, Tamil, Telugu feeds — runs a step behind even the English feed, and demand for correction is lower there because fewer questions are asked. The desks that were already doing verification work — the print sports pages, the regional-language sub-editors who ask for the source name before a number goes in type — are the strictest auditors in the building today. I learn from them. I do not call them old-fashioned.

Now turn it over. The easy conclusion is that analysis did not happen because data was absent. That is half true and half an excuse. The real trouble is that our analytical apparatus is not built to falsify itself in the face of missing data. Eight dimensions, three or four metrics inside each, and nowhere a rule for what an output ranks as when the inputs do not exist. In such a system, filling in N/A becomes the path of least resistance. I remember the same thing happening with my pressure index — the index said five, the dressing room refused to believe it, and I spent a week working out why nobody believed it. The answer was that my input was broken: I was treating a dot ball as pressure when, for that batsman, the dot ball was setting his timing. The model was not right; my assumption was wrong.

So my verdict on this empty file is plain. An analysis that cannot publish its own null state cannot ask anyone to trust any number it does publish. Instead of filling the template, the correct move is what happened here: mark N/A on every dimension, attach the reason to each N/A, and close with a clean instruction — re-run Stage-1, populate the headline, source, information points and time sensitivity, then come back. That is not a weak output. It is an honest one, and an honest output can be re-run.

My forward signal for the coming cycle is this. Every dashboard should carry a null register alongside it — which cell is empty, since when, which source left it empty, and whose job it is to fill it. On the day a channel's pressure index is broadcast with its null register attached, that index will become arguable for the first time.

Before the number there is a ball. And the ball's absence also has a date on it.