Empty Sheet, Deep Ditch: Why 'N/A' Is Not an All-Clear in Cricket Analytics

Tuesday afternoon, Delhi. On my desk lay an open file — seven analytical dimensions, twelve tables, rows in every table, a slot in every row, and in every slot the same four characters: N/A. The file was titled 'Stage-2 Deep Analysis — Cricket Domain'. Seven pages. Not one score, not one name, not one date across all seven. And the strangest part — not a single spelling error. The format was intact. Headings where headings go, tables where tables go, a risk matrix where a risk matrix goes. Only the substance was missing.

Fifteen years ago I would have called such a file broken, closed it, and ordered it again the next morning. Now I stop and read it, because I know this is the kind of file that does the most damage. Bad data makes noise. Empty data stays silent, and silence gets read at the desk as 'all clear'.

What happened inside that file was not a cricket event but a pipeline accident in cricket analysis. Stage-1's job is to pull facts from a source — title, source, format, three or four discrete information points, names. This time Stage-1 returned zero information points. And Stage-2, which had to rule on seven dimensions, had no table to sit at.

I have learned one discipline across my whole career, and it is this: in cricket the fact comes first, the story second. The day the story arrives first, the next morning's paper is carrying somebody else's story.

Cricket desks have done the same job for seventy years: a scorecard arrives, somebody looks at it, and a sentence comes out of that look. In the twentieth century the scorecard was paper and the man looking at it was singular — the reporter in the commentary box. In the twenty-first century three more layers were welded under the card: ball-tracking, biomechanics load files, and auction valuation models. A single T20 match now yields close to twelve thousand data points, and maybe forty of them end up in any given sentence.

That architecture has three stages, each with a separate responsibility. Stage-1 extracts. Stage-2 distributes the extracted facts across dimensions — format, player, team, league, governance, risk, narrative, industry. Stage-3 gives the analysis language. Every stage trusts the one below it, and that trust is the weakest joint in the whole structure.

I wrote my first workload ledger by hand in 2026, at a Delhi sports-science lab, logging spell lengths for an IPL franchise. That year Bhuvneshwar Kumar took 26 wickets to win the Purple Cap. I pulled the delivery data and found his death-over volume had risen 41 per cent from the 2026 season. No desk wanted the raw spreadsheet, so I turned it into nine hundred words on yorker volume and injury risk. A digital desk ran it in August.

That day I made a rule I have not broken since: I kept a workload ledger; the byline found me inside it. Not once since have I filed a tactical claim without a measured input.

Back to the empty file. Three things in it stung most.

First, the information-point list was empty. This is not a minor gap. The information point is the unit everything else stands on. If the list holds zero entries, all seven Stage-2 dimensions collapse on their own, however beautiful their tables look. A conclusion drawn from zero input is not a conclusion — it is a guess that has put on good clothes.

Second, the article type was unclassified. This is where the largest error hides. If you do not know whether the source is a match report, an auction story, a board decision, or an obituary, you will default to the most probable assumption, and that assumption will be 'match'. I have watched this be the most common desk error: reading every story as a match, because a match is the easiest thing to read.

Third, the domain label read cricket_world rather than the canonical Cricket. That small mismatch is the biggest clue. In cricket we would call it schema drift — the machine that extracted the data was not running on fully correct settings. And a machine with the wrong settings does not announce its error. It quietly returns an empty file, and an empty file looks exactly like a complete one.

Here I think of the Russia summer of 2026. Through that World Cup I watched geometry rather than France — built a twenty-zone grid, mapped where sides pressed and where space opened between the lines. In July I wrote about what a press can teach a ring field, arguing that a T20 inner ring is a compactness problem. Two months later India lost the Test series in England 4-1. I took the same grid and showed that India's slip cordon stood half a metre too deep for the seam movement at Edgbaston. The twenty-zone grid was not a map; it was an argument.

But that same experience taught me the second thing: format must be settled before the grid. A Test average, an ODI strike rate and a T20 economy rate are not comparable — and an analyst who picks a metric without knowing the format is quoting a bowling average without looking at the pitch.

That file contained no format. Which did not mean format was irrelevant; it meant the file had no licence to record any tactical conclusion at all. Fifteen hundred clean words saying 'insufficient information' everywhere was, in fact, a refusal — and in my reading, the correct refusal.

Let me be plain here, because desks read it backwards. When an analyst says 'I don't have the data', that should not be read as 'he isn't working'. It should be read as 'he declined to invent numbers'. Inventing numbers is easy. A strike rate, an economy, a ranking — all three are so familiar that a reader cannot disprove them. The truth is that in this industry wrong numbers live very long lives.

One morning at the Kotla nets I watched a fast bowler — I will not name him, because the name is not required by my argument — into his third spell. The ball left his hand, the seam held straight for half its journey, and in the last two feet before the file pitch it drifted slightly. The sound off the crease was a little different each time. The coach said something, he laughed, and bowled the same ball again. A man on the bench got up and walked off. I sat there twenty minutes and had not one number to write. When I came back my first entry read: 'third spell, sound changing, no laughter.' That entry never earned a column in a table, and it is the entry that told me his yorker data was lying that day.

I repeat this because my largest professional test in recent years has been which comes first, the grid or the cricket. The grid has never disappointed me enough to abandon, but a grid can never see where the seam is pointing. The empty file is a portrait of that gap: twelve tables and not one ball.

This is not a digression — the gap is the file's real lesson.

Now look at its risk matrix. Six rows: sporting, personnel, commercial, rules and integrity, public opinion, systemic. All six rated 'insufficient information'. That matrix looks useless and is not. It is itself a finding. Do not read a null as clean. 'Unassessed' does not mean 'no risk found'; it means 'no search was run'.

Cricket needs this distinction because it already applies it elsewhere. When no DRS outcome is logged for a match, that does not mean the umpiring was spotless. When a bowler's injury history sits blank in a file, that does not mean he is fit. The most useful entry in my workload ledger was never one I filled; it was the one where I wrote that spell length had not been recorded.

The same building in Delhi houses a Hindi desk, and on this file that desk raised a finger before I did. Their point: if seven dimensions contain not one name, it isn't analysis, it's a form. I would extend it — the form isn't even right by accident. It is the form of a machine that ran and brought nothing back.

The most useful thing the file produced was a minimum list for the next Stage-1 run. Five items. Title, source and publication date — without them neither time sensitivity nor source tier can be fixed. Format — Test, ODI, T20, The Hundred or league; the door without which no metric is comparable. At least three discrete information points, each a separate fact, not a summary. At least one named entity — team, player, league or board. And a source-quality tier: official board announcement, named journalist, general media, or unattributed aggregator.

The cheapest upgrade on that list is three items — source, time sensitivity, three information points. Fill only those and roughly sixty per cent of the seven-dimension framework becomes executable. I learned this in February 2026, during the IPL mega auction in Bengaluru, when a Bengaluru-based data engineer and I built a cost-per-win-above-replacement model across more than six hundred players. Ishan Kishan's ₹15.25 crore purchase came out as a mispricing; Deepak Chahar's ₹14 crore looked roughly fair. Nine months later I covered the Qatar World Cup as a calendar problem — thirty-two teams, eight stadiums, twenty-nine days — and wrote about what that compression would do once the field expanded to forty-eight.

That model gave me a habit I have never dropped: read the auction as a market, not as theatre. The transfer market is a spreadsheet pretending to be a soap opera. And that habit taught me to count the rows of a file before reading its cover page.

The empty file's biggest discovery was not that the data was missing. It was that the file was built well enough for anyone to pass it on as complete. The danger is not that a machine says something wrong; the danger is that it says nothing in the correct format and somebody reads that as fine. Had this file entered an automated chain — an alerting system, an editorial brief, a trading-style signal — it would have arrived without a single error and become the most trusted document in the room. An error message at least states its case. An empty but tidy report does not.

Which is why, before rejecting it, it should carry a label: invalid input, insufficient evidence. The label is not part of the report; it is a condition placed in front of it.

Now the most inverted aspect of the whole episode.

Most of my trade would call this file a failure. I call it half a success. A success because a system under pressure to fill seven dimensions did not fill them. It could have added forty or fifty ratings, a few invented economies and a couple of phantom auction prices and produced a very readable piece, and nobody would have known. It didn't. In an industry where a new 'take' becomes a headline every hour, refusing is a discipline.

I do not trust a narrative until it survives a load calculation. This file honoured that rule from the opposite direction: no load, therefore no narrative.

Second inversion: we read zero as fear, when in cricket zero is often the most informative figure on the sheet. When a bowler has no wickets in a particular zone, the zero is telling you the captain is not stationing a fielder there. When India's slip catching in a series is zero, the question it raises is about cordon depth, not about hands. Zero is a pointer, not a vacancy.

There is a third, less comfortable angle. cricket_world versus Cricket — a label, five minutes for a technician to fix. In my experience that label was the most honest thing in the entire run. A system that misspells its own name has probably misspelt other things, and it did not complain about any of them. When the 2026 IPL was played in empty UAE stadiums — sixty matches, Dubai, Abu Dhabi and Sharjah, home advantage set to zero and dew the only honest variable — I learned the same lesson. When the crowd vanished, the experiment became uncomfortably honest. In Sharjah I logged second-innings scoring rates and found chasing sides gained roughly 0.8 runs an over after the sixteenth over, and that the gain belonged to dew, not to the crowd.

That experience gave me a habit: every tactical piece now opens with conditions — dew, humidity, boundary dimensions, toss — and names the team later. I deleted 'home advantage' as a default explanation and replaced it with something measurable, which made my previews much harder to argue with. The empty file was delivering the same instruction: conditions first, story second. It had no conditions at all.

So the next time a seven-dimension analysis report lands in front of you, count three things before reading a word. The number of information points. The name of the format. The date on the source. If any one of the three is blank, turning the remaining pages buys you nothing — you have already passed the most important part of the file.

My next test is exactly this. The next time Stage-1 runs, I will not read page one. I will count the rows first. If the count is zero, I will close the file and send it back on the spot, and I will not give some elegant sentence the chance to tell me the momentum had shifted.