The Empty Row: When the Tennis Record Returns N/A

It was two in the morning. I opened the column nobody would ever read — serve_zone_3. Forty-one of 3,142 rows were blank. In January 2026, in the media box at the SDAT Stadium in Nungambakkam, I hand-keyed every point of all 31 main-draw matches at the Chennai Open — serve direction into three zones, rally length, error type. The client was a European data aggregator. Not one Indian coach ever asked to see that file. I kept a personal copy of every sheet, and that copy became the first slab of an archive I still maintain.

What I found that night was not simply forty-one empty cells. I found that the skeleton of the sheet was entirely intact. The header row had every column name in its proper place, the cell borders were drawn, the formulas worked. And there were no numbers inside. The structure was alive; the content was dead. I found the error at 2 a.m., in a column nobody would ever read, and I had to wait until morning to fix it, because the source sheet sat on a vendor's server that was shutting down that week.

This essay is not about that incident. It is about the situation that repeated itself twenty-five years later.

Last week a document of analysis reached my desk. Its title was N/A. Its source was N/A. Its type was Unclassified. Its one-line summary was empty. The entire Information Points section was empty. The Entities field read 'identify from the information points above' — and there were no information points above. Source Quality was deferred to 'judge from the source field of the information points', when no information points existed. Time Sensitivity was never assessed.

The only living thing in the whole document was Domain Label: tennis. Everything else was template.

Context: the hand-written record and its rules of provenance

The first lesson I learned in data work was not about matches. It was about records. When I began as a reporter and sub-editor in 2026, the file was paper. In 2026 I hand-wrote 3,142 points from 31 Chennai Open matches, and with every point I recorded three things: where the serve went, how long the rally ran, and what kind of error produced the point. It was labour. A match runs past 200 points, and if a scoreboard logs a point every twelve seconds, breaking that point into three fields takes three times as long.

I keyed every point by hand, and the scoreboard never learned my name.

The purpose of that work was to build a record that anyone could later open and check to see where the number came from. We now call this provenance. With every point I also kept the match date, the round, the court surface, and the name of the person who wrote the row — in most cases, myself.

Once you have written a record by hand, one habit sets permanently: you keep the address of every number. Where a number is missing, you write the gap as a visible gap. You do not fill it with a guess.

From 2026 to 2026 I was performance analyst for the Indian Aces in the International Premier Tennis League. In the inaugural 2026 season, the Aces won the title. The league had its own machinery — a twenty-second shot clock, power points, a super shoot-out in place of the tie-break. No tour event produced data like it. I built a model of how each player's first-serve percentage moved under the clock. That work I am still proud of.

Three seasons with the Indian Aces, then a spreadsheet with no next row. By the end of 2026 the league had folded, and players were left chasing unpaid money. A franchise dies in the footnotes long before it dies in the press.

January 2026 brought the twenty-first and final Chennai Open. I held files on seventeen of the twenty-one editions. I coded the last one from the media box, where Roberto Bautista Agut beat a 20-year-old Daniil Medvedev in the final — 8 January 2026, 6-3, 6-4. That date, that score, those two names: this is the one concrete fact in this essay that I am writing straight out of my own record rather than lifting from a summary.

The following year the tournament moved to Pune as the Maharashtra Open, and my accreditation lapsed. They moved the tournament to Pune, and I moved the archive with it. I watched the 2026 edition on television, logging points off a broadcast feed with a delay I could not correct. It took me three months to accept that the access was gone for good.

In June 2026, between tennis contracts, I spent the Russia World Cup inside a Mumbai broadcaster's studio. I had built the xG and PPDA models behind the analysis segments for all sixty-four matches. The producers wanted a single number on screen. I insisted on sample size and a visible margin of error. Twice I was told my graphic was too complicated for air, and twice I simplified the language rather than the maths.

All of that context matters here, because to read last week's empty document you need to know what a filled record looks like. When you know what the full record looks like, the empty shell becomes recognisable.

Core: a line-by-line audit of an empty document

Now the real work. I open that document the way a ledger is opened. No complaint, no accusation. Only the question of which cell is empty, and for what reason.

First cell: the title. Value N/A. This is the largest clue, and the easiest to overlook. The title is the topmost cell of any document, and it is filled first. If the title is empty, the document was not built from a sentence or a name; it was built from a form. An empty title means everything upstream of it failed.

Second cell: the source. Value N/A. This wound is deeper. A lost source name means the thing that was to be analysed never reached a table. In a data pipeline, the source is the first brick on which everything else rests. No source, no citation. No citation, no verification.

Third cell: the type. Value Unclassified. Something unusual shows here. The type classifier did not leave the field blank; it wrote 'unclassified'. It received something, but could not place it in any known category. Yet Domain Label: tennis is present in the same document. Two classification engines, one document, two different outcomes. The domain classifier ran; the type classifier did not. In my experience the most direct cause is that the two engines are being fed text from different sources, and one of them received empty or truncated text.

Fourth cell: the one-line summary. Empty. This is where all the failures gather into view. A summary exists only when text exists. No text, no summary.

Fifth cell: the information points. The entire section is empty. This is the place where sentences get broken into facts, claims, forecasts and rumours, each kept separately. An empty section here means the text was never decomposed. The entire machinery above depended on the output of that decomposition.

Sixth cell: the entities. Here the document wrote out its own failure. The field instructs: 'identify from the information points above.' This is the moment a record system points at its own empty rows and admits that the material was supposed to come from there. The information points do not exist, and the rest of the document is still waiting for them. This is what a template running live looks like — the machine does not say that it has nothing; it says that it is trying to build something out of what it never received, and is so far failing.

Seventh cell: time sensitivity. Never assessed. This is the field that tells you when a conclusion will go sour. In tennis it means: which season is this article from, what is the ranking date, is this score from last month or last decade. If the field is empty, a fact may be true today and true tomorrow, and nobody will ever know which.

Eighth cell: source quality. The value is also an instruction, not an assessment. The document says: judge from the source field of the information points. The information points are empty, so the judgement never happened. For me this is the costliest omission. In a record, source quality states which claim is first-hand, which is second-hand, which is anonymous. Without that, you cannot even identify stakeholders — who benefits, who loses.

Ninth cell: the author's stance. N/A. Nobody knows whether the document is reportage, analysis, or forecast. In a tennis record that distinction changes everything. 'Medvedev lost to Bautista Agut in 2026' is a recorded fact. 'Medvedev will reach the top five in 2026' is a forecast. Filing both in the same type is the fastest way to ruin any record.

Reading the ledger this way lets me state one thing clearly, and it is the central finding of this essay:

This is not a problem of tennis knowledge. It is a problem of source governance.

Things did not break because nobody knows tennis. Things broke because some part of the machine lost the ability to report that it had received nothing.

The divergence of the two classification engines is the best evidence of that. In any system where two engines read the same text, they should act as checks on each other. When the input is empty, both should fall silent together. Instead, one fell silent and the other answered. That means their inputs were not the same. Where two checks never look at each other, an empty row is more dangerous than a filled one, because the empty row is visible and the filled row is never audited.

One more thing stung. Paragraph-level metadata. Nowhere in the document is there a trace of which sentence came from which part of the source. When text is broken into pieces, each piece must carry its address. It did not, so the source-quality field could never be filled. I call this the cutting error. If a piece is cut without its ticket, it becomes an orphan. And every person who later uses an orphan piece starts speaking in the language of their own assumptions.

I think of 2026, logging points off a broadcast feed. The feed ran with a delay — a second or two, sometimes four. I could not correct the delay, so I began writing 'delay: unconfirmed' against every row. Many of my colleagues thought it a waste. To me it was the most valuable column, because when I reopened that data I knew where to trust it and where not to.

That is what the empty document lacks — a way of saying where the delay is, where the fault is, where nothing yet exists.

The contrarian angle: the empty row is honest, the filled row is suspect

Now the part where I argue with myself.

The first reaction to reading this document is that the system cheated. The machine never ran. All the labour was wasted. That is true, but only half true. The other half is that the document practised an honesty that most data products never practise: it wrote N/A against its own ignorance.

Consider what could have happened instead. The engine knew the domain was tennis. Had it decided that 'tennis' alone was signal enough, it could have invented a title, invented four information points, invented two entities, and returned a document that was fully filled, fully clean, and fully false. I have seen such documents a hundred times. They are so beautiful that nobody bothers to check them.

The empty document at least told me that something was missing. It did not save me from being wastefully idle, but it saved me from being misled. And in sports journalism, being misled costs far more than being idle.

Here I remember those two occasions in the Mumbai studio when I was told my graphic was too complicated for air, and I simplified the language, not the maths. The producers wanted a number. They did not want the sample size, they did not want the margin of error. Had I yielded, a very clean number would have glowed on screen — and that number would have shown the truth of a match not less but more sharply than it deserved. The same logic applies here. A pipeline that returns empty makes noise. A pipeline that returns full and wrong does damage in silence.

So the complaint — that Stage-1 came out hollow — is in fact good news. The bad news would be a Stage-1 that came out hollow and showed no sign of it.

One more thing must be added here, which I call the theft of correlation. The domain label is tennis, so the easiest path is to assume the source document was about a tennis match and write on that basis. That is exactly like deciding the result of a match from the outcome of a single point. One variable is present; treating it as the cause is our error, not the system's. The rush to write a whole story from a partial signal is the oldest disease of sports data. This is where the hands that write forecasts move forward and the hands that record facts stop. I prefer to stand on the other side.

In more than fifty years around this work I have seen that those who insist on recording are called boring, and those who forecast are called visionary. I have also seen whose work survives when old scores are pulled out twenty years later.

Those who stress governance often slide into blaming a federation or a system. Here I must be careful. Nothing in this document tells us who erred. There is no error message, no exception. This is nobody's malice; it is a quiet omission. Keeping documented decisions separate from structural constraints is the whole of governance. The rest is accusation, and accusation is not a record.

Takeaway: a checklist for next time

So what comes next.

I will not throw this document away. I will file it in the same drawer where I kept those forty-one empty rows from 2026 — not as a specimen of failure, but as the first page of a failure ledger. Because a governance that records only successful rows is only half a governance.

In this essay I typed every figure twice: once by hand, once from memory. The final of 8 January 2026, 6-3 and 6-4, 31 main-draw matches, 3,142 points, three seasons, sixty-four matches — all written the way I wrote them in 2026.

The next time a system hands me a clean summary, I will ask three questions first. One: what is the address of this sentence — which paragraph, which document, which date? Two: was the empty cell filled with a guess, or left empty? Three: if two separate engines were reading the same text, why did they not flag each other?

Those three questions hold on a tennis court and in any data system. And if none of them can be answered, the wisest act is to accept an empty row, rather than fill it with a beautiful, confident, false number.