The Record That Came Back Empty: How Tennis Data Dies Quietly Inside the Pipeline

Last week a data vendor's file opened on my screen. Twelve columns — match ID, tournament, surface, first-serve percentage, return points won, break-point conversion, winner-to-unforced-error ratio. Every header sat in its place. Under every header, not a single number. The file was named match_report_tennis.csv, its size was 340 bytes, and its domain label was perfectly correct — tennis. Everything else was blank. The system raised no error. It said: done.

This is not merely a technical glitch. It is a new form of an old illness in tennis that I have been fighting for twenty-five years — the disappearance of the record, and nobody watching it disappear. The only difference is that records used to die on paper. Now they die on servers, and while dying they look entirely valid.

January 2026. The Chennai Open was running at the SDAT Tennis Stadium in Nungambakkam and I was thirty, three years into a statistics job at a Chennai sports-data vendor. I was handed the full main draw — thirty-one matches. I hand-coded 3,142 points. Each point carried three tags: serve direction into three zones, rally length, and error type. The client was a European aggregator. After that tournament, not one Indian coach asked to see the file.

I kept a personal copy of every sheet. That was the first slab of an archive I still maintain, and its most useful property is that it also records what is missing. I keyed every point by hand, and the scoreboard never learned my name. That was the definition of my job, and also my complaint about it.

A record is a custody problem, not an analytics problem. Until someone decides which document sits in which cupboard and who holds the key, even the best data remains only a good memory. Of twenty-one Chennai Open editions, I hold seventeen. Four are gone — into a closed agency, a reformatted laptop, an intern's inbox. I do not know. And that is the real problem: I do not know why they are missing.

In January 2026 the tournament was played for the twenty-first and final time. I was in the media box coding the final — Roberto Bautista Agut beat a twenty-year-old Daniil Medvedev 6-3, 6-4. I did not know that day it was the last one. The following year the event moved to Pune as the Maharashtra Open, my accreditation lapsed, and the archive travelled with me while the camera stayed behind.

I watched the 2026 edition on television, logging points off a broadcast feed with a delay I could not correct. When the ball was on court, it had not yet bounced on my screen. Gilles Simon beat Kevin Anderson in that final, and not one of my serve splits from it is trustworthy. It took me three months to accept that the access was gone for good. That was my slowest admission, and I am ashamed of it.

Meanwhile another record slipped out of my hands, faster. From 2026 to 2026 I was performance analyst for the Indian Aces in the International Premier Tennis League, which the Aces won in the inaugural 2026 season. The league introduced gimmicks no tour event carried — a twenty-second shot clock, power points, a super shoot-out replacing the tie-break. I built a model of how each player's first-serve percentage moved under the clock. It was a good model. It had data no other tour possessed.

By the end of 2026 the league had folded, with players chasing unpaid money. Three seasons with the Indian Aces, then a spreadsheet with no next row. Nobody opens that file now. The clock-pressure model was never requested by an academy, because academies never took seriously a league that could not survive. A franchise dies in the footnotes long before it dies in the press.

Put the three cases side by side. The Chennai record broke because the tournament changed city and no custodian travelled with it. The IPTL record broke because the company closed and the data was the company's asset, not the sport's. And now the third, newest case: the record is breaking because the pipeline stopped working and never said so.

All three share one property — none of these events registers on any scoreboard. No 429, no 500, no alarm. The system returns a perfectly valid, entirely empty payload and reports success. I found the error at 2 a.m., in a column nobody ever reads: file size. 340 bytes. Thirty-one matches of data cannot fit in 340 bytes. That was the whole test.

And here I have to admit the most uncomfortable truth of my profession. In tennis we talk constantly about a shortage of data. The shortage is not of data. It is of the checking that should follow the data's arrival. Once you have built a schema, an empty schema is still valid. Once you have built a dashboard, a dashboard can display zero, and that zero will hide beneath a graph.

In June 2026 I spent the Russia World Cup inside a Mumbai broadcaster's studio, having 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. Twice I simplified the language rather than the maths. That was my most honest compromise, and I still call it a compromise.

The same rule applies to the Indian tennis record, and here the divergence is sharpest. The market keeps selling a singles dream, while India's real Grand Slam inheritance sits in doubles and mixed doubles. Leander Paes holds eighteen Grand Slam titles — eight in men's doubles, ten in mixed. Mahesh Bhupathi holds twelve — four in men's doubles, eight in mixed. Sania Mirza holds six — three in women's doubles, three in mixed. Those figures come from the official Grand Slam record, not from an estimate.

Now look at this country's tennis coverage. The column inches spent on a single singles quarterfinal exceed what eighteen titles ever received. This is not merely an editorial preference. It is a property of the record-keeping system in which we filed doubles as a subsidiary category. When an achievement is filed as subsidiary, its data is filed as subsidiary too — less coding, less verification, less archiving. Twenty years later someone asks where the Indian tennis record is, and the answer is simply this: the thing we never counted, where would it be?

Here I have to fight my own habit. I am one of those who believe every figure should be typed twice — once by hand, once from memory — and if the two disagree, a third time from the source. It is a slow method and I keep it deliberately slow. But I must also admit that hand-keying became a defence mechanism for me, and defence mechanisms die late. Even when the pipeline improves, I keep watching it with suspicion, because my entire method stands on that suspicion.

That suspicion has a second face I will not hide. The old method's greatest cost is that it conceals error, because correcting error is labour. The new method's greatest cost is that it conceals emptiness, because emptiness never arrives as an error code. Neither is safe. The only safe thing is to insist that every figure carry, beside it, where it came from, when it came, and who checked it.

Is this a private grievance? I do not think so. Tennis is the sport that generates the most data and curates the least. A Grand Slam records the speed of every serve, the length of every rally, every movement of every player. A year later, ask for the original point-by-point record of the same match and you will hit a vendor's wall. Even the tour does not hold it complete. This is strange, and it has been normalised.

The example that stings most: among those seventeen Chennai editions, several matches exist for which I hold point-level data but no sourced official scorecard. I have information nobody else has, without the proof everyone should have. The reverse is also true — many matches have an official scorecard with no underlying data held by anyone. That is not a record. It is half a record, and half a record is more dangerous than none, because half a record does not lie. It simply falls silent.

I rebuilt from the bottom up. I cannot say that sentence about every project, but I can about the archive. After 2026 I digitised every sheet, stamped each file with a date and a version, and separately flagged every place where the data is incomplete. Today the most valuable part of my archive is not the columns that are full. It is the columns beside which I have written: source unknown.

That is the real lesson, and it comes directly from the empty CSV that opened this piece. A good data system is not one that fills every cell with a number. A good data system is one that shouts when a cell is empty. That 340-byte file contained not one wrong figure. It also contained the largest possible error — the illusion that the work was finished.

I will not forecast what happens next, because forecasting is not my job. But one thing I can fix: the next time a tournament changes city, the next time a league closes, the next time a vendor swaps its feed, the question will not be how much data existed. It will be who held the key. The day Indian tennis answers that question, it will not need a foreign aggregator to count its own inheritance.

Until then, seventeen editions sit in my cupboard, four spaces are empty, and I have deliberately left them empty.