An Empty Spreadsheet Before the First Serve: When a Data Journalist Must Say “Insufficient Evidence”

The night before a Challenger-tier draw began, I opened my spreadsheet and found exactly three things with content: a column of player names with a typo, a column marked “hard” for surface, and an empty column for first-serve percentage. The sheet had four columns. The number of real data rows: zero.

I sat looking at that zero for about twenty minutes. In my profession, that kind of silence is not comfortable. Hundreds of professional tennis matches are played around the world every week, and most of them leave behind nothing but a final score. A three-set match lasts two and a half hours, roughly one hundred and eighty points, each point a measurable event — and all of it evaporates from history simply because nobody sat down to record it.

Viewers see 6-4, 3-6, 7-5. I see an empty spreadsheet and a question with nothing to hold onto. The zero sitting in that fourth column was, in fact, my first piece of data for the day.

I have been tracking sport through spreadsheets for twenty-five years. My career began at the Daily Mail, where I spent fourteen years, then moved through Sports Illustrated as a fact-checker — a job that paid me to say “this number is not right yet”. Born in the United States and now living in Hai Phong, I write for Vietnamese readers, and that geographic shift taught me something no classroom did: data infrastructure is not evenly distributed, and it follows the money.

Tennis has the clearest tiered data structure of any sport I have covered. At the top, a Grand Slam match is recorded point by point: serve speed, bounce location, spin, distance covered by the player. Hawk-Eye arrived at the majors in 2026 and opened an era in which every shot has coordinates. In the middle tier, the ATP and WTA Tours have dedicated data providers. Down at Challenger level, the numbers thin out quickly. At ITF World Tennis Tour level and in domestic events, what remains is usually just a scorecard written by hand.

That gap has real consequences. It determines who gets to be described, and described with what.

In 2026, mid-way through the V-League season, I published the first series applying expected goals to Vietnamese football. In a match between Hai Phong and SLNA at Lach Tray stadium, the home side generated 1.92 xG but lost 0-1 to an individual error. The media called it a slump. I called it random injustice, and pointed out that the opposing goalkeeper had made eleven saves, 3.8 times his own average. I was mocked for two weeks — until the Hai Phong head coach publicly cited my numbers in a press conference.

Since then I have held one inviolable rule: no verified numbers, no conclusions. That rule has cost me a few pieces that were doing very well at the time, and I accept the cost.

Germany collapsed in my spreadsheet before it collapsed on the pitch. In June 2026, ahead of the World Cup group match against South Korea, I published an analysis showing Germany’s pressing coefficient had fallen from 8.1 PPDA in 2026 to 12.6, with average distance covered down 6.2 km per match. On the field: 74% possession and a 0-2 defeat. The lesson was not that I was right. The lesson was that data is never in a hurry.

When a spreadsheet is empty, a writer falls into one of two traps. The first is inventing a story and attaching numbers to it afterwards. The second is staying silent, as if the match never happened. Both are different ways of lying.

The unit of tennis data is a single point, and it is an expensive unit.

A three-set match contains roughly one hundred and fifty to two hundred points. A five-set Grand Slam match lands between two hundred and fifty and three hundred and fifty. That sounds like plenty until you divide it up. In a typical three-setter, a player serves about eighty to ninety times, converts roughly half of those into first serves, and wins perhaps forty to fifty points behind the first serve. Break points converted across the whole match usually sit below ten.

That is the sample size. Anyone who has done statistics knows a sample under ten cannot support a firm conclusion.

The four core metrics every tennis dataset must carry are first-serve percentage, first-serve points won, second-serve points won, and return points won. Together they can reconstruct a match without watching it. They can also mislead a careless reader.

Suppose a player wins 74% of first-serve points in one match and 68% the match before. The press writes that his serving form is declining. In reality, with roughly forty-five first-serve points per match, that gap sits comfortably inside the noise band. Six percentage points looks like a trend, but most of the time it is just random variation in a sample that is far too small.

This is where I learned the most from my own mistakes. In my first year writing about tennis with numbers, I built an entire piece around one player’s declining ability to convert chances, based on 3 of 14 break points across three matches. An old colleague from the Daily Mail, who had built tables for twenty years, messaged me one line: what exactly are you measuring? I reopened the file and realised I was measuring luck.

Break-point conversion is the most treacherous metric in the sport. A player gets eight break chances in a match. Convert four and the rate is 50%. Convert two and it is 25%. Those two outcomes are two points apart — two rallies, about thirty seconds of one afternoon. Yet in the next day’s coverage, one is character and the other is a weak mentality.

I am not saying break points do not matter. I am saying eight chances is far too few to name a quality.

Every serve is a hypothesis. First-serve points won is how we test it — but only when the number of serves is large enough for the test to mean anything. At Grand Slam level, where a player contests seven matches and serves several hundred times, the metrics begin to stabilise. At Challenger level, where a player may lose in the first round and serve forty times all week, every table is fragile.

Another favourite of the press is average rally length. Players who win a lot tend to have shorter rallies — statistically true. But short rallies are the result of serving well and returning early, not the cause. Plenty of analysis reverses that causal arrow and then advises players to finish points faster. That advice is meaningless: finishing points quickly is what happens when you are hitting well enough, not a decision you make.

That is why I always print the sample size beside every figure. Not to make the piece look more professional, but so readers know which parts they are allowed to trust.

Tennis data infrastructure is changing faster than most Vietnamese readers realise. The creation of a joint data venture between the ATP and WTA marked the first time the two major tours sat at the same table over ownership of statistics. But that change happened at the top of the pyramid. At the bottom, where thousands of players grind for ranking points, little has changed in twenty years.

I once spent time at a Challenger event held in Ho Chi Minh City across several seasons. The organisers did the stands, the media and the sponsorship extremely well. Nobody was responsible for data. The electronic scoreboard showed the game score, and that was the entire systematic record. To learn what percentage of first serves a player made, I had to count by hand.

So I counted. In one quarter-final I counted all one hundred and seventy-three points across two hours and forty minutes, logging each one into four columns of a notebook. The final tally showed the winner had taken 71% of first-serve points and only 44% of second-serve points — a hole the scoreline never revealed. Reading only the score, I would have written that he won comfortably. Reading my hand-counted sheet, I knew he won on one leg.

That is the kind of information the Vietnamese market lacks, and it lacks it for a very concrete reason. Nobody pays anyone to sit and count.

At national-team level the problem is starker. SEA Games and regional team events come with full scorelines, full news coverage, full photography — and near-total emptiness on process data. When Ly Hoang Nam held the Vietnamese number one spot for years and reached the world’s top two hundred and fifty according to the public ATP rankings, most of what the public knows about him came from description, not measurement. Nobody knows precisely what percentage of second-serve points he won at his peak, because that data was never systematically recorded.

For a player, that is a lost legacy. For a tennis nation, it is a lost capacity to learn.

There is a paradox I keep running into: the sport’s own governance rules are a source of new data. When the majors introduced a twenty-five-second serve clock in the late 2010s, it became possible for the first time to measure the interval between points in a standardised way. When the ATP permitted off-court coaching at certain events from 2026, analysts gained an entirely new variable. Rules on medical time-outs, on shoe-change breaks, on foot faults — each one creates a column.

But those columns exist for match management, not for analysis. Inexperienced analysts mistake data generated by governance for data generated by performance. A player who always uses the full twenty-five seconds before serving may be deliberately controlling tempo, or may be losing rhythm. One number, two opposite stories, and the spreadsheet will not adjudicate.

During a major tournament season this multiplies. Data volume surges at the same moment commentary volume surges, and the two are not of equal quality. Each Grand Slam produces thousands of articles inside fourteen days, most drawing on the same very narrow set of figures. The repetition makes a claim look more credible than it is. It is a crowd effect wearing a digital coat.

The counter-intuitive angle sits here: I have spent most of this piece on what numbers cannot do, but the exit is certainly not to throw numbers away.

People remember results. I remember the conditions that produced the results. But a memory of conditions is only worth something if it was recorded before the result arrived.

An empty spreadsheet does not say nothing happened. It says nobody measured. The distance between those two statements is far wider than it looks — because the first is a claim about reality, and the second is only a claim about people.

The second danger is using humility as a shield. After presenting the data, stating the sample size and marking the error band, refusing to offer a judgement is no longer science. It is avoidance dressed in terminology. Spectators can leave the stands, but physical data never takes a day off — and neither does the writer, when a hard question arrives.

The less-discussed blind spot runs the other way: thick data at the top and thin data at the bottom means the public only ever sees roughly twenty leading players, while thousands more work at this sport seriously. Every story about a golden era or a next generation is written from a tiny sample sitting at the summit. Most of the real story of this sport lives in the tier with no numbers at all.

From my years of watching matches, one simple principle: if forced to choose between a good story and a boring table, read the table first, then tell the story. The order decides everything.

The signal I will track in the next round is not who wins the title. I will track how many Challenger and ITF matches publish point-level data, and whether anyone at a Vietnamese domestic event will sit down and count a match from first point to last. Those are dull indicators, and nobody writes headlines about them.

But if a young writer in Hai Phong sends me a spreadsheet with one hundred and eighty rows for a match where I previously had only a scoreline, I will know this tennis nation has started measuring itself. Data is never in a hurry. Only people are.