The Night the Data Returned Blank: The Discipline of Verification in Tennis Analysis

Three numbers to open with: 6 hours 36 minutes, 26-24, and 0.

The 2026 Wimbledon semifinal between Kevin Anderson and John Isner lasted 6 hours 36 minutes and closed with a fifth set at 26-24 — the longest Grand Slam semifinal on record. The third number does not belong to that match. It is the number of records my charting system returned on a different night of play, a night when every point on court had already happened and the umpire had already announced the final result.

My desk in Hai Phong has two monitors. On the left runs the point-by-point data collector; on the right sits an open draft. That night, the left screen returned an empty column: no point score, no serve type, no serve direction, no rally length. Only a timestamp running steadily and blank cells where characters should have been.

I wrote nothing for the next forty minutes.

My job is to reconstruct the truth of a match through numbers, and that job rests on one foundational rule: data is never in a hurry. The person in a hurry is the one who gets it wrong. An empty column is not evidence that nothing happened. It is only evidence that my data pipeline broke somewhere between the court and the hard drive.

That is why I am devoting this piece to a subject rarely mentioned in tennis analysis: what happens when the spreadsheet returns zero, and why the correct reflex for a writer is not to fill the blank with narrative, but to stop.

Context: why tennis is harder to chart than people assume

Tennis has the cleanest data structure of any adversarial sport. Every point begins with a serve, most end within seven shots, and the point winner is always clearly identified. Compared with football — where a single passage of play can involve eighteen players and nobody touches the ball twice — tennis is almost designed for counting.

But clean does not mean easy. The data string of a three-set match contains hundreds of points, and each point needs at least seven fields: server, service box, spin type, ball direction, rally length, point winner, and how the point ended. At Grand Slams, automated camera systems handle most of that. At most tournaments in Vietnam, no such system exists. We chart by hand, by eye, and with a spreadsheet designed years ago.

Manual data has a property I learned across many seasons: it does not err randomly. It errs in patterns. Long, tense points are often under-recorded because the chartist gets pulled into the rally. Quick service winners are fully recorded. The result is a data table that looks complete while describing a far easier match than the one actually played.

The core point: a blank cell in a spreadsheet is not a verdict

That night of lost signal taught me a principle I still use. When a data field comes back empty, the right question is not what did this player do wrong, but at which layer did my collection system fail.

There are three layers that can fail, and I check them in a fixed order.

The first is the source layer: the connection to the tournament's scoring system. If this layer fails, the entire table is empty from the first point. The tell is obvious: the timestamp keeps running but no records appear.

The second is the mapping layer: a field code was renamed after an update. This case is far more dangerous, because the table is still full of numbers, only the labels are misaligned. I once had a week in which the column for points won on first serve actually contained points won on second serve. Without manually cross-checking against the official scoreboard, that error would have gone straight into print.

The third is the interpretation layer: the data is correct but I asked the wrong question. This is the hardest layer to detect, and the one responsible for the most wrong conclusions in the tennis analysis I read.

People remember results. I remember the conditions that produced them. Anderson beat Isner after 6 hours 36 minutes, and most fans remember 26-24. What is more worth remembering sits in the point structure: in a match where both players served at an extremely high level, very few break points were created, and almost the entire fate of the match was settled in tiebreaks and long deuce games. A 26-24 scoreline does not measure a gap in level; it measures how many times both men held serve. Those are two entirely different statements, and only one of them is supported by the data.

Using the same test, I look at the 2026 Australian Open final between Rafael Nadal and Daniil Medvedev. Nadal lost the first two sets 2-6 and 6-7, won the next three 6-4, 6-4, 7-5, over 5 hours 24 minutes, and took his 21st Grand Slam title. Read only the result line and the story is the grit of the older man. Read the set-by-set data and the story is more complicated: Nadal's points-won-on-first-serve rate shifted markedly from the first two sets to the last three, and his return points won rose precisely as Medvedev began to wobble on second serve.

To me, that is the most predictive indicator in tennis: points won on second serve, and points won returning second serve. The first-serve points-won rate depends too heavily on an individual's serve quality and the surface, so it says little about the probability of winning a specific match. The second serve is the shot where a player has nowhere to hide: they must choose between safety and ambition, and that choice repeats often enough within one match to become a signal rather than luck.

Data pressure on the rules of play

There is a dimension of the story rarely discussed: data does not only describe the rules, it also creates pressure to change them.

After the 6-hour-36-minute semifinal of 2026, Wimbledon decided to introduce a tiebreak in the deciding set at 12-12, starting in 2026. By 2026, all four Grand Slams had unified on a 10-point tiebreak at 6-6 in the final set. That decision was not purely technical. It was the product of an accumulated body of data on match duration, physical load, players' schedules, and the commercial value of broadcast windows.

When I write for the Vietnamese market, I always place rule changes beside this body of data, rather than presenting them as administrative events. A rule change does not fall from above; it is pushed up from a spreadsheet someone patiently maintained for years.

The Vietnamese context: when the sample is smaller than you would like

In Vietnam I work with a paradox. Fans follow international tennis in depth, yet data on Vietnamese players themselves is thin. Ly Hoang Nam won the 2026 Wimbledon boys' doubles title alongside Sumit Nagal — the biggest milestone for Vietnamese tennis on the junior Grand Slam stage. But his number of matches at ATP Challenger level in a single season is not enough to produce a statistically meaningful series in the way we routinely do for players inside the top 50.

With a small sample, the error margin grows very fast. A player who wins 4 of 5 break points in one match can look like a break-point specialist, but 5 points are not evidence. That is one click of a mouse in a spreadsheet.

Based on my experience following matches, I set a minimum threshold for any claim about a Vietnamese player at twelve consecutive matches on the same surface. Below that threshold, I describe; I do not conclude.

The counterintuitive angle: an empty cell is not truth, and neither is a full one

There are two traps, and they are symmetrical.

The first is believing that an empty field means the event did not exist. In journalism this is the deadliest trap, because it turns the writer's technical error into an attribute of the person being written about. I once saw an analysis conclude that a player had lost the ability to save break points, purely because the author's table was missing the break-point column. With no data, the conclusion was still published, and readers had no way to detect it.

The second trap is subtler: believing that a complete dataset is a complete verdict. In 2026, I wrote about a V-League match in which the home side generated 1.92 xG but lost 0-1. The media called it a collapse of the attack. The data said otherwise: the opposing goalkeeper saved 11 shots, 3.8 times his own season average. The result was decided by an anomaly, not by a trend.

Twenty days later, the home team's head coach cited my figures in a press conference. I did not treat that as a victory. I treated it as a reminder that two people can read two stories out of the same table, and only one of them survives as the sample grows.

The empty stadiums of 2026 were not an exception, but the cleanest laboratory of modern football. No crowd, no roar, no emotional home advantage — and the physical and pressing metrics of that period became easier to compare than ever. I keep that lesson for tennis: look for the windows in which the confounding variable has been removed, because that is when the signal shows itself.

That is why I always reserve a section at the end of every analysis to state clearly what I do not know. xG cannot measure spirit. A first-serve points-won rate cannot measure fear at minute 260 of a fifth set. A spreadsheet cannot capture luck. But admitting those limits is the precondition for the rest of the piece to carry weight.

What I refuse to publish

Across twenty-five years of writing, I have built myself a short list of things that never leave the draft.

First, any claim about injury and return timelines based solely on communications-team statements. Return schedules are controlled by the PR department, and the phrase wait until the weekend in press releases usually means the injury has not healed. I only write about injury when there is data on training workload, or when the player has already competed in three consecutive matches.

Second, any conclusion drawn from a single match. One match is an observation. Three matches are a trend. Ten matches are a characteristic.

Third, any number without a public source. If I cannot show readers where to verify it themselves, I do not put it in the piece. This is why I state the source of every metric in the accompanying data tables, rather than leaving them embedded in prose as self-evident truth.

These rules make me slower than my colleagues. I accept that.

Signals for the next cycle

If you follow tennis in the period ahead, here is what I will have on my desk.

First, the second-serve points-won rate of seeded players, split by surface. This is the first indicator to react when pressure rises, because no player keeps a perfect second serve across five sets if the technique is not genuinely sound.

Second, return points won in the third set of each match, measured across an entire tournament run. This indicator shows who still has legs and still has focus in the phase where most matches are decided.

Third, the quality of the data pipeline you are using yourself. I check mine periodically by manually cross-checking three matches a month against the official scoreboard. If the discrepancy exceeds two percent of points, I stop publishing any metric from that system until I find the cause.

As for the night of lost signal in Hai Phong, the ending was not dramatic. Close to dawn, I found the cause: an update had renamed a data field, and my collector was still calling it by the old name. Nobody did wrong. Nobody lied. A string of characters had simply shifted by one place.

But had I filled that gap with a commentary on a player's form, the shifted number would have become match history in readers' eyes, and nobody would ever have corrected it.

My spreadsheet was empty that night, and the only correct thing to do was to leave it empty until I understood why.