The Data Sheet Came Back Empty: The Discipline of the Number Writer

02:47

In Shenzhen, August is hot enough that the office air conditioner has to run at its lowest setting and still hums. I was sitting in front of two screens. The left one held a tracking sheet for thirty league matches. The right one held an analysis file with nine sections, each with its own heading, each heading with its own data cell, and nearly every cell carrying the same value: N/A.

I ran the extraction three times. On the first pass I assumed I had entered the wrong match ID. On the second I switched data providers. On the third I changed the way I framed the query. The result did not move. The sheet came back empty.

A newcomer panics. Someone who has been at this long enough notices something else: an empty sheet is not a technical failure. It is a result.

In 2026, when I was an intern at a football newsroom in Shenzhen, I asked a question at a post-match press conference about the home side's hybrid 4-4-2. A senior male reporter cut me off mid-sentence. In 2026, the press conference door closed in front of me. Today, I read it through data.

Tonight, the data closed its own door. I stayed seated and read that closure instead.

There is a trained reflex in this profession: when there is nothing to say, say something about having nothing to say. It is a comfortable reflex, because it is always technically correct and almost never challenged. It is also the first trap. A piece about emptiness can easily become an empty piece dressed up in elaborate sentences.

I reopened the file. Nine sections. Thirty-seven cells. Thirty-seven instances of N/A. And at the bottom, a note I had written myself months earlier: "Do not infer from an empty source."

That note was the only item in the file with genuine analytical value.

The architecture of a data report

A decent data report is built on an investigative architecture, not a commentary one. It starts with a hypothesis that runs against the crowd, then builds a data frame, cross-checks multiple metric sources, and only then commits to a conclusion. That order cannot be reversed. If you commit to the conclusion first and go looking for numbers to defend it, you are not writing a data report. You are doing public relations for a prejudice.

The skeleton I use has five joints: an opening on an anomalous metric, a context joint on method, a core joint of data, a counter-intuitive joint, and a closing joint on the signal for the next cycle. Those five joints work like the five beats of a defensive table tennis rally: wait, read the spin, absorb the pace, redirect, counter.

One layer upstream, a data report needs a source deconstruction step. That step must return at least eight things: the original headline, the source, the genre, the core viewpoint, the list of information points, the list of entities involved, the time sensitivity, and the quality tier of the source. When all eight are missing, what I am holding is not a report. It is a blueprint with no building, a nine-part frame with no material, a board with no chalk.

The uncomfortable part is that the frame still looks beautiful. It has room for technique and tactics, for player data and head-to-head records, for event systems and ranking rules, for the competitive landscape between one dominant table tennis nation and the rest of the world, for rules and governance, for coaching staff and talent pipelines, for the risk surface, for public narrative and expectations, and for the transmission line of an entire industry.

Because it looks beautiful, it is dangerous. A beautiful frame can be misread as a finding. I have seen enough of that in this industry to know that the fatal error of data work is not misreading a number. The fatal error is reading the frame as if it were the number.

Nine empty cells and what they accuse

I went through each section of the file, not to list them, but to read what each empty cell accuses.

The technique and tactics section is empty, and it accuses the input of containing no technical content to assess. No description of playing style, no tactical deployment, no single-match review, no equipment change. In table tennis this is the most important section of all, because table tennis is a sport where a small change in rubber or in the ball itself can bend a whole run of results. With no data here, there is nothing to discuss.

The player data and head-to-head section is empty, and it accuses the input of naming no entities. In table tennis, a head-to-head file can say more than a ranking: win rate abroad, consistency at major events, performance at deciding points. All of it needs a name to start. Without a name, there is no analytical funnel.

The event system and ranking rules section is empty, and it accuses the input of mentioning no specific event. No ranking points for the champion, no prize money, no strength of the entry field, no position in the Olympic cycle. In a sport where ranking points decide major-event entries, missing this section means missing the entire decision context of the athlete.

The competitive landscape section is empty. There is no comparison table between a leading table tennis nation and its main challengers: seats in the world top ten, titles at the last five majors, depth in the under-21 cohort. Those three metrics are the three legs of a stool. Without all three, the stool does not stand.

The rules and governance section is empty. No competition-rule reform, no event-system rules, no selection rules, no disciplinary sanctions. This is the section I watch most closely as a reader of rules, because rules are the only thing in sport that can change outcomes without needing talent.

The coaching staff and pipeline section is empty. No team, no coach, no age structure in the main tier, no conversion efficiency from the youth tier upward. In table tennis this is the section that decides an eight-year cycle, not a single season.

The risk surface section is empty. No competitive risk, no selection risk, no generational-gap risk, no governance and public-opinion risk, no systemic risk, no opponent risk. When no cell is filled, the overall risk level is not low. It is undetermined, and those two things are very far apart.

The public narrative and expectations section is empty. No sentiment line, no heat-cycle position, no analysis of the gap between market expectation and objective assessment. This is the section where writers most easily fool themselves, because crowd emotion always leaves a trace somewhere, and a trace can always be read as evidence.

The industry transmission section is empty. No equipment market, no grassroots training base, no commercial ecosystem around events, no commercial value of players, no policy and capital flow, no international ecosystem.

Nine empty cells. And the notable thing is that none of them is empty because I was lazy. All of them are empty because the source input was empty.

| Type of gap | How to recognise it | Correct handling | |---|---|---| | Source gap | No original article to cross-check | Stop analysis, request the source | | Entity gap | No names of people, events or teams | Do not guess a topic from a vague prompt | | Time gap | No absolute dates | Mark as "not assessed" rather than assigning one | | Sample gap | Sample too small to conclude | Downgrade the conclusion to a hypothesis | | Definition gap | Metric formula unknown | Trace the original definition before use | | Motive gap | Source has a stake in the outcome | Separate the fact from the source's framing |

Table tennis: where data is as thin as the rubber

I came into number-writing through table tennis, and that was a harsh start.

Table tennis is a sport where public data is far thinner than in football. In football I can pull expected goals, passes allowed per defensive action, distance covered, pressure metrics. In table tennis, most of what decides a match sits in things that resist quantification: spin, placement, tempo, and the ability to read an opponent's intention inside roughly a tenth of a second.

Thin does not mean absent. There are four metric families I always try to secure before writing anything about a table tennis match.

The first is average rally length. A player with a short average rally length usually wins by finishing early, which means through the serve and the third-ball attack. A player with a long average rally length usually wins by dragging the opponent into the mid-rally zone, where fitness and consistency decide.

The second is serve efficiency and receive efficiency. Those two together tell me what kind of match this is: a match where both sides hold serve is a match of good servers; a match where returners break often is a match of good spin-readers.

The third is the win rate at deciding points, usually counted from the eighth point onward in each game. This is the single metric that most clearly separates a good player from a champion.

The fourth is the win rate in matches played abroad. In table tennis this matters more than it appears to, because competing abroad means time zones, climate, the table, the ball, and a hostile crowd.

When those four families are missing, I do not write. I sit and wait. And waiting, in this profession, is a professional skill rather than a form of slowness.

I have hosted broadcasts of several major events, including editions of the Table Tennis World Cup and badminton events of comparable tension. That experience taught me something no data sheet can teach: most of the information in a match is not in the scoreline. It is in the silence between two points. That is where players wipe sweat, rotate the racket, look down at the table, and rebuild order. No metric measures that silence. But if I ignore it, I will write the match wrong.

Football: where data is thick and still full of holes

In 2026, when the World Cup was held in Russia, I was a second-year student hired by a sports outlet to write data pieces. One laptop, 64 matches, and a world telling the story of the World Cup through numbers. That summer shaped how I work.

I built my own spreadsheet tracking each team's pressing metric across every match. The metric measures how many passes an opponent completes before your team makes its first defensive action. The lower the number, the more aggressively the team presses.

What I noticed as the sheet filled up: Croatia averaged a pressing metric of 9.2 across the knockout rounds, well below the early-exit group. I wrote a prediction that Croatia would reach the final, despite being underrated. The piece drew 120,000 reads, four times the site's normal level. From then on the newsroom treated me as its own tactical writer, and the gender story from the press conference receded.

But I have to tell the part fewer people tell. All three of Croatia's knockout matches that year went to extra time: 1-1 against Denmark and a penalty win, 1-1 against host Russia and a penalty win, and a 2-1 win over England after extra time. Three matches, three trips to the final limit of physical capacity.

A low pressing metric can explain why Croatia controlled the tempo. It cannot explain why Croatia survived three consecutive periods of extra time. If I attribute that entire achievement to a pressing metric, I have turned a correlation into a cause. That is the most common error in data writing, and it is more dangerous than getting a digit wrong.

The truth is that the rest of the Croatia story sits in columns my spreadsheet did not have: the capacity to absorb pain, the willingness to take risk in a shootout, and a goalkeeper who read direction better than average.

Tactics are what people draw on a blackboard. Data is what they draw on reality. And reality, in Russia that year, was thicker than the blackboard.

Referees and VAR: the gap is wider than people think

There is one domain I have tracked for years where I always find the data insufficient: officiating.

VAR entered the World Cup in 2026 and has since become part of nearly every major tournament. Technically, it does things the human eye cannot: drawing offside lines, measuring the point of contact, rewinding from multiple angles. In terms of judgement, it does not solve the underlying problem, because its intervention threshold is defined by a vague phrase: clear and obvious error.

I once tried to quantify that phrase. I took reviewed incidents from a domestic league across three seasons and classified them by degree of contact, position on the pitch, match minute, and player movement speed. The result: at the same level of contact, in the same position, the intervention rate differed markedly by match minute and by the scoreline at the time.

That is a finding I cannot prove from public data, and I will not call it a fact. I will only say that the space for subjective judgement inside VAR is wider than the way it is presented. A system sold to the public as an objective mechanism operates on a subjective variable. Nothing is technically wrong. Something is wrong in the communication.

In table tennis the pressure takes a different form but shares a nature. Table tennis has the serve, and the rule against hiding the ball has been enforced since the early 2000s. But determining whether a service motion hides the ball depends on the umpire's angle, and in a sport where the ball travels faster than the eye can follow until it has already passed the net, angle is everything.

I note this in every relevant report: when a judgement system is built, measure the subjectivity left inside it, rather than only the technology added to it.

Injury and return: the schedule is written by communications

No domain distorts data as much as injury.

An injury case is a time series with many markers: scan date, consultation date, surgery date, first light jog, first session with the ball, first contact session, return to the matchday squad. Of those markers, only two are usually published: the injury date and the expected return date.

The second marker is the most interfered-with. It passes through communications, through the agent, through the coaching staff, and each time it passes, it shifts toward whoever is speaking.

In my profession, the phrase "we will assess at the weekend" has become a signal. When a club says it will assess further and decide at the weekend, it means the injury has not healed enough to allow a clear plan. When they give a specific date, it means recovery has passed its least predictable stage. The difference between those two statements corresponds, in my data, to roughly a two-week average gap between the announced return and the actual one.

In table tennis this is more complex still. Table tennis is a sport where the wrist, shoulder and back must work near maximum range on every shot. An unhealed wrist injury destroys the ability to read spin before it destroys the ability to score. The player still walks out, still holds the racket, still serves, but performance at deciding points drops first.

That is why I never use the scoreline of a comeback match to judge recovery. I use the win rate at deciding points across the first three matches after the return. That metric tells the truth more reliably than the result, because it only appears when a player is forced to compress maximum force.

The transfer market: the noise of the agent

I work at the edge of the transfer market, and I have learned that the largest expense in a deal is usually absent from the prospectus.

In 2026 a transfer worth 222 million euros was completed, breaking every previous record and reshaping the entire price scale of the market. Since then, fees at or above 100 million euros have stopped being exceptions. But inside every published fee there is a portion that never appears on the board: the cost of the people standing in the middle.

Agents do not create players. They create information about players. And when information is a commodity, the market stops pricing ability. It prices narrative.

I once followed a domestic transfer window in China with a single purpose: counting how many times a name appeared in transfer news before any official confirmation. Some players were mentioned more than ten times in one window without a single deal happening. In my analysis, that mention count predicted the final fee better than the player's own performance the previous season did.

That is a correlation, not a causal relationship. But it is enough for me to ask one question of every report I read: who benefits if this name appears here, at this moment, written this way?

The most useful tool I have in this domain is simple. I keep a three-column table: where the information came from, who benefits if it is true, and who benefits if it is false. Those three columns filter out most of the noise before it reaches my writing.

The data storyteller with no data to tell

In 2026, when the pandemic shut down the entire sports system, clubs in Shenzhen lost their live training data streams.

I was a final-year student then. I wrote code to analyse roughly 5,000 historical matches, including matches before and after the closure of stands in a European league in the 2026-20 season. What I found: teams with an average age above 28 lost about 17 percent of their attacking efficiency in away matches played without crowds.

I sent a 32-page report to a sports analytics company in Shenzhen. They hired me after a single interview. Colleagues called me the data monk, because of my habit of sitting motionless in front of a screen for ten hours.

The empty stadiums of 2026 taught me that football never goes fully silent. It simply moves its noise to another channel. The crowd disappeared, but the noise of the crowd left a gap behind, and that gap is also a form of data.

What I discovered was not in the 17 percent figure. It was elsewhere. Young teams did not lose attacking efficiency. Which means the lost efficiency was not caused by the absence of crowds in general, but by a specific group of players reacting differently to that absence. That is a psychological finding drawn from data, but not fully explainable by data.

And that is where I stop. My prediction model has no heart, and that is why it is never wounded. But for the same reason, it will never understand why a 30-year-old needs a crowd more than a 21-year-old does.

What if this metric is wrong

At this point I have to turn the question back on myself. That is a mandatory step in the process, not a ritual of humility.

What if Croatia's 2026 pressing metric is wrong? If my provider defined a defensive action differently from the definition I assumed, the 9.2 could become 11.4 on the same raw data. My conclusion would flip without a single on-pitch event changing.

What if the 17 percent from 2026 is wrong? That figure depends on how I defined attacking efficiency, on whether I normalised per minute, and on whether I excluded matches with red cards. Three technical choices, three different answers.

What if the transfer mention count is wrong? It measures media coverage, not player ability. A frequently mentioned name may only reflect a harder-working agent.

Those three questions do not destroy my work. They are the foundation of my work. Every report I write carries a closing paragraph acknowledging its limits, because I know the reader today is smarter than the reader of 2026, and they will find the hole if I do not point to it first.

The discipline of the empty cell

Back to the file at 02:47.

Nine sections, thirty-seven cells, thirty-seven instances of N/A. I had three options. Fill the cells with speculation and write something that looks profound. Refuse to write and write the night off. Or write about the emptiness itself, but with discipline rather than emotion.

I chose the third. Not because it is easy, but because it is hard: writing about emptiness demands describing exactly what is empty, rather than turning it into an excuse for an essay on humility.

There is a permanent temptation in this profession: going against the crowd. It starts as a professional principle and easily slides into a habit of contrarianism. I check myself with a single question before publishing: would this metric force a coach to change a decision? If the answer is no, I do not write it. If the answer is yes, I ask one more: change in which direction, and is that direction right?

The second temptation is self-isolation. Deep-diving writers easily develop a thickness of prose. Inside a long data report, the reader needs a breath. I learned that from table tennis: after every long defensive rally, there must be a change of direction. After every block of metrics, I place a real person's story, so the numbers have a face.

The third temptation is letting caution become paralysis. Knowing the limits of data too well can spiral into unanswerable self-interrogation. I handle it with a small administrative rule: write the firm conclusion first, then reserve a closing section for the limits. The conclusion must exist. The limits must exist. Neither may eat the other.

The signal for the next cycle

I closed the file and opened a new one. The clock in Shenzhen read 03:20.

There is one thing I know for certain about this season, even with an empty analysis file in front of me. The signals worth watching always sit in the same three places. In football, it is the length of a run in which a team must sustain high pressing, because that is where fitness and squad depth surface before the table does. In table tennis, it is the deciding-point win rate of the cohort moving from the twenties into the top ten. In both sports, it is injury statements issued without a specific date.

Those three places do not need a perfect dataset to begin. They need someone willing to sit long enough to read what is not yet there.

I do not need a press conference to prove I understand football. I have 64 matches in my laptop. But tonight, all I have is an empty sheet and a self-written note. For someone who works with numbers, that is already enough of a result to start tomorrow.

The crowd leaves the stands. The players leave the pitch. The agent is still on the phone. The referee is still at the monitor. And I sit here, waiting for data that will not arrive, in order to record precisely that it did not arrive.

Appendix for readers who want to check

A decent data report must let the reader verify it. So I record here what a standard input source should carry before any analysis is written.

Original headline and publishing source, with an absolute date. No "yesterday" or "this week", because relative markers decay within days. The article's genre: news, interview, analysis, or opinion, because each allows a different degree of inference. The core viewpoint in one sentence, so the reader knows what the source is trying to prove. A list of information points, each a verifiable fact. A list of entities: people, teams, events, organisations. Time sensitivity: whether this is worth a day, a week, or a season. And a source quality tier, to separate verified information from rumour expressed carefully.

When those eight are present, I can write a complete data report within six hours. When they are absent, I can write something three times as long and say nothing.

The difference between those two cases is not the writer's talent. It is the quality of the input. And in an industry that rewards speed, saying the input is not enough is a professional act, not a refusal.

I left that line in the file, right under the thirty-seven instances of N/A, as a reminder to myself for the fourth extraction pass.