The Empty Data Pipeline: A Lesson on the Limits of Badminton Analysis

Four in the morning in Surabaya, the third monitor in my office showed red. I was preparing an analysis for a series of important badminton matches, and the automated data pipeline — the thing I rely on before my coffee even cools — returned an empty file. No tournament name. No source. No information points. No identifiable entities.

In my profession, an empty file is more frightening than a wrong file. A wrong file at least gives you something to argue with. An empty file is silent, and silence in an analysis room is always filled with the most dangerous thing: speculation dressed in the clothing of conclusion.

I sat there, watching the cursor blink for forty minutes. In those forty minutes, I could have written a fluent analysis of a match that never existed. I could have assigned some player a form, an index, a tactical weakness — all of which would have sounded very reasonable, and all of which would have been fabricated. This is the greatest professional temptation of a data storyteller: we are trained to find a story, and a good story is always more seductive than a gap.

But a gap in sport is not a flaw. It is data.

That night I shut down the machine, and I understood that I had just touched a subject the badminton world rarely dares to address directly: when the source of information is empty, the boundary between analysis and speculation is more fragile than any line on the court.

What I learned from that empty file was not how to fix the pipeline, but how a professional must maintain honesty when there is nothing to say.

The context of this story is larger than one sleepless night. Badminton over the past fifteen years has undergone a silent revolution, and I was present at its edge, first as a data adviser to clubs and later to a sports platform. When I started, a badminton analysis session was a notebook, a few sheets of carbon paper, and the memory of an old coach. We recorded by hand the number of bad serves, the number of net errors, the number of winning smashes in the third game. Those numbers were scattered across three different notebooks, and each coach remembered a different part.

Today it is different. Every Super 1000 tournament and above has dozens of data points recorded automatically for each rally: smash speed, shuttle trajectory height, the position of a player's feet at contact, reaction time after a short serve, the average length of a rally. National teams hire dedicated analysts. National federations invest in tracking cameras and motion-recognition software.

This revolution has a halo: it makes everything sound like exact science. But badminton data has never been exact science. It is a pile of unrefined raw material, and people often pour that raw material straight into a conclusion written before the match took place.

I have seen this everywhere in the China–Indonesia data corridor, where I worked for many years. The two largest badminton nations in Asia produce champions according to two opposing philosophies. China believes in the system, in the curriculum, in standardized process and repeatability. Indonesia believes in instinct, in personality, in the explosive moment of a player who can reverse a match with an unpredictable stroke.

Both philosophies are correct, and both produce numbers that lie in different ways. The Chinese system produces players whose indices are so stable you would think they were machines, but on a cold day in a cold arena, that stability can become rigidity. The Indonesian instinct produces players whose index volatility is so wide you don't know which number to use to back them, but in a decisive series, that volatility is the weapon.

The problem lies right there: we learned to trust data before we learned to understand it. We built complex pipelines to collect figures, then forgot that a pipeline is only good when it knows how to say the most important sentence: "I do not have enough information."

I call this the empty-input problem, and it has haunted me since a World Cup I never attended.

In 2026, I followed the World Cup in Russia from a café in Surabaya and wrote a series of analysis pieces for a local sports site. I remember spending over a week just looking at Croatia's numbers. At the time, most people judged Croatia by results and the beautiful strokes of Luka Modric and Ivan Rakitic. But I found something else.

Croatia's PPDA in the group stage was only 9.2 — not the highest-pressing team in the tournament. If you stopped there, you would conclude Croatia was a slow team, controlling in their own half and waiting for opportunities. That conclusion was wrong. Because when I split the pressing index by pitch zone, Croatia had the highest recovery rate in the opponent's half of the tournament: 12.4 times per match. They did not press everywhere; they chose the right moment to close down in the right place, and most of those close-downs occurred after Modric or Rakitic had read the opponent's next move.

That piece was shared more than two thousand times in tactical communities in Vietnam and Indonesia. It did not spread because I wrote well. It spread because I did not tell readers a number — I told them a tactical intention behind the number.

But the bigger lesson from Croatia was not about PPDA. It was about the period before I had any data at all. In the early days of following Croatia, I had no PPDA, no zone indices, no analysis time. I had only footage and my eyes. And I learned that the eyes must work before, not after, the data pipeline.

The irony is that I myself made the opposite mistake when I was young, and that mistake taught me a whole career.

In 2026, when I was sixty-one, I served as a data adviser for a club in the second division with promotion as its goal. I built an xG model and advised the coach to push the line high in a play-off match. The model predicted my team would reach 1.8 xG, enough to win a match in which, according to the model, we needed only a bit of finishing luck. We lost 0-2. The opponent deliberately played counter-attacking defence, dropped deep, and turned every one of our shots into harmless long-range efforts from outside the box.

My mistake was not in the model. My mistake was that I read only total xG without splitting it by zone and without cross-checking against footage. I ignored the PPDA index and the shot-origin positions — two things that could have told me the real story.

From that day, I set myself a professional mantra: The model is not wrong, I was wrong to let it speak for my eyes.

That mantra led me to a conclusion I believe is truer than any index: in sports analysis, the highest value is not the correct number, but the boundary between what can be asserted and what must be said to be unassertable.

Eight years later, I applied that conclusion to one of the discoveries I am most proud of: the analysis of Roberto Mancini's Italy at Euro 2026.

At the time, everyone looked at Italy through the lens of pressing. But I was interested in something else: the average distance between positions on the pitch. I found that Italy controlled matches by compressing space horizontally, not by pressing continuously. They had the highest rate of symmetric flank circulation in the tournament: 18.3 times per match. It was precisely that stretching of the opponent horizontally and then switching flanks that created gaps for midfielders to break through.

I predicted Italy would reach the final from the group stage, and an international data magazine published my piece. But what I remember most is not the correct prediction. It is the feeling when I finished writing and realized I had not relied on a single index to assert anything. I relied on a chain of evidence, each piece cross-checked against context, and I left at least two scenarios open if that chain broke.

Then the pandemic arrived, and it taught me a lesson no index could teach.

In 2026, when every league stopped, I was thirty-nine and served as a data adviser for a club during the lockdown. Management asked me to predict form after football returned. I used data from the first fifteen rounds to build a model and advised the team to maintain a possession-based style. The result: the team lost three straight matches when the league resumed.

Opponents exploited the empty stadiums to press harder, causing my team to lose the ball in their own half. My model lacked two variables it could not know: spectators and on-pitch distancing. I had to admit that my data had expired.

I wrote a short piece titled "Data can speak, but it needs to listen," and changed my writing style: from then on, every analysis must contain at least two scenarios instead of one assertive conclusion.

The pandemic taught me that data too knows fear — when the world stops, figures are meaningless.

Those three lessons — Croatia, the play-off, the pandemic — converge into what I want to say about the empty file I met at four in the morning today.

In badminton, the empty-input problem appears more often than people think, and it appears at the most important moments.

Let me speak about the data gaps that any badminton analyst has faced.

The first gap is the historical gap. Badminton has a structural obstacle: most detailed data exists only since tournaments installed automatic tracking systems. Earlier matches — including legendary matches of earlier generations — exist mainly through footage and memory. When I analyse a young player today, I have hundreds of data points. When I compare him to a legend from twenty years ago, I have only low-resolution footage and scattered notes. Placing those two datasets side by side is a methodological insult, because they do not share an origin and do not share collection conditions.

The second gap is the condition gap. A badminton index means something only when you know the conditions under which it was measured: whether the arena has air conditioning, which shuttle model was chosen, what the humidity was, whether the court lighting causes glare. A player's average smash speed at an arena without air conditioning is significantly lower than that same player at a climate-controlled arena. If you place those two numbers side by side without context annotation, you have created a false story about a player's progress or decline.

The third gap — and the most serious — is the process gap. This is exactly what I met in my empty file. Every data pipeline is designed to handle non-empty input, not empty input. When the input is empty, the system does not report an error. It returns an empty result, and that empty result flows into the next steps of the analysis chain as if it were a fact. An inexperienced operator will not realize he is analysing nothing; he will only see empty tables and assume that is a normal result.

This is the most dangerous blind spot of an entire industry: we have complex systems to recognize truth, but no system to recognize the absence of truth.

I have seen the consequences of this in pre-tournament tactical sessions. A team can spend three days building a plan to counter a player based on data from an unreliable source. A coach can be persuaded by a miscomputed index because no one rechecked the source. And in a fierce badminton competitive environment, where a single point can overturn a career, such confusion has a price.

There is a story I rarely tell. At a tournament in Southeast Asia, a team prepared a plan for a semifinal based on the assumption that their opponent had a higher-than-average net-error tendency. That number came from an old data file, collected at a different tournament, with a different player in the doubles lineup. When the match took place, the opponent in fact did not net much at all; on the contrary, they played extremely securely at the net. My team lost, and in the post-match analysis room, no one dared face the truth that the mistake did not lie with the players on court.

The mistake lay in the grey zone between number and context — where data looks reasonable but has no basis to exist.

This is where I want to state a view I have held for twenty-nine years of observing the industry and sometimes express more bluntly than usual. In badminton, no data sample allows you to despise intuition. This sounds paradoxical coming from a data analyst. But I believe the eyes of an old coach watching a player perform a serve motion can capture what no camera records: hesitation, a change in breathing rhythm, a slight shift of weight in the back foot.

Figures are prayer, but intuition is the candle. I light both every time I read a match, and when both flames waver at once, I learn to stop.

The true value of a player lies not in the hardest smashes, but in the space he occupies when he does not hold the shuttle, in the position he chooses after losing a point, in the moment he decides to pull back rather than lunge forward. These things are hard to measure with existing indices, and precisely because they are hard to measure, they are often ignored in modern analysis.

A good analyst must periodically watch footage muted, turn off all data overlays, and try to answer a single question: what is this player trying to do?

That question has no answer from aggregate data. It only has an answer from the human eye.

But I must say the opposite too, because honesty does not permit me to take one side. Intuition can also be wrong, and it often errs in frightening ways. Intuition is built from memory, and memory is selective. A coach may forever remember one net error by a player and form a lifelong prejudice. An expert may remember one defeat and attribute it to a single cause, then apply that cause to every subsequent match. Unchecked intuition becomes prejudice, and prejudice in elite sport is a tax on player development.

So where is the boundary? The boundary lies in this: data answers the question "how much," intuition answers the question "who." A dataset can tell you what percentage of net drops a player misses after the third game. It cannot tell you why his hands shake in that third game. And to answer the second question, you need something no table can contain: an understanding of the person.

Now I want to return to the subject of this article in a more systematic way, because I believe the story of the empty file is not merely a professional anecdote, but a sign that the badminton world needs to rethink its relationship with data.

Let me begin with a fact rarely mentioned: badminton is a sport with a higher scoring density than any other. A match can stretch over three hundred rallies, each rally containing dozens of technical decisions. This means badminton has enormous data potential. But that potential also means enormous noise. In such a large pile of data, one easily finds a beautiful correlation between any two variables, and turns that correlation into a tactical story.

I have witnessed this in my consulting work many times. A team can find that their wins correlate highly with the number of short serves, then conclude that short serving is the key to victory. But that correlation may reflect something else: they serve short more when leading, and when leading they win more easily. Short serves do not create victory; they are a sign of being in the lead.

This is a methodological error so common as to be worrying, and it exists at every level of the badminton world, from national teams to small academies.

Earlier I mentioned three data gaps. In this section, I want to go deeper into the mechanism that creates them, because only by understanding the mechanism can one find prevention.

The first mechanism is source-selection filtering. When source data is lacking, the analyst tends to use the most accessible sources: aggregate articles, rankings, old reports. Those sources have varying reliability, but they are treated equally in the analysis process. A ranking compiled by an industry outsider and an official federation ranking both become "sources" in a data file, though their evidentiary value differs by a chasm.

The second mechanism is chain collapse. In an analysis pipeline, each step depends on the previous. If the first step returns empty, all subsequent steps still run and return empty results with a valid shape. This is what I call silent collapse. It is dangerous because no sound signals it. No alarm sounds. No red warning. Only a result that looks normal and contains nothing.

The third mechanism is time pressure. During the transfer window and the season, the pressure to produce a conclusion before a deadline makes the analyst tend to fill gaps with speculation. A report that is wrong but looks complete is still better than a report admitting it has nothing to say, because in our culture, emptiness is treated as failure, while inaccuracy is treated as a fixable mistake.

These three mechanisms explain why the badminton world is full of stories told from very little evidence. And they also explain why silence is sometimes the most honest answer.

I want to spend the next section on an aspect badminton analysts rarely face: structured uncertainty.

In statistics, one distinguishes two kinds of uncertainty. Random uncertainty is the kind that cannot be reduced even with more data — for example, where a serve lands within a region. Systematic uncertainty is the kind that can be reduced if you collect more data correctly. In badminton, these two kinds of uncertainty mix, and people often mistake systematic uncertainty for random uncertainty to avoid work.

For example, when a player has a low win rate against a specific opponent, people often say it is "fate" or that "this opponent is his bogey." This is a way of turning an analysable problem into an un-improvable destiny. In reality, a low win rate against an opponent often reflects a specific tactical blockage that can be identified: an awkward serve type, an unfavourable movement pattern, a rhythm weakness in a series of matches. These factors can be studied and corrected.

In other words, the phrase "bogey opponent" is a way to stop thinking.

A good data analyst must constantly ask: in this portion of uncertainty, how much is irreducible and how much is merely unreduced? This question is hard to answer, but it is the question that distinguishes the true professional from the storyteller.

I remember one time working with a team in Southeast Asia. Their head coach believed that one of his players had reached his technical ceiling. He told me: "He plays as well as he can. Nothing more can be improved." I watched footage for three days, then I noticed something unusual: this player lost points mainly in long rallies, not short ones. This meant his problem was not technical but aerobic fitness and focus in long rallies. This was an improvable problem, but no one improved it because no one saw it. They attributed a solvable problem to an insurmountable limit.

That is the direct consequence of ignoring context analysis.

I also want to speak about another aspect of modern badminton rarely mentioned in data analyses: the shifting of centres of power.

Over the past twenty years, the map of world badminton has changed in ways that simple indices cannot capture. Japan rose as a power through systematic investment in sports science. India emerged as a new centre with characterful players. Taiwan and South Korea maintained strength in specific disciplines. Europe, through Denmark and Spain, produced players with a completely different style from Asian players.

Each of these centres defines success differently. For China, success is stability at the system level, expressed through medal counts in team events. For Indonesia, success is the explosion of players who can single-handedly change the course of a match. For Japan, success is sustainable development across generations. These definitions cannot be compared with each other through simple numbers, and any direct comparison is a cultural distortion.

In the China–Indonesia data corridor where I worked, I saw fiery debates over the question: who produces better champions? There is no answer. China produces more stable champions, Indonesia produces champions more capable of creating miraculous moments. They are two different kinds of champion, serving different purposes of the sport. Ranking them is an attempt to impose a single measure on values that cannot be measured with the same yardstick.

The lesson I want to draw from this is: when you place two sets of data from two different cultures side by side, you are not comparing them. You are writing a story. And that story may be good, but it is not analysis.

Now I want to speak about what I consider the centre of this entire subject: the sacrifice of accuracy in exchange for appeal.

In sports media, there is a constant pressure to turn every observation into a compelling story. A player whose form dips due to injury is not told as "injured"; he is told as "lost himself." A team that loses because the opponent played well is not told as "the opponent played well"; that team is told as "mentally collapsed." These stories have great emotional appeal, but they do not help the reader understand the sport.

I once wrote that way. In the early years of my career, I embellished every match into a tragedy or a glorious victory. Then I realized those pieces helped no one who was genuinely trying to understand the sport at a deeper level.

One of the reasons I pursued the data path was a desire to bring truth back. But I learned that data, when abused, can become an even more dangerous form of storytelling, because it wears the armour of objectivity.

In an empty data file, there is nothing to embellish. And that is precisely what makes it honest. The empty file forces you to face the truth that you know nothing at all. In an industry built on saying things you do not fully know, that silence is a rare gift.

The true value of a player lies where he runs and when he stops. This is true also for analysts. The true value of an analyst lies in what he writes and in what he refuses to write.

I want to close this section with an observation on how the badminton industry can learn from other industries in dealing with uncertainty.

In medicine, there is a concept called evidence-based consensus. A treatment protocol is only recognized when there is enough high-quality evidence from multiple independent studies. In aviation, every procedure has mandatory stop steps: if a key parameter cannot be determined, the pilot stops rather than continues.

Badminton has not yet developed similar standards. We have not built a culture in which an analyst is respected for knowing how to say "I don't know," rather than being considered weak for lacking a conclusion. We do not yet have mandatory stop procedures in data pipelines.

I believe building that culture is the next great challenge of modern badminton, greater than improving player pay or expanding the number of tournaments.

Back to my empty file. After many hours facing it, I decided that what I had to write was not an analysis, but a piece about why there was no analysis to write.

That decision sounds minor. But for someone with my temperament — fond of order and results — it was a milestone. I had spent nearly thirty years in the industry building systems to help me reach conclusions faster and more surely. And that night I had to admit that sometimes the best system is one that knows how to say: "I do not yet have enough information to make any judgment about this subject."

I want to emphasize that this admission is not surrender. It is an act of discipline. In every field that uses data to make decisions, the most important discipline is not the discipline of collecting more figures, but the discipline of not concluding when evidence is insufficient. An analyst without that discipline is like a pilot without a pre-flight checklist: he may fly well, but one day he will fly without fuel.

I do not want this article to end with a warning or an advice. I want it to end with a more open question.

When the empty file sat on my screen, I asked myself: if badminton built systems capable of recognizing the absence of data, what would change? Would we produce fewer false stories about players? Would we stop misattributing failures to mentality and ignoring fitness? Would we stop comparing legends of different eras as though they competed in the same arena?

And what would happen if we ourselves — the analysts — were also entities in some data file? If someone were measuring us with indices we do not know, which index would speak the truth about our work? Number of articles published? Number of shares? Or the number of times we dared to say we do not know?

I leave that question open. Because after nearly thirty years in the profession, I have learned that an open question is kinder than a confident but wrong assertion. And also because I believe in my model, but I still pray before every match. Not because I believe in gods, but because I believe there is an unfillable gap between what can be measured and what can happen.

That gap is the sport. That gap is why we return to badminton every week, even knowing that no model can fully predict a shuttle flying over the net in the darkness of a crowded arena.

Now I want to spend the final section of this article on a topic I have touched on in many parts but have not given its due attention: the relationship between badminton and esports in the context of data analysis.

There is a strange resemblance between these two worlds that analysts often ignore. Both are sports with extremely high event density: in a thirty-minute badminton game, there are hundreds of technical decisions; in a forty-minute esports match, there are thousands of tactical decisions. The tempo of the match never lies, and both sports record that tempo in structurally similar data files.

I have spent time studying the esports world over the past few years, and what I realized is this: tempo is one of the few indices that can migrate between these two sports without losing meaning. A badminton team whose tempo is controlled by the opponent will have longer average rallies, shorter rest between rallies, and slower decision speed. An esports team whose tempo is controlled will have fewer average actions per minute, longer response time after key events, and a higher map-control loss rate. These indices are not alike in form, but they share the same essence: they measure who is imposing tempo on the match.

However, there is a fundamental difference in ethics and governance that I believe badminton must learn from.

In esports, betting is eroding competitive integrity faster than in any traditional sport, because betting regulations in esports lag behind the industry's growth rate. Esports tournaments can be organized within weeks, with hastily written rules, and loopholes can be exploited before regulators respond. Meanwhile, badminton has gone through decades of building regulatory systems, and although imperfect, it is slower and has more layers of protection.

The lesson badminton can draw from esports is not their analytical technique, but the speed at which a young industry can collapse when regulations cannot keep pace with growth. This relates directly to a subject I mentioned above: when systems lack mandatory stop mechanisms, they keep running until they hit a wall.

In both sports, building a culture of responsibility is a precondition for healthy development. And in both, responsibility begins with admitting what we do not know.

I also want to add one thing about the relationship between data analysis and officiating in badminton, because these two fields share a resemblance I find interesting.

In recent years, badminton has adopted many officiating aids, including shuttle-landing recognition systems and video review tools. These technologies, like VAR in football, were expected to reduce controversy and increase fairness. But in my experience observing officiating decisions, the opposite often happens: technology does not erase controversy, it shifts it from the court into the review room and into the grey zones of the law.

A shuttle landing near the line can be determined by technology to have touched the line, but the question of whether the umpire should review the situation remains a human decision. A net touch can be replayed from many angles, but the question of whether the player touched intentionally remains a matter of subjective judgment. Technology cannot resolve questions of intent, and most badminton controversies involve intent.

This has direct meaning for the subject of this article. In both data analysis and officiating, the stronger the technology, the more the grey zones are brought into light — but light does not automatically resolve those grey zones. It only lets us see them more clearly, and sometimes seeing a grey zone more clearly makes the controversy fiercer, because everyone has evidence to defend their position.

The lesson I want to draw is: better technology does not mean better decisions. A system with more data does not mean more accurate analysis. And an umpire with more camera angles does not mean a fairer umpire.

In every case, what determines final quality is not the volume of data, but an understanding of context and the ability to make a judgment when data conflicts.

Now I want to return to an aspect I consider most important for readers interested in badminton as an elite sport: how a match-watcher can develop the ability to read a match.

Over the years, I have taught a small class for those who want to learn badminton analysis, and I always begin with a lesson students find strange. I ask them to watch a match without sound and without any data overlay, and to try to write down ten things they observe. Then I ask them to rewatch that match with full sound and overlay, and write down ten new things.

The result is always similar. The ten things from the first viewing concern what the players do. The ten things from the second viewing concern what the numbers say the players did. And there is often a significant difference between the two lists.

I tell my students: that difference is the starting point of all analysis. If you cannot see the difference between what you see and what the numbers say, you are not an analyst. You are a number-reader.

And a number-reader cannot detect when a number is wrong.

I believe the most important skill of a modern badminton analyst is the ability to grasp the tempo changes of a match. A badminton match, like a piece of music, has silences and crescendos. A good analyst can hear when a player begins to run out of breath, when a player changes serving tactics, when a player begins to shrink in long rallies.

These signals are not in the summary table. They are in the gaps between numbers.

And to read them, you need something that cannot be generated from data: attention.

I want to close this analytical section with an observation about the future of badminton analysis.

In the next ten years, badminton will have more data than anyone can imagine. Cameras will track every movement of a player. Sensors will measure every breath. Algorithms will predict every serve. And in that data-rich world, the greatest challenge will not be collecting data, but knowing when to stop trusting it.

The best analysts of the future will not be those with the most data. They will be those best able to distinguish truth from appeal, signal from echo, what can be asserted from what must be admitted as unassertable.

And among them, I hope there will be those who dare to face an empty file without filling it with stories.

That is the future I want to see. That is the future I am trying to build, one article at a time, in a small room in Surabaya, while outside the sky is still dark.

And tonight, looking back at the empty file on my screen, I no longer see it as an error. I see it as a reminder that, after nearly thirty years in the profession, I still have one thing to learn: how not to say anything when there is nothing to say.

That is perhaps the hardest lesson of all. And it is also the lesson I believe the badminton world will have to learn, whether it wants to or not, as data grows ever larger and truth becomes ever harder to find.

The value of a truth lies not in how loudly it is asserted, but in how it is not distorted when diagnosed by numbers that lack sufficient basis. And in a sport decided by brief moments, where a player can become champion or collapse within a single rally, honesty about what is unknown may be the most valuable quality we can keep.

I will continue to look at empty files. I will continue to ask questions when the numbers say nothing. And I will continue to believe that, in the end, the winner is not the one with the best model, but the one who knows when his model has reached its limit.