The Empty Lane: When Vietnam's Swimming Data Is Not Enough to Conclude

9:40 p.m. I sat in front of the screen and opened the data file of a national swim meet that had just ended. The leaderboard was complete. The athlete names were complete. But the first 50m split column — the column that decides how to read a 200m freestyle race — was blank. Not a typo. Genuinely blank. And in that moment I understood that every conclusion I could draw from that file would be fabrication, unless I accepted a single sentence: the data is insufficient to conclude.

That sentence is harder to say than people think. In Vietnam's sports analysis industry, silence in front of an empty cell is read as weakness. People want a forecast. They want a name, a number, a verdict. No one pays an expert to hear that he does not know. But the very moment I saw that empty column, a familiar warning fired in my head: if I fill the empty cell with guesswork, I am no longer an analyst. I become a storyteller, and the story I tell will be my own story, not the race's.

This piece is not a complaint about one broken file. It is about a chronic disease of Vietnamese swimming, and that disease is quietly shaping the medals we have not yet won. Every shock has its own probability. We call it a shock when we have not yet checked the tables. But when a page is torn out of the tables, everything that remains becomes a shock — including the things that should have sat inside the forecast for years.

Context: A swimming nation rich in medals, poor in data

I grew up with Vietnamese swimming from 2026, when I started my career at a sports newsroom as a swimming reporter. Back then we wrote results into notebooks with a ballpoint pen. Everyone had a notebook. Each notebook was its own data store, and no two matched. Twenty-two years later, I still see the image of those notebooks inside every swimming data system I touch. Technology changes, but habits outlive any model's forecast.

After two decades, Vietnamese swimming has produced names the whole of Southeast Asia has to mention. Nguyễn Thị Ánh Viên is tied to an entire generation of SEA Games achievement. Nguyễn Huy Hoàng left a mark at continental level. Young swimmers such as Trần Hưng Nguyên, Phạm Thành Bảo and Võ Thị Mỹ Tiên are steadily claiming their places. That is the tip of the iceberg. But below the waterline, where data must live, we still lack the most basic things: a standardized split system, continuous year-by-year athlete records, and an archive tight enough that anyone — a reporter, a national-team coach, or a data consultant like me — can trace a specific race back.

When I worked at a private sports data company and later became a consultant to football clubs, I noticed a counterintuitive thing: swimming is the sport with the largest volume of raw data of all. A single 200m race generates hundreds of measurements if the system is complete: reaction time off the blocks, underwater speed after the dive, dolphin-kick count, time per stroke, stroke rate, distance per stroke, turn-split times, finish speed. One 200m race can spawn more than 40 separate variables.

Yet swimming is the least digitized sport in Southeast Asia. That paradox does not come from a lack of equipment. It comes from a lack of process. No one signs a document forcing a meet organizer to record splits. No one keeps the original of the timing system. No one cross-checks the electronic file against the paper record. The result is that we have results, but we lose the process that produced them — and that process is precisely what can forecast the future.

Core: The anatomy of a race and the death of split data

Let us put the empty file aside for a moment. I want you to see why splits matter so much, so that when they disappear you understand what we lose.

A 200m freestyle race, at international level, is not swum. It is built. Coaches divide the race into tactical segments: the first 50m, the second 50m, the third 50m, the final 50m. Each segment has its own function, and that function shifts according to each swimmer's physical character.

For a typical 200m sprint swimmer, the ideal race takes this shape: a fast opening segment to seize position before encountering a teammate's wash, a slightly slower second segment — the 'rhythm zone' — a third segment beginning to accelerate again, and a fourth segment as an absolute sprint. The gap between the fastest and slowest segment for an elite swimmer can be just 1.5 to 2.5 seconds. But that gap decides who wins and who finishes fifth.

Splits do three things that a final time can never do.

First, splits expose physical structure. If your fourth segment is more than two seconds slower than your third, that is not a technical problem. It is a problem of energy allocation and lactic-acid resistance. It tells you what to train, how many sessions, and for how long.

Second, splits reveal hidden technical faults. When distance per stroke drops mid-race while stroke rate rises, that is the classic signal of arm-muscle fatigue. The swimmer compensates by spinning the arms faster, but each stroke is shorter, and efficiency falls. And distance per stroke can only be computed if you count strokes within each segment. Without splits, you cannot make that division.

Third, splits build the prior probability for the future. When I forecast that a swimmer will break a personal record at an upcoming meet, I am not relying on feeling. I am relying on the average improvement rate of each segment over the past twelve months. A swimmer improving their finishing segment faster than the others is a swimmer growing in the right direction. A swimmer improving the opening segment while the final segment stalls is optimizing the wrong link.

Now return to my empty file. Without splits, all three possibilities collapse to zero. I do not know the athlete's physical structure. I do not know whether technique is stable. I have no prior probability. And yet I am still expected to write a forecast.

The simple amateur question I ask before every table of numbers: if this number vanished, what could I still say? If the answer is 'nothing', then that number is the backbone, and I must state clearly that it is missing. Splits are the backbone. The final result is only the skin.

Three data layers, and which one is collapsing in Vietnam

I divide swimming data into three layers.

Layer one is output data: finish times, results, rankings, medals. This layer we have. And have it well. National meets, SEA Games and ASIAD all publish complete final results. A ten-year-old can look up who won the 100m breaststroke at a SEA Games.

Layer two is process data: splits, reaction times, distance per stroke, stroke rate, underwater kick counts, turn speed. This layer we have, but unevenly. Some meets preserve it, some lose it, and there is almost no standardization between meets.

Layer three is physiological data: training load, high-speed running distance per week, fatigue indices, recovery cycles, sleep and nutrition variation. This layer is nearly empty in Vietnam, except at a few large centers that use monitoring devices.

The most serious problem is not that layer three is empty. The problem is that layer two is collapsing, while the whole sport behaves as though we live in the full layer.

When a coach says 'this swimmer finished slowly because of weak mentality', that is a layer-three judgment issued from layer-one data. No recovery index, no load data from the previous week, no split analysis from the last three races. The judgment sounds reasonable, but it is built on nothing verifiable. It is a hypothesis in the costume of a conclusion.

I once saw exactly this trap in the GPS data of a Saigon club in 2026. Reviewing players' high-speed running distance, I found a pattern: before a muscle injury occurred, high-speed running distance rose by about twenty percent over a short window. That was not the direct cause of injury. It was a signal that the player was compensating for fatigue by running more at high intensity. The injury came later.

That pattern only surfaced because we had layer-three data. With only match results, we would forever talk about injuries as accidents. But injuries in elite sport are rarely pure accidents. Most are the endpoint of an accumulation that data saw weeks in advance.

In Vietnamese swimming, we are analyzing swimmers' physical structure by feel, in a sport where feel is systematically deceived by water. Water feels gentle even when the body is at overload threshold. That is why this sport needs numbers more than any other.

Reading a race with and without data

To keep this concrete, I will build a controlled comparison. I will not assign specific numbers to a real athlete, because that is exactly what I am condemning. Instead I describe the structure of two ways of reading a 400m freestyle race — a race where pacing allocation is decisive.

The 400m freestyle is the harshest pacing test in short-distance distance swimming. It is four times longer than the 100m, and if you swim the first 100m too fast, the rest will pay with what I call the 'segment collapse effect'.

With complete data, I read a 400m race through three indices. The first is opening-segment deviation: the gap between the first 100m and the average speed of the remaining 300m. The second is the decay slope: by what percentage speed falls after each 100m. The third is the ability to recover rhythm in the 250-350m stretch, the point at which most swimmers begin to fade.

These three indices let me sort swimmers into physical archetypes: the 'even' type, the 'strong finisher' type, the 'fast starter who cannot hold it' type. Each archetype needs a different training plan. A 'strong finisher' needs to raise mid-race endurance. A 'fast starter' needs to learn restraint in the opening lap.

Without splits, these three indices vanish. I am left with the final time. And with only the final time, I am no longer analyzing. I am left with a choice between two attitudes: either say 'insufficient data', or invent a plausible story about tired legs or weak spirit.

In Vietnam, the second attitude is systematically winning. Not because reporters deliberately lie. But because the pressure to have a story every day exceeds the pressure to have data. A news brief filed in fifteen minutes cannot wait for a two-week verification process.

This is the mechanism of unconscious fabrication, and it deserves to be named.

The mechanism of unconscious fabrication

I once wrote that a stroke appears once, but its trajectory lasts for years. That holds true for analytical data. A race with a broken split today can become a wrong conclusion in a coaching curriculum three years from now. A groundless judgment about 'weak mentality' can cost a young athlete the chance to test themselves at a bigger stage.

Unconscious fabrication follows a predictable process. I have watched it enough to describe it in four steps.

Step one, an appealing observation appears. For example, a young swimmer's results jump at a meet. Step two, a single cause is attached. For example, the athlete changed coaches, or changed technique, or raised training volume. Step three, a story is told persuasively enough to spread. Step four, the story becomes accepted truth because no one checks it.

The frightening part of step four is that no one actively refutes it. A wrong forecast carries no immediate penalty in sport. No one sends an invoice for a bad prediction. If the forecast booms, the forecaster is praised. If it fails, attention moves to the next meet. The result is that the self-correcting mechanism of the sports industry is extremely weak compared with an industry like finance, where you lose money when you predict wrong.

Data analysts are entering the dressing room, and their conclusions often detach from the rhythm of reality. But the reverse is also true: the dressing room is entering data analysis, and when there is no table of numbers to brace against, a coach's feeling becomes the model. That is why I never praise a tactic without pointing out its operating conditions. A good technique in one condition can backfire in another, and the split table is where operating conditions are tested.

Boundary conditions: when the stands, the pool and the calendar become variables

One of my biggest lessons came in June 2026, when I was 31 and working as a data expert for an online football site in Saigon. The U20 World Cup in South Korea was the milestone when I used an expected-goals model to analyze Vietnam U20's matches.

Across all three group games, the team generated a total of 2.1 expected goals, but scored only one, from a free kick worth merely 0.08. I pointed out that if they maintained high pressing against a strong opponent, they could spring a surprise. But poor chance conversion, plus missing data on shot quality at youth level, meant my conclusion could only take a probabilistic form: about a one-in-three chance of advancing if they held their defensive structure across the final two games.

That article taught me three things I carry into swimming.

First, every shock has a prior probability. Do not call any result a surprise before checking the historical data. In swimming, this means a young swimmer beating a veteran for the first time is not a surprise if their improvement rate over the previous six months already crossed the threshold. The prior probability sits in the table, not in the spectator's feeling.

Second, dissect boundary conditions before analyzing any race. For swimming, boundary conditions are not just home pool or crowd. They include water temperature, lane-rope height, number of head breaks after the dive, the day's schedule, the gap between heats and finals, and water quality. A small shift in a boundary condition can change the entire outcome of a 50m race.

I once saw a domestic meet where water temperature was about two degrees Celsius below competition standard. Sprinters were mildly affected, but distance swimmers were hit harder because the body spent more energy maintaining temperature. Without recording this parameter in the minutes, every comparison between races at that meet and other meets is a biased comparison. That is why I say: when boundary conditions are eliminated or unrecorded, the whole analytical model collapses into a number near zero.

Third, I learned to present forecasts as probabilities, not absolute assertions. After the 2026 tournament, every piece of mine carried a section on 'conditions for the prediction to hold', helping readers understand the nature of the data and avoid misreading. In swimming this matters especially because 50m and 100m races carry very high noise, while 400m and 800m races are more stable. Applying one forecasting standard to both groups is statistically wrong.

Evidence chain: from a single race to an Olympic cycle

One of my principles is never to analyze a single match but always to extend the observation window. In swimming this is even truer, because the international competition cycle lasts four years and each race is a data point on a long curve.

Picture a Vietnamese swimmer aiming at an Olympic Games over four years. I will track them across at least four continuous indices.

The first is average improvement rate by quarter. A swimmer improving steadily by about one percent per quarter accumulates about four percent per year. If their gap to the Olympic standard is five percent, that trajectory is enough to predict a ticket within two years. If the gap is twelve percent, the model says a structural change is needed, not a volume change.

The second is segment stability. A swimmer whose gap between best and average race is low is one who can control pace under pressure. High stability is a more valuable asset than a single personal record, because big meets are decided by repeatability.

The third is biological age versus competition age. For men, peak in sprint events usually falls around 22 to 26. For women, peak comes earlier and can hold to about 24 to 27. If a 19-year-old has not yet hit the Olympic standard but their improvement trajectory points the right way, dropping them from a long-term plan is a mistake based on a wrong standard.

The fourth is load tolerance across a dense competition sequence. At a SEA Games, a swimmer may race four to six events in a few days. That is a test of recovery, not only speed. If recovery records are not kept, we cannot know whether a drop in performance is due to cumulative load or a single technical issue.

These four indices form an evidence chain. With enough of them, I can draw a trajectory. With one or two key indices missing, I must narrow the scope of my conclusion. I once missed a chance to consult for a professional team because I delayed to perfect an injury-forecasting model. That perfectionism cost me a contract I considered valuable. But it is also why I do not invent an Olympic prediction for a swimmer when I lack the four indices. I would rather lose an opportunity than tell thousands of people something false.

Counterintuitive angle: correlation is not causation in the pool

Now I reach the part I consider most important in this piece, and also the one most easily misunderstood.

In swimming analysis there is a kind of conclusion that sounds highly professional but is in fact pseudoscience: causal conclusions from correlation. I call it the trap of the pretty number.

For example. A swimmer raises stroke rate from 38 to 42 strokes per minute over a season and improves their 100m freestyle markedly. The conclusion sounds compelling: higher stroke rate leads to better results. But it may be wrong. Perhaps the swimmer simultaneously cut underwater kick rate, or improved turning at the wall, or entered the peak phase of the training cycle. Higher stroke rate and better results may be two independent variables both driven by a third: better propulsive force.

Before saying 'X leads to Y', I must point to a concrete physical or behavioral mechanism linking the two variables. In swimming, the mechanism must lie in fluid mechanics and muscle physiology. If I say higher stroke rate leads to better results, I must explain why that rate raises speed without cutting distance per stroke enough to destroy efficiency. If I cannot explain it, I use the word 'associated', and note that the link has not been established as causal.

This is harsh linguistic discipline, but it is necessary because Vietnamese swimming lacks a cross-verification system. In finance, independent bodies audit every profit claim. In swimming, no one does that. If I do not audit myself, no one will audit me.

Another counterintuitive angle: peak performance usually comes from simplification, not complication. I have seen many teams cram too many indices into a forecasting model until the model becomes uninterpretable and unmaintainable. A model with five variables you understand well beats a model with fifty variables you do not. The biggest shock does not come from too little data. It comes from too much data with no structure.

I once worked with a swimming data system that logged over a hundred indices per race. But when I asked about standardized 50m split data, the answer was that the system did not record it. They logged a great deal, but missed exactly the number that was needed. This is the common tragedy of sports data systems in developing countries: abundance at the display layer alongside poverty at the foundation layer.

Error prevention: a four-layer verification process

Since I do not want this piece to stop at a warning, I will describe the process I propose for any body that wants to build a serious swimming data system.

The first layer is collection. Every race must be captured electronically with timestamps, not just the final result. Every wall touch must have its own time. This is the minimum requirement, yet in Vietnam many meets still stop at the overall result.

The second layer is standardization. Data from every meet must be converted to a common format, with the same units and the same athlete IDs. If a swimmer changes name in the records, there must be a traceability mechanism so previous races still link to the right person. This is a small technical issue with large consequences for long-term analysis.

The third layer is cross-verification. Every important number needs at least two independent sources. Electronic results must be checked against the referees' paper record. Splits must reconcile with total time. If there is a discrepancy, there must be a clear handling process, not silent editing.

The fourth layer is open archiving. Data must be stored in a form retrievable years later, even if the original collecting body no longer exists. This is what Vietnamese swimming lacks most severely. A great deal of valuable data from earlier generations has been lost simply because no one kept the original file.

These four layers do not need expensive technology. They need discipline. And discipline is the cheapest thing to build and the most expensive to maintain.

I have built forecasting models for football for years, and data discipline there is much higher than in swimming because football has money from betting. Vietnamese swimming has no similar financial driver, but it has a larger one: every Olympic or ASIAD medal can be the result of a correct coaching decision, and that decision needs data. If we build a serious data system now, in ten years we will have a generation of swimmers guided by probability, not by belief.

A note on the limits of the model

I must confess something many data analysts dislike confessing. There are parts of a race that data cannot explain.

In my model, the crowd, pressure and emotion are often treated as variables that can be switched on and off. But when a meet carries large social meaning, or when a swimmer competes in front of family and hometown, unexplained variance rises markedly. I cannot precisely measure what percentage a swimmer loses from nerves, just as I cannot measure what percentage a swimmer gains from cheering.

What I can do is record a confidence interval. When I forecast a swimmer's result at a big meet, I give a range, not a point. That range widens when psychological conditions are abnormal. This is how I acknowledge the model's limits without abandoning data discipline.

At a national championship, where the atmosphere is usually less tense, the interval can be narrow. At a home SEA Games, it must widen. At the Olympics, it is widest, yet it is also where a small error is worth an entire four-year cycle.

A tactical era dies when no one reads its data table any more. That is true of football, and it will be true of Vietnamese swimming. When a new generation of coaches grows up with split data in hand, they will no longer accept judgments based on feeling. That is an irreversible shift, and it is arriving more slowly than our potential deserves.

Match-watching experience: reading the race to understand the years

Ordinary people look at a medal to understand the race. I look at the race to understand the years.

In more than twenty years of watching swimming, I have learned that the real story of an athlete is not in their best race. It is in their worst. The best race tells you the ceiling of ability. The worst race tells you the foundation. And in an Olympic cycle, the winner is usually not the one with the highest ceiling, but the one with the most stable foundation.

Based on my experience watching matches and races, I have found one thing repeating. Swimmers who make a marked jump in results usually have a quiet stretch before it. They do not improve evenly. They stall for months, then explode. That quiet stretch is not a sign of failure. It is a sign of restructuring. But to see the quiet stretch and the explosion as a causal chain, you need continuous data. If you only have results per meet, you will see the explosion as a miracle.

I sit far from the pitch to see the match more clearly than the referee. That is true of football, and it is true of swimming. The person closest to the race is the swimmer. The person with the most holistic view is the one far away, with data in hand.

In 2026, when the pandemic paralysed world football, I reviewed the GPS data of twenty-nine players at a Saigon club. The stadium was empty but the data kept running. I found high-speed running distance rose by about twenty percent before a muscle injury occurred. I proposed a load-reduction process splitting training into four stress thresholds, and when the season returned the club recorded a significant fall in injury cases versus the previous season. That episode taught me that data does not need a crowd to speak. In swimming, when the pool is empty, data is the only thing still telling the truth.

Since then, I have brought physical indices into every technical analysis. I do not see a race as just one minute or four minutes. I see it as a link in a long chain of movement, where mistakes usually appear weeks earlier. A weak arm pull in the final 50m does not begin in the final 50m. It begins in a training session three weeks earlier, in a decision to raise volume at the wrong time, or in a recovery regime traded for more sessions.

Football and esports do not differ in essence, only in reflex tempo. That line of mine is often misread. I am not saying they are technically alike. I am saying they are alike in the data structure needed to evaluate talent. And swimming belongs to that family too. The career of an esports player is far shorter than a footballer's, yet their youth-development and post-retirement support systems are near zero. Vietnamese swimming has a similar problem in the opposite direction: longer careers but a data system to extend those careers that is nearly empty. Both are infrastructure problems, not talent problems.

Tactical blind spot: when coach and data do not speak the same language

There is a gap I want to name clearly. Coaches and data analysts often look at the same race but measure it in two different frames.

Coaches measure by feel from the poolside. They see a swimmer's shoulder tighten on the return lap. They see breathing rhythm change. They hear the water. This is real data, but it is encoded in a system different from numbers. Analysts measure by splits and stroke rate. They see the number drop, but not the body behind the number.

The common mistake on the analyst's side is to treat the coach's feel as noise. The common mistake on the coach's side is to treat numbers as detached from reality. Both are wrong.

The right way is to translate the two frames into each other. When a coach says the shoulder is tight, that is a testable hypothesis: does distance per stroke fall in the later phase, does time per lap rise, does stroke rate rise to compensate. If those three indices confirm, the coach's feel has been converted into data. If they do not confirm, that is a chance to revisit an impression.

This translation process can only happen when both sides accept that neither holds the whole truth. The analyst does not hold the whole truth because numbers cannot measure emotion. The coach does not hold the whole truth because feel cannot measure probability distributions. When both fall silent before a data gap, the system has a chance to become honest.

That is also why I return to the story at the top. My empty data file is not the failure of the file itself. It is the failure of a process that allowed the file to be exported without its backbone. And that failure belongs to no single person. It is a system failure.

Why a race is never its own present

I want to close the core analysis with an idea I consider the most important in all data thinking.

A race is never its own present. It is the convergence point of everything that happened before it: training tactics, psychology, opponents, schedule, health, nutrition, even weather. I never analyze a single race but always pull it back into a long chain.

When I watch a race and want to understand it, I ask three questions. First: what happened three weeks ago? Second: what will happen three weeks from now? Third: if this race were repeated in a different probability space, could the result be reproduced?

The third question is the hardest. It separates a real result from a fluctuation of luck. A swimmer can win once because their rivals swam badly. But a swimmer who wins three times in ten months under three different conditions is one who has built a foundation. In my model, foundation matters more than peak. And foundation can only be measured by tracking repetition, not by celebrating the moment.

That is why I say: a stroke appears once, but its trajectory lasts for years. When we build a Vietnamese swimming data system patient enough to track that trajectory, we will stop calling shocks shocks. We will call them by their real name: events with a prior probability, already seen in advance by the table.

A counterintuitive angle on expectations: don't ask who wins, ask why they reached the final

There is a question I am often asked over coffee with sports people: who wins this meet? I usually answer with another question. Do you want to know who wins, or why they are likely to reach the final?

These two questions lead to two entirely different kinds of analysis. The first demands a prediction. The second demands a trajectory. And in swimming, the trajectory is what helps a sport that is building a development system.

I have written many things since 2026, and some of them have by now become the headlines of other articles. What I learned from that is that a prediction can be right for the wrong reason. If I predict a winner and they win for reasons I did not foresee, my correctness has no value. It is a random hit. The real value of analysis is not in being right, but in being right for verifiable reasons.

This is something the sports industry usually ignores because no one pays for humility. People pay for confidence. But confidence with no prior probability behind it is a cheap commodity, and the domestic market is consuming far too much of it.

A forward thought: a cycle to build the data backbone

If I were given a small mandate to build Vietnamese swimming over the next four years, I would not set medal targets first. I would set data targets first.

Year one is collection. Every national meet, every youth meet, every selection trial must be recorded by an electronic system with standardized splits. No expensive technology needed. Only process.

Year two is standardization. Data from year one is converted to a common format, with a unified identifier for each athlete. Monthly continuous records begin for the pool of potential athletes.

Year three is analysis. Trajectory-forecasting models begin to have enough data to run. Training-load injury models can be calibrated.

Year four is action. Selection decisions, training-camp budget allocation and international competition plans are made on probability, not instinct.

Those four years need no genius swimmer. They need people willing to sign a document about data discipline. That is the kind of document that brings no glory, but it is the foundation for all glory that follows.

I do not know whether those four years will happen within my career. But I know one thing for certain. Every time I open an empty data file and choose to say 'insufficient data to conclude' rather than invent a story, I am signing a small signature into that foundation. A contract is not a signature, it is a hypothesis signed. And an honest hypothesis, even when it fails to satisfy today's reader, remains more useful than a compelling but false conclusion.

Numbers do not lie. Social media does. But when the numbers are absent, the only one who can lie is ourselves. The question left for the next cycle is not who will win the next national swim meet. The question is: do we have enough courage to look at the empty lane, admit it is empty, and start rebuilding from the first data cell.