Data Blanks in Badminton: From the 565 km/h Smash to the 1.15-Metre Stick

Data Blanks in Badminton: From the 565 km/h Smash to the 1.15-Metre Stick

An Afternoon Without Numbers

On the livestream of a BWF World Tour quarter-final in Europe, the television graphics showed everything: the player's latest smash speed, the average number of shots per rally, the scoring rate when attacking the net, a heat map of where the shuttle landed. Three weeks later, the same player competed at a Super 100 event in Vietnam. The screen showed one thing only: the score, updated point by point, with no accompanying metrics.

The distance between those two frames has nothing to do with the standard of the player. It sits in the measurement infrastructure.

I have recorded this observation in my notebook since the 2026 season, and it remains the first thing I check whenever someone asks why badminton is harder to analyse than football, despite having fewer variables on court. Football has 22 players, one ball, 90 minutes and hundreds of data companies competing over every decimal. Badminton has two people, one shuttle, and a public data system so thin it is genuinely strange.

That thinness has real consequences. It shapes how fans remember a player, how a national federation allocates its sports budget, how markets price an individual match, and how a young talent in Binh Duong or Bac Giang is perceived compared with a young talent in Copenhagen.

When every tournament stops, I finally hear my own pulse. But in badminton, there are tournaments that never made a sound loud enough to be heard as data in the first place.

The Measurement Infrastructure of a Global Sport

Every BWF World Tour match is logged in the World Badminton Federation's Tournament Software: the score of each point, match duration, umpire names, court, finish time. That is the first layer of data, present in every match, including the qualifying rounds of a Super 100 event in Ho Chi Minh City. But this layer answers only one question: who won, who lost, and by how many points.

The second layer is performance metrics: smash speed, shots per rally, front-court scoring rate, successful net approaches, unforced errors. This layer appears only when a venue has a shuttle-tracking camera system and a team encoding events in real time. The cost of installation, operation and staffing limits it to a small number of televised courts each year.

The third layer, the deepest and most closed, is workload and physical condition data: distance covered, jumps, heart rate, knee and ankle load after each match, required recovery time. This layer belongs to national teams and training centres and is almost never published.

Stacked together, these three layers form an inverted pyramid. The layer most important to an athlete's health and career is the one least visible. The layer most frequently shown on television exists for only a few dozen matches a year.

BWF once tried to close this gap at the level of the rulebook. In 2026 the federation proposed replacing the 21-point system with five games to 11 points, arguing that shorter matches would be more television-friendly and reduce the load on players. The proposal was rejected at the annual general meeting. What interests me is not the vote but the argument: both sides talked about audience experience and broadcast windows, and neither raised which system would produce cleaner data for evaluating the players themselves.

I walk into the church of data not to pray, but to listen to the noise of the truth. And the loudest noise in that room is the sound of everything that was never recorded.

The Small-Sample Problem of the 21-Point System

A badminton game ends at 21 points. A match usually lasts two or three games. The total points in an elite match typically falls between 75 and 95, split between two players. Each side touches the shuttle in roughly 40 to 50 decisive situations across the whole match.

Set beside football, where a team makes around 500 passes a match, forty-five sounds small. But the issue is not the number of actions. It is the number of independent observations an analyst can gather about a single player in one season.

Suppose a player competes in 20 matches a year, each around 45 points, roughly 900 points in total. That sounds like plenty. But if I want to test a narrow hypothesis, for instance how this player handles situations at 18-18 or later, the sample collapses to a few dozen situations per season. If I want to isolate 18-18 in a deciding third game, when the legs are gone, the sample falls to somewhere between ten and fifteen. At that size, a swing of one or two successes can reverse the entire conclusion.

I once built such a tracking sheet for a men's singles player across three seasons. The first result looked excellent: he won 68 per cent of points late in games, well above his own baseline. I nearly wrote a piece about an instinct for pressure. Then I recalculated the confidence interval. With 34 observations, the interval around 68 per cent stretched from roughly 50 per cent to roughly 82 per cent. In other words, my data was fully compatible with the possibility that this player had no special late-game ability at all.

That is the first lesson, and the one that repeats most often in this job: the 21-point system produces tremendously compelling matches and extremely small datasets for any question narrow enough to be interesting.

A shuttle clipping the net cord is not destiny — it is only a very small deviation between expectation and probability. But in badminton, the number of net cords a player experiences across an entire career is sometimes small enough to count on two hands.

The 565 km/h Smash and the Problem of Measurement Standards

In 2026, Malaysia's Tan Boon Heong was credited with a smash reaching 493 km/h, a figure once treated as the physical limit of the sport. A decade later, in 2026, international reports said India's Satwiksairaj Rankireddy had reached 565 km/h.

I spent many hours trying to trace the origin of both numbers. I found more than I failed to find.

What I found: both depend on the type of device used, where it is positioned, and when the sample is taken. Some systems measure racket-head speed at contact, others measure shuttle speed leaving the string bed in a short window after impact, and others again apply aerodynamic corrections. These three methods return three different numbers for the same shot, differing by as much as tens of per cent.

What I did not find: any public document describing a smash-speed measurement standard applied uniformly across the BWF World Tour.

The result is that the most widely shared number in badminton — smash speed — is the number least comparable across tournaments. A 400 km/h smash recorded in Copenhagen and a 400 km/h smash recorded in Jakarta may not be measuring the same physical quantity at all.

The more precise the number, the wider the distance between the person and the match. Three decimal places on a smash tell me nothing about whether it came in the third minute of game one, when the legs were fresh and the opponent had not yet read the direction, or in the fortieth minute of game three, when it was the last option of a player who had run out of choices.

I suspect this is one reason predictive models in badminton struggle to reach the accuracy seen in sports with standardised data. Not because badminton is inherently harder to predict, but because the input variables are inconsistent across sources.

The 1.15-Metre Stick

If the smash is measured by technology, then one of the most consequential decisions in a match is measured by the human eye and a stick.

Since 2026, BWF has applied a service rule requiring the point of contact to be no higher than 1.15 metres above the court surface. The service judge stands beside the net, compares the contact against a fixed marker attached to the net post, and delivers a verdict in under a second.

While electronic line-call systems are now present at many major events, helping resolve shuttles landing near the lines, the service-height decision has no equivalent technological support. It remains one of the largest grey zones in modern badminton law.

The flaw is not in the source code. It is in the eyes of the person reading the source code.

I have no public data on the error rate of service decisions at World Tour level. Nobody publishes that figure. And the very absence of that data is itself information: in a sport where every point is logged to the second, a decision capable of swinging an entire game is not audited.

At the same time, BWF also had to deal with the controversy over the spin serve, a delivery in which the server imparts spin on the shuttle with the fingers before striking it. From May 2026, the federation banned it on the grounds that it did not fit the definition of a service action. The fact that a technique existed for months, was used successfully by some players, and was then removed by administrative text rather than experimental data shows a sport still defining itself through the instincts of its rulemakers.

This does not make me doubt the legitimacy of those decisions. It makes me notice the order of priorities. Badminton buys expensive technology to determine whether a shuttle landed two centimetres out, yet keeps a manual method for a decision that is physically more complex.

Medical Data and the Crowded Calendar

The most closed layer is the one that determines an athlete's career length.

The case of An Se-young is the clearest example of the most recent Olympic cycle. The Korean player injured her knee at the 2026 Asian Games, kept competing, won women's singles gold at the Paris 2026 Olympic Games by beating He Bingjiao in the final, and then publicly criticised her national federation's management and medical care. What the public knows about the state of her knee during that period is very little, and most of it came from the player's own account rather than from verifiable medical data.

Kento Momota is the reverse case in terms of information. The car crash on the way to Kuala Lumpur airport on 13 January 2026, following a 2026 season in which he won 11 titles, was documented in full because it was a news event. But the metrics explaining why a player who once dominated could not return to that level over the following four years do not exist in any public database.

I tried to reconstruct that period using what was accessible: matches played, games played, win rate, average match duration. The picture I obtained showed a marked decline in win rate in matches that went to three games. But that picture could not distinguish between three different hypotheses: incomplete recovery from injury, loss of confidence after the event, or changes in how opponents prepared for him. All three produce the same statistical fingerprint. This is the inherent limit of retrospective analysis: I can describe the shape of the decline, but I cannot identify the cause.

Alongside this sits calendar pressure. The BWF World Tour runs roughly thirty events a season, from Super 100 to Super 1000, plus team events and the World Championships. Federation regulations require the top-ranked group of singles players to compete in all Super 1000 and Super 750 events, with financial sanctions for withdrawals that lack an approved medical reason.

Seen through a data lens, this is a system that manufactures a conflict of interest over information. Players have an incentive to report injuries as milder than they are in order to avoid sanctions. Organisers have an incentive to announce a full entry list. Broadcasters have an incentive to tell stories about players giving everything for the audience. None of them has an incentive to publish real data on the physical load carried by athletes.

Every price in the transfer market is the whisper of a fear. In badminton, the greatest fear of a twenty-five-year-old player is not losing a match, but losing a season to an injury nobody recorded early enough.

The Vietnamese Case

Vietnam has one of the most interesting data stories in the region, and it is among the least analysed.

Nguyen Tien Minh is the definitive example of a long career with extremely thin public data. Born in 2026, he competed at four Olympic Games and won bronze at the 2026 World Championships, Vietnam's first badminton world medal. Yet anyone wanting to reconstruct the answer to "what keeps a Vietnamese player at the elite level across two decades" will find very little: match lists, opponents, scores. No data on training load, on hours on court each week during his peak, on recovery routines, on the number of injury treatments.

Nguyen Thuy Linh and Le Duc Phat represent the next generation. Both qualified for the Paris 2026 Olympic Games and exited in the group stage. They compete mainly at Asian events and a handful of European ones, many of them Super 300 and Super 100 level, where the performance-metric layer barely exists.

Based on my experience tracking matches, a Vietnamese player at domestic and small Asian events generates far less public data than a European player of the same ranking. This does not affect match results, but it affects three other things.

It affects how opponents prepare. A Danish player can review dozens of matches by an opponent of similar level with full landing-point metrics. A Vietnamese player often has to work from raw video with no event-encoding layer.

It affects the market. Sports data providers serving the betting industry only invest in collection where there is liquidity. A Super 100 in Vietnam carries less betting value than a Super 750 in Europe, so it is not measured. The loop reinforces itself: fewer data means fewer analysts, fewer analysts means less betting, less betting means even fewer data.

And it affects decision-making at federation level, which is what concerns me most. A federation deciding how to allocate budget to a player needs evidence of improvement potential. If that evidence must come from international data collected for broadcasting rather than for athlete development, decisions will tilt towards players with denser international schedules rather than necessarily towards those with higher potential.

The Vietnam Open, a Super 100 event, illustrates the paradox. It is the largest international tournament staged in Vietnam, drawing players from many countries, yet most of its matches take place without an accompanying metric-encoding system. A young Vietnamese player who performs brilliantly there generates less quantitative evidence than a peer of the same age competing at a Super 300 in Europe.

The Counterintuitive Angle: Missing Data Is Not Random

There is an implicit assumption in most sports analysis: that missing data is randomly missing. That the observed sample still represents the population, only smaller. In badminton, that assumption fails at nearly every layer.

Performance-metric data is missing systematically at smaller events, on non-televised courts, in qualifying rounds, and in less-covered disciplines such as women's doubles and mixed doubles. The deeper into the system you go, the less data exists. That is also where most players actually compete and develop.

Medical data is missing systematically for injured players, for those returning after long absences, for those nearing retirement. This is statistically serious: the group we know least about is precisely the group where information has the highest value.

Withdrawal data is missing systematically because withdrawals are announced in a single short line with a generic reason. Those withdrawals carry information about physical condition that no model can infer from scores.

This is why I distrust badminton prediction models built only from public data. Not because the models are weak, but because they are fed a sample selected in ways nobody controls. The correlation between a metric and a match outcome can be strong in the available dataset and still be meaningless in the population, because both variables are filtered through the same mechanism: this was a televised match.

I made this mistake publicly once. At Euro 2026, I predicted Belgium would win the title because they held the tournament's highest attacking output. Italy beat Belgium in the quarter-finals and went on to lift the trophy. I spent sixty hours rewatching all seven of Italy's matches and realised I had asked the wrong question: I measured chance creation, while the title was decided by the ability to restrict an opponent's chances. Since then I weight defensive and structural metrics more heavily, even though they are far harder to collect in badminton.

Error in badminton analysis mostly does not live inside the data. It lives in the fact that the data was never generated to answer the question I am asking.

Signals for the Next Cycle

I am watching three signals in the period ahead.

The first is the pace of measurement standardisation. If BWF publishes a technical standard for smash-speed measurement and applies it uniformly at Super 1000 and Super 750 events, the entire range of power data becomes comparable for the first time. If it does not, record numbers will continue to be reported as sporting events when they are merely events of measurement equipment.

The second is open data at player level. If a national federation begins publishing workload data for athletes under its management, even in aggregate form, that would be the biggest break in badminton analysis in decades. I put the probability low over the next two years, but it is not zero.

The third, and the one closest to me, is Southeast Asian tournaments building their own data layer. A Super 100 in Vietnam may never carry enough broadcast value for an international company to install a camera system, but a manual event-encoding operation run by trained staff and funded by a national federation is entirely within budget. That would not produce smash speeds, but it would produce unforced errors by court zone, landing distribution by situation, and conversion rates when an opponent is defending. That is enough to change how a young player is evaluated.

When the crowd counts goals, I count the chances dropped on the way to the goal. In badminton, I count the rallies nobody ever entered into a machine, so that years later nobody can remember exactly what happened.

There is one thing I always remind myself before every analysis: if my conclusion cannot be verified by another person holding the same raw data, it is not analysis, it is an opinion decorated with tables. In badminton, that verifiability condition remains far harder to meet than in football. But the standard should not be lowered merely because data is scarce. The scarcer the data, the more obligatory it becomes to state clearly what is missing.

A shuttle landing near the line that an electronic system resolves, a smash measured tens of kilometres per hour off true, a game whose final numbers were never recorded: these are all slices of the same problem. This sport measures very carefully what is easy to measure, and leaves alone what matters.

The question I carry into the next season is not who will win. It is this: of all the players I have assessed, how many did I assess on the basis of what I actually saw, and how many on the basis of what I was handed?