Badminton Measures Smash Speed, but Ignores What Actually Decides the Point

Istora Senayan, second game, 18-18. The shuttle drops just inside the sideline; the line judge calls it out. The player stands still, no reaction, eyes fixed on the big screen. It takes roughly seven seconds for the system to rebuild the trajectory. The graphic appears thin and curved, the contact point a few millimetres from the line, and the arena splits into two different sounds: half the crowd roars, half hisses.

I am in row eleven, pen still in hand, my notebook open at a page two-thirds full. There is no smash speed in it. No score. Only symbols I invented myself: a slanted arrow for a long return of serve, a dot for a short serve, a slash for a rally that ended with an error in the front court. For four years I have logged matches this way.

On screen, once the point is confirmed, the stats panel shows the familiar line: fastest smash of the rally, in km/h. The crowd applauds. The broadcast replays it three times. While the whole arena looks at that figure, I am thinking about the third serve of the game — the one nobody replays, nobody measures, and which, according to my notebook, decides more than any smash.

The map of data in a sport that measures the wrong thing

Over roughly fifteen years, professional badminton has built a fairly dense data system. The Badminton World Federation (BWF) operates an electronic review system called the Instant Review System, built on the Hawk-Eye platform, allowing players to request a review in a limited set of situations. World Tour events are equipped with high-speed cameras from multiple angles, trajectory reconstruction software, and a parallel data feed that drives the broadcast graphics.

That feed produces three main groups of information. The first is speed: smash speed, serve speed, sometimes drive speed. The second is rally outcome: who won the point, with which stroke, from which position. The third is a set of composite metrics such as average rally length, net approaches and unforced errors.

The third group is the one rarely shown on air. The first is shown on almost every rally worth showing.

Based on my experience tracking World Tour matches across several seasons, this misalignment is not harmless. It shapes how audiences understand the sport, how broadcasters build narratives, and at a deeper level, how the betting market prices each player's chances. When a metric is repeated often enough on screen, it stops being information and starts being a bias.

In Indonesia, where badminton is not just a sport but part of national identity, the misalignment is sharper. Fans follow Anthony Sinisuka Ginting and Jonatan Christie with a different intensity. Every one of their smashes is replayed, counted, compared. But the few who study technique know that matches are usually decided in rallies where the camera never stops.

I walk into the church of data not to pray, but to listen to the noise of the truth.

Where smash speed is measured, and how that changes how we read a match

A smash is measured at the moment of contact between racket face and shuttle. That is the highest point in the shuttle's entire flight. After that instant, the shuttle can only slow down.

The distance from the hitting position to where the opponent stands is usually eleven to thirteen metres. Over that distance, air resistance pulls the speed down significantly, and the decay is not uniform. A flat, hard shuttle passing through the warm air above the court holds its speed better; a shuttle that arcs upward loses speed far faster and gives the opponent more reaction time.

So when the board flashes a very high smash speed, what the audience learns is: player A swung very hard. What the audience does not learn is: at what speed the shuttle reached the opponent, at what height, and from what body position.

In my notebook I keep a separate column I call "arrival speed" — a relative estimate in three bands: fast, medium, slow, based on the frame count from racket contact to opponent contact when I rewatch the video. The method is crude and subjective, I know. It cannot replace radar data. But after roughly two hundred matches coded this way, I noticed something broadcast graphics never mention: speed measured at contact correlates far more weakly with speed at the opponent's racket than people assume.

Some players appear constantly in a tournament's fastest-smash table, yet their direct smash-winner rate does not follow. Conversely, some players with average smash speed achieve a higher scoring efficiency because they choose placement, rhythm, and the moment when the opponent has just moved the wrong way.

A hard smash is a statement about physical capacity. A well-placed smash is a tactical decision. The scoreboard measures the first and presents it as if it were the second.

An evidence chain from rallies that never make the broadcast

From 2026, when the international calendar shut down, I started building a personal database by hand-coding every rally in old matches. It took months, mostly to fill the gap when there were no new matches to dissect. Today it holds more than three thousand rallies from various events, mostly men's singles and men's doubles.

The three findings below come from that dataset. I should say up front that the sample is small, the coding was done by one person and therefore carries systematic error, and the confidence intervals are wide enough that further data could reverse several conclusions. What follows is a hint, not a law.

First, most points are decided within the first four shots.

When I group rallies by shot count, the band ending in one to four shots holds the largest share. Rallies longer than twelve shots are a small minority, but they cluster at the most important scorelines. Long rallies are rare, and they arrive exactly when the match is tightest.

The implication is direct. If most points come from the first four shots, then serving and receiving carry far more weight than their place in news coverage suggests. A short serve at the wrong height gets driven flat into the open corner, and the rally ends before the crowd has settled.

Second, front-court errors are the most expensive and the least counted.

In my dataset, errors arising between the net and the short service line — net cords, shuttles hit long from an attacking position, net shots that drift out — account for a substantial share of points lost. These rarely appear in post-match statistics. They are folded into a single label, "unforced error", which says nothing about cause.

A net cord from an attacking position can come from legs that are gone, from eyes that missed a small movement by the opponent, or from a player who decided before the shuttle left the opponent's racket. Three different causes, three different fixes, counted as one.

Third, return-of-serve quality separates winners from losers more clearly than smash speed.

Comparing wins and losses for the same group of players, the largest gap is not in smash winners but in the rate of returns of serve that put the opponent on the defensive. Winning matches show a clearly higher rate. Direct smash winners differ far less, and in some pairings are almost identical.

The most reasonable reading is that a good return creates the conditions for a smash winner, rather than the smash creating its own conditions. This is a reversed causal relationship when you look at the end-of-match statistics table. The reader sees many smash winners and concludes that smashing wins matches. In reality the causal chain starts earlier, at a shot with no row in the table.

The grey zone of the review system

The Instant Review System launched with a simple promise: reduce controversy by supplying visual evidence. After several seasons of operation, I would argue that promise is only half kept.

The system is very good at determining whether a shuttle landed in or out when the contact point is a few millimetres from the line. That is the kind of dispute the human eye cannot resolve, and machines can. But it also creates a new kind of dispute: an argument about the system's own margin of error.

In many sports using trajectory reconstruction, there is a tolerance zone in which the displayed result still upholds the on-court umpire's decision. That zone exists because no system is perfectly accurate. But to viewers, a line drawn on screen looks perfectly accurate, and when a decision is upheld despite a line that appears to have crossed, a sense of injustice appears immediately.

The problem is that viewers never see the tolerance zone. They see a line. They do not see the uncertainty on either side of it.

While the screen rebuilds the trajectory, players usually stand still. Nobody measures that interval, no metric records it. Yet it is the longest pause in a badminton match, longer than the interval between games. In those seven to ten seconds, a player can reload, break the opponent's rhythm, or simply escape a run of points flowing the wrong way.

When every tournament stops, I finally hear my own heartbeat.

From another angle, using a review has itself become a tactical decision. A player can challenge at a scoreline going badly, not necessarily because they believe the umpire is wrong, but because they need a legitimate pause. Reviews are limited per match, so spending one is an opportunity-cost calculation. That is a dimension of the match with no column in any table.

Live data and the market behind the line

Here I have to be blunt about my own trade, because it governs how I see everything.

Sportsbooks do not measure smash speed themselves. They buy data from specialist collection providers or contract directly with tournament organisers. The feed sold to markets is usually more granular than the feed given to broadcasters: updated within seconds, attached to individual rallies, coded with end position and final stroke type.

The result is a paradox. A viewer watching television receives less information than a bettor watching a live odds board. Both are watching the same match, but one sees it through the lens of storytelling and the other through a grid of numbers.

This is the darkest side effect of the digitisation of sport, and it is rarely named correctly. Data created to serve understanding of the match becomes an asymmetric advantage for a very small group of users who do not need to understand the match, only to know where the price is wrong.

At a broader level, this shapes how the sport is packaged. Metrics that are easy to sell to markets get prioritised. Smash speed is a beautiful metric to package: an absolute number, easy to compare, easy to place side by side. Return-of-serve quality is complex, requires manual coding, and sells poorly. Markets reward what is easy to measure, and over a few years, the easy thing crowds out the hard thing.

The more precise the number, the wider the distance between the people and the match.

Undervalued skills

If smash speed is the most inflated metric in modern badminton, the most undervalued skills are three: defence from an off-balance position, deception through hand movement, and the quality of recovery movement after being pushed to the back court.

The first rarely makes highlights because it does not produce a beautiful image. A retrieval from a near-fall does not draw the roar a flat smash does. But in my data, the rate at which players win the next point after a difficult retrieval is notably higher than the average rate.

The second concerns a very small technical detail: racket-face angle in the final instant before contact. Good deceivers hold the same arm shape for several different strokes, changing only the wrist angle within a very short window. At full speed from a wide angle, the difference is nearly invisible. In my notebook I mark it with an asterisk, and I found that players with many asterisks tend to have strong third-shot winning rates.

The third concerns fitness, but not strength. It concerns the speed of recovering the central position after each stroke. A player who smashes hard but lands slowly is always passive on the next shot, no matter how good the previous smash was. In many matches I coded, the gap between winner and loser was not in the decisive stroke but in where they stood when the next rally began.

The counterintuitive angle: reading correlation backwards

There is one analytical mistake I have made many times and still keep making, despite reminding myself constantly.

When I calculate the correlation between smash speed and match win rate in my dataset, the coefficient is positive and fairly clear. The naive reading is: hit harder, win more. The correct reading is: better players tend to hit harder, and they win because they are better, not because of speed.

Separating those two possibilities requires comparing within a group of players of similar standard, and there the gap in smash speed between winners and losers narrows considerably. Most of the effect disappears once skill is controlled for. The same pattern repeats across many sports metrics: it is attractive, it is easy to measure, and it correlates with success — but most of that correlation comes from a hidden variable behind it.

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

There is a more dangerous error still: reading a match backwards once you know the result. Knowing that player A won, every decision player A made looks sound and every decision player B made looks wrong. Data does not produce this reading by itself. The analyst produces it, and then credits the data.

The only defence I have found is to write pre-match expectations down, in words, dated, and never edit them afterwards. That forces me to reread lines I wrote and see I was wrong. It is the price of keeping data honest.

A shuttle clipping the net cord and tumbling over is not fate — it is only a very small deviation between expectation and probability.

I once sat down to calculate the probability of a net cord flipping in favour of the striker. The rate was not nearly as small as the crowd's feeling suggests. The real figure makes the concept of luck much paler. The same applies to losing streaks the press calls a form crisis. In most cases, it is an ordinary run of results read emotionally.

When the crowd counts the smashes, I count the returns of serve that were thrown away.

What to watch in the next cycle

Four signals I will be tracking over the coming months, listed so readers can verify for themselves rather than trust my conclusions.

First, whether tournaments begin publishing defensive and recovery-position data. If that happens, the quality of tactical discussion changes within one season, because viewers finally have a tool to evaluate rallies they have always watched but never had measured.

Second, how the review system handles the communications problem of the tolerance zone. If organisers start displaying a confidence band rather than a single line, controversy will fall in some situations and rise in others. It is worth watching because it reveals what the sport understands about audience psychology.

Third, the evolution of the short serve in doubles. If the short-serve rate keeps climbing while scoring efficiency does not follow, that is a sign of copying the form without understanding the mechanism. I have logged a similar pattern in football, and it usually lasts several seasons before being corrected.

Fourth, how much of the commercial data feed is allowed into public view. This is the most important and hardest signal to track, because it sits inside contracts the public cannot read. But if the gap between what is on air and what is on the odds board keeps widening, the sport will eventually have to answer a question it has postponed for a long time.

I will keep logging my notebook by hand, because it is the only way I know to stop data from speaking on my behalf. Those pages will never appear on a big screen, and probably nobody but me will read them. Still, in every match, when the screen shows the smash-speed line, I will be looking somewhere else.

---

Note: All figures cited are the author's hand-coded data, with a limited sample, systematic error, and should not be read as statistical conclusions. This content is for sports information only and does not constitute betting advice. Competitive sports outcomes carry high uncertainty.