The Empty Data Sheet on a Badminton Analyst's Desk

5:12 AM, Beijing. Outside the window it is still dark; only the streetlamp lays a strip of light across the wooden desk. In front of me is a match-tracking sheet for a BWF World Tour Super 750 event — the kind of sheet I have used for more than fifteen years. Every cell has its own name: rally-length distribution, net winners, service-return efficiency, error rate at decisive points. All of it is empty. Not a single number. Not a single note. Only the letters "N/A" blinking in the top right corner of the screen, like a tired blink. I sat looking at it for nearly forty minutes. What stopped me was something else: my first reflex — a reflex honed over forty-three years — was to write something anyway. An observation. A prediction. Some sentence that would read as thoroughly professional. I did not write. I sat still. And I began to understand that the empty sheet was teaching me something no metric could teach. To a deep badminton analyst, a match-tracking sheet is not a piece of paper recording a score. It is a multi-layered structure. The top layer is what everyone sees: the score of each game, the number of service points won, the number of service faults. The middle layer is what you have to rewind video to read: average rally length, the share of short rallies under five shots and long rallies over fifteen shots, net-win rate, the depth of the service return, the frequency of coming to the net after the first smash. The bottom layer — the layer I care about most, and the one few people bother with — holds the signals no scoreboard displays: the rest interval between rallies, the drift in shuttle trajectory when a player tires, and the clustering of net errors across the final ten points of a third game. The Badminton World Federation's tournament system has been tiered since 2026: Super 1000, Super 750, Super 500 and Super 300. The difference between tiers lies not in prize money — though that figure is large enough to change many players' entry decisions — but in opponent density within a single week and in accumulated physical load. A player reaching a Super 1000 semifinal has usually played four matches in five days, at least one of them over three games. At a Super 300, the same number of matches comes with a noticeably lower intensity of opposition. Placing two sets of metrics side by side without separating tiers is the fastest way to manufacture a wrong conclusion. And that is exactly where I recognised the value of the empty sheet. Over fifteen years I have fallen many times into a familiar trap: find a number, then write around it. I was once infatuated with smash speed — measured in km/h, a beautiful figure, easy to quote, easy to impress with. A 420 km/h smash always reads better than a metric about pace distribution. But after rewinding hundreds of matches, I had to admit: smash speed correlates very weakly with winning. The fastest smashers are not the most decorated champions. Champions are the ones who control the rhythm of the rally — who know when to extend it, when to end it, and when to let the opponent end it themselves. The numbers are not wrong; I simply forgot to ask where they were standing. Smash speed sits on the top layer of the sheet. It is the product of a chain of decisions: court position, timing of the jump, racket angle, and — most importantly — the quality of the opponent's previous shot. Pull it out of that chain and it becomes a decorative figure. I remember a semifinal at a Super 1000 some years ago. One player's average smash speed was nearly 30 km/h higher than his opponent's across the match, yet he lost 1-2. The speed statistics said he was stronger. The video said the opposite: most of his smashes came after he had already lost the rally shape, in deep retreat and with his body rotating away. Those smashes were beautiful, fast, and meaningless. A number detached from context is only a lie that has been beautified. So when the data sheet is empty, what I must do is return to the bottom layer and ask myself: what do I actually need to see in order to say something useful? I need the rally-length distribution. Not the average — the average is the enemy of what is really happening in a sport whose pace shifts from game to game. In the first game, rallies tend to be shorter because both players are fresh and opt for early attack. In the second game, once one side has lost the opener, longer rallies appear more often. And in the third game, that distribution shifts once more — but in different directions depending on how much energy the first-game winner spent. A player who wins the first game in 25 minutes and a player who wins it after saving three game points enter the third game in completely different states, despite the same 1-0 scoreline. I often use a few examples to explain this to younger colleagues. In women's singles, An Se-young of South Korea is the kind of player the top-layer statistics cannot fully explain: her smash speed is not in the highest bracket, yet her win rate in rallies over fifteen shots is among the highest in the world. In men's singles, Kunlavut Vitidsarn of Thailand shows the same thing in a different way — sheer rally tolerance. And Viktor Axelsen of Denmark is a fascinating case: he has speed, but what sets him apart is not speed, it is his control of the mid-court region, forcing opponents into a shot they do not want to play. I also need the rest interval between rallies. This is the most underrated metric in professional badminton. There are 60 seconds between rallies; 120 seconds between games. How a player uses that time — the number of steps, the number of times he wipes sweat, how he places the towel, his breathing rhythm — tells a story the scoreboard does not. I have spent hours rewinding footage of these intervals. A player who begins shortening his steps midway through the second game will usually lose the third. Not because of technique, but because the energy system raised the alarm before the scoreline did. When the arena falls silent, that is when I hear the whisper of the underlying data clearly. I learned this in 2026, when tournaments were suspended and I sat at home rewinding hundreds of matches played in empty arenas. What I found went beyond the question of noise. When the stands are empty, players change how they handle the shuttle at key moments. Net errors rise. Short serves that miss their target rise. But at the same time, the server's win rate also rises slightly at decisive points. With no cheering, there is no social pressure from the crowd — but there is also no release. Pressure moves from outside to inside. And it shows up as very small numbers, sitting on the bottom layer of the tracking sheet. If I looked only at the top layer, I would conclude that matches without crowds proceed normally. The bottom layer says otherwise. For badminton, I use a tool close to the PPDA metric in football — something I once used to read the pressing rhythm of the Italy national team at Euro 2026. In football, PPDA measures the number of passes an opponent is allowed before facing pressure. In badminton, the equivalent is the number of exchanges a player allows the opponent to control before acting himself. But like PPDA, it is only a stethoscope. PPDA is only a stethoscope, but the one listening to the patient must be a monk who knows how to stay silent. Meaning the metric only brings me to the right room. Sitting in it, listening to every pause, and accepting that not every sound carries meaning — that is the part of the work that cannot be outsourced. Here I have to say something the sports-analytics industry rarely says outright. Most deep data in professional badminton today is collected systematically, and its largest commercial value does not lie in articles or television analysis segments. It lies with betting companies. Every detailed metric, every probability model, every bottom-layer tracking sheet can be packaged, resold, and fed into a pricing algorithm that never sees the light of a court. Fans think data helps them understand the match better. It does. But it also helps another party price that understanding. I do not reject data. I only refuse to let data become a curtain. There is another layer to this story I rarely write about. Streaming platforms are paying enormous sums for the rights to major badminton events while subscription revenue fails to keep pace. I saw this happen to television twenty years ago, and I see no reason it ends differently this time. When rights fees exceed the value viewers actually generate, the first thing cut is always deep analysis — because it takes time and generates no clicks. The mistake is not believing the model; it is failing to ask what the model left out. The empty sheet that morning asked me exactly that. What did it leave out? It left out everything — which means it also promised nothing. An empty sheet is the most honest sheet in my entire drawer. In this industry, the person who writes "I don't know" is usually seen as weak. But across forty-three years of watching sport, I have found the opposite: the person willing to leave a cell blank is the person who understands that cell. The person who fills every cell with a number — any number — is usually just protecting his own reputation. I once thought data was truth, until the 2026 World Cup taught me fear. That year, at fifty, I analysed the entire group stage using xG and concluded Croatia would lose to France in the final, simply because the numbers did not favour them. I ignored something xG cannot measure: how a team changes the tempo of a match when it falls behind, and how penalty shootouts collapse every model. After that tournament, I spent a full month rewinding twenty Croatia matches, noting every transition phase, and building my own coefficient for a team's volatility. It did not save me from every mistake. But it taught me that every metric has a boundary, and the analyst's job is to stand on the right side of that boundary when speaking. For badminton, where is that boundary? It lies in the fact that this sport, through its 21-point rally-scoring system, has turned every rally into a unit of absolute value. There is no gradual clawing back as in time-based sports. Every point carries equal value on the sheet, but by no means equal psychological value. The nineteenth point at 18-18 weighs many times more than the nineteenth point at 18-11. Any model that merely counts points without weighting them by score-state is ignoring half the match. That is why I began taking notes by cluster. Not point by point, but in clusters of three to five points. Within each cluster I record: who changed the tempo, who reacted, and whether the cluster ended in a short or a long rally. After a few hundred matches, a pattern emerged more clearly than any percentage: players who win three games in a row are usually those who win the third cluster of points in each game, not those who win the first point. This sounds small. But it changed how I watch badminton entirely. I no longer wait for the decisive smash. I wait for the moment a player decides whether the next rally will be long or short — and that moment usually does not appear on the scoreboard. So when someone asks me how an upcoming Super 750 will unfold, I usually answer with another question: which data do you want me to read? If you want a number, I can give one. If you want an opinion, I can give one. But if you want to understand the match, I need time, and I need the right to say that I do not yet know. The empty sheet that morning taught me more than any sheet full of numbers. It reminded me that my job is not to fill in the cells. My job is to know which cells must be filled, which must be left blank, and which — even with a number already in them — still need to be checked again on another morning, when the court is empty and I am calm enough to listen. Three signals I will write onto the sheet in the next round, this time not left blank. The clustering of net errors across the final ten points of a third game, where pressure shifts from technique to psychology. The rest interval after every rally longer than fifteen shots, where energy is redistributed. And the depth of the service return at scores from 17 upward, where the best players separate themselves from the rest. If one of those three signals disappears from the sheet next round, I will not rush to conclude that the player has changed. I will first ask what I have overlooked.