Five Empty Volleyball Metric Columns and the Analyst's Discipline of Not Concluding

At 2:14 in the morning I reopened the folder labelled "VNL — pool stage" on my hard drive. Five columns, five metrics I still use to read a volleyball team: spike success rate, blocks per set, ace-to-error ratio, perfect-pass rate, dig rate. Every cell empty. Those cells are empty because there is nothing underneath to fill them, not because I was lazy. One month, 64 matches, and every dead ball written down in my notebook — and here I do not have a single line to start from. For twenty years my trade has rested on one assumption: the data always exists, you only need enough patience to find it. That night the assumption moved.

This five-metric panel works like a chain. The quality of the first pass decides the quality of the attack, the quality of the attack decides how the opposing block sets up, and how the block sets up decides the rhythm of the counter-attack. Every link needs its own denominator: number of passes, number of attack attempts, number of opponent attacks, number of sets, number of service errors. Without a denominator the first link hangs in the air, and the other three collapse into guesswork dressed in terminology.

I watched 17 matches only to find the gap Elsinho left behind him. What I brought home was not a conclusion but a set of repetitions thick enough for a conclusion to stand on. In volleyball that standard is harsher, because every rally lasts a few seconds and is cut apart by the whistle. In the Japanese market, the perfect-pass rate is graded on how precisely the ball reaches the setter's hands. In Vietnam, the measure leans towards whether the ball reaches the setter at all. Two frames of reference, two numbers, one player. Putting them into the same ranking without stating the convention is a category error, not a small one.

Take each column in turn. Spike success rate requires knowing how many times a player attacked, from which position, and against which block. Without those three, the ratio only describes the recorder: whoever records more has a bigger sample, whoever records less has a smaller one.

Blocks per set hides a trap inside its own denominator. A three-set match gives each set more weight than a five-set match does. Fair comparison demands normalising by the opponent's number of attack attempts, not by sets.

The ace-to-error ratio is the most fragile figure in the group. One ace and one error gives 1.0. Three aces and three errors also give 1.0. Those two players are not the same, and neither number says who served better. The perfect-pass rate depends entirely on the grader's convention. And the dig rate depends on a question few stat sheets bother to answer: did that rally end with a set that could still be attacked, or merely with a high ball sent back over the net?

When the denominator is zero, every metric is right and every warning is meaningless; that is why an empty table is more dangerous than a wrong one. A wrong table can still be corrected by cross-checking. An empty table can only be filled with guesswork, and guesswork never admits that it is guesswork.

The risk checklist attached to the metric set falls into the same trap: insufficient data to support a tactical claim, single-point dependency on one player, a first-pass system whose swings collapse the structure, a new lineup still gelling, being countered by a specific opponent type. With no data, all five are simultaneously plausible. A checklist where every box can be ticked is equivalent to a checklist with no boxes at all.

Nine corners repeated, until a 2-0 scoreline became a warning. The principle carries across sports: a warning has value only when it comes from a sequence that has been counted, not from a possibility that has not been eliminated.

In the summer of 2026, with stadiums closed, I pulled data from 120 matches across the J.League and the Bundesliga and found high pressing roughly 30 percent more effective, simply because players could hear each other call. I doubted the 30 percent figure, so I watched 15 Bayern matches before believing it. Six weeks of verification showed the effect came from empty stands, not from the midfield. In volleyball a similar effect could come from the absence of crowd noise masking the setter's signals.

The counter-intuitive point sits here: an empty dataset is not the analyst's failure, it is a finding about the collection system. When the spike success rate is missing, the right question is not "did this player attack well" but "who failed to record it, and since when". Sport has a habit: every match must produce a conclusion. Content calendars, sponsor demands and the pressure to say something push writers towards the fastest available conclusion.

Before publishing any model, I try to break it first. With every predictive model I ask myself: if this evidence is right, where is my system wrong. That question is not a ritual. It is the only way to tell a finding apart from a preference presented in numbers.

There is another temptation, subtler. The story of a small club using data to beat a big one sounds wonderful, and it conceals two things: the financial gap has not disappeared, and the small club's data quality is usually worse than the big club's. When neither side has data, that story becomes even easier to tell, because nothing can refute it.

Next match I will still carry that notebook into the stand, but the first page will have a column that never existed before: a column for what nobody managed to record. Spike success rate unmeasurable for this reason. Blocks per set left blank for that one. That column may end up fuller than the other four.

If a team wins a match and nobody manages to record a single metric, is what we are doing analysis, or just retelling an evening?