Nine Dimensions of Volleyball Analysis and a File Holding One Surviving Word: Verification Discipline When the Input Is Empty

I opened the input file at nine in the morning, before anyone else reached the data room. Nine dimensions of volleyball analysis were laid out as nine blocks, each block holding a few tables, each table holding a few rows. The same sentence sat in every cell: insufficient information for assessment. Article title field, empty. Source field, empty. Publication date field, empty. Competition field, empty. Team field, empty. Coach field, empty. Statistics field, empty.

The file still carried an introduction, a conclusion, a risk assessment, a five-star information-value scorecard, and a disclaimer at the bottom. Correct structure. Correct format. The only missing element was content.

The one living signal in the entire file was a single-word domain label: volleyball. One word holding up a nine-storey structure.

I read that file three times, the way I still rewatch match footage — once at normal speed, once in slow motion, once looking only at the places I would rather not look at. All three passes produced the same result. The pitch does not ask the reader's gender, it only asks how deep you read. In this file there was nothing to read deeply.

To understand what happened, the workflow needs explaining. A deep volleyball analysis in a modern data room runs through two layers. Layer one deconstructs the source article: it pulls out atomic facts — each fact a sentence that cannot be split further while remaining true, such as “Team A won 3-2 on 12 July” or “Setter B entered in the third set” — and then identifies entities: team names, people, coaches, competitions. Layer two takes that output and analyses it across nine dimensions: tactics, statistics, competition system, landscape, rules and governance, roster building, risk, public narrative, and industry transmission. Without layer one, layer two is only a skeleton.

Layer one returned an empty list this time. Not one fact. Not one entity. An output file that states in its own first line that it is empty.

Two ways to handle that situation. The first: fabricate something to fill the frame. The second: state clearly that it is empty and stop. This file chose the second, and added a hypothesis about cause: the source-article fetch failed — the piece may sit behind a paywall, the page may render only through JavaScript so the reader tool sees no text, the link may be dead. The conclusion sits at the technical layer and it is verifiable: retrieve raw text longer than three hundred characters with real content, re-run layer one, done.

In other words, this is a pipeline failure. A pipeline is the chain of steps running from data retrieval to final conclusion; when one link breaks, the whole chain stops, but the broken link sits at the intake end, not at the reasoning end.

This connects to volleyball in a more specific sense than a forgotten domain label. Volleyball is a sport whose public data layer is far thinner than football or basketball. At national-team level and at major tournaments, statistics systems run automatically ball by ball. Down at mid-tier national championships and youth events, most numbers are entered by hand by one or two people sitting in the stands with a notebook. Every number has a person behind it, and every person behind it carries a personal error margin.

That is why, when I tell volleyball stories with numbers, I always translate the terminology into ordinary language before using it. Perfect-pass rate is the share of first passes delivered to the position that leaves the setter's full tactical menu available; put simply, it is the share of first contacts good enough to keep the team's options open. Rotation is one of six service-order configurations deciding who stands in the front row and who stands in the back; in volleyball, a rotation with only two front-row attackers is a structural weak point, and any opponent who reads it will funnel the ball there. An out-of-system attack is an attack played after a broken first pass, when the team has lost the right to run coordinated plays and must live on individual ability. Olympic-cycle positioning classifies a moment into one of four types: Olympic year, qualifying year, adjustment year, generational-transition year — because the same win means different things depending on which type you are in.

Those four concepts are four inputs. Remove one of the four and every conclusion downstream is an ornamented guess.

Most people misread where a data system breaks. They assume breaking means producing wrong numbers. The more dangerous break is producing a correct format with no content, because correct format carries visual weight.

An empty cell on screen is instantly visible as empty, and people go looking for the cause. A cell filled with the sentence “insufficient information for assessment” looks like a conclusion somebody worked to reach. The two differ in substance, match in appearance, and most skimming readers remember only the appearance.

One detail in this file deserves a pause. The information-value scorecard has four rows on a five-star scale. Competitive value scored one star, with a note that it earned that single star only because the domain label survived. Industry value scored one star. Timeliness scored zero, because there is no date against which to assess timeliness. Reference value scored zero, because there is nothing to cite.

One word, “volleyball”, was enough to keep the whole scoring system from collapsing to absolute zero. That is a compact illustration of how scales behave: a single surviving fragment of signal is enough for the system to start awarding points rather than returning empty. Analysts need to know this in order to guard against themselves.

That mechanism has an exact twin in how we read match statistics, and this is the part I watch most closely. A surviving fragment of signal, pushed up into a conclusion with weight.

The most recent example from the season I follow: an attacker introduced with the line “seventy-eight percent perfect-pass rate”. Readers take it to mean this player receives well. The question to ask before believing it: seventy-eight percent of how many contacts? If the sample is twelve contacts, each correct or incorrect contact shifts the figure by more than eight percentage points, and the number says nothing about long-run ability — it says the team met a server who was not heavy enough in one match. If the sample is three hundred contacts across a season, each contact moves the figure three-tenths of a percentage point, and only then does the number begin to mean something.

The danger does not sit in small samples. It sits in small samples entering the story as an attribute of a person, rather than as a sample.

Based on my experience tracking matches, this class of error shows up most often in two places: reception metrics and rotation metrics.

The rotation question matters more. The standard analysis for a two-attacker rotation is to track the point-loss rate while the team occupies that rotation, set it against the team average, then split further by the opponent's serve quality — because the same rotation dies against a heavy-serving team and survives normally against a light-serving one. To perform that split, the scoresheet must record which rotation the team occupied at the moment each rally ended. Most amateur scoresheets record only the final point. You cannot ask a notebook about something it never wrote down.

So when someone tells me Team X is weak in rotation three, I ask two things back. First: does the scoresheet record rotation rally by rally. Second: was the recorder sitting close enough to see where the setter stood. If both answers are no, what is being called analysis is one person's memory of one match.

There is a further layer few people notice: material conditions decide which conclusions can be drawn at all. Where multi-angle footage exists, foot position and body rotation can be cross-verified. Where a single camera sits high in the stands, every conclusion about footwork is an estimate with a lower hit probability than we assume. Matches per season, paid recorders per match, matches with sufficient camera angles — those three variables decide how far a team can analyse, before coaching quality even enters the conversation.

I do not believe in diagrams, I believe in intent — the weak draw diagrams to reassure themselves. A fully filled table is a diagram. What needs finding is the intent behind it: who recorded it, when, and to answer which question.

Twice in my career I had to return to the source before I was allowed to write, and both times the conclusion changed.

The first was the round-of-sixteen match at the 2026 World Cup in Russia, where Japan led Belgium 2-0 and lost 2-3. The decisive goal came from a fourteen-second sequence starting immediately after a Japanese corner, launched by the Belgian goalkeeper's quick throw. The press called it a mental collapse. I rebuilt it frame by frame at slow speed and cross-checked against positional data. The fourteen seconds in Rostov do not live in the goal, they live in the silence between two touches — and that silence has to be measured: how many fractions of a second the ball took from the touchline to the halfway line, how many fractions of a second the Japanese defender took to turn his back, whether that span exceeded the time needed to reorganise the defensive line. Measurement first, conclusion second. Without measurement there are only adjectives.

The second was the pandemic season. I tracked thirty-seven matches across the first two rounds played in empty stadiums and found the home-win rate falling from 43.2 percent to 29.7 percent, while away teams' pressing intensity rose by more than eight percent on the passes allowed before the opponent crossed the halfway line. The conclusion was not “away teams got stronger because the crowd was gone”. The conclusion was: crowd noise is itself a variable in the decision to play out from the back, and when that variable disappears, the risk level the home team accepts shifts automatically.

Both times began with re-checking the source, and both times proved more useful than any commentary written overnight.

A team does not need eleven geniuses, it needs eleven people who own their roles. In volleyball the number is six and the principle holds: a team does not need six outstanding individuals, it needs six players in the right roles. Data systems work the same way. They do not need ten elegant metrics, they need each metric to own its role — and they need a gate that knows how to block when a role stands empty.

The natural reflex on seeing a system return empty is to buy more data and add more models. That reflex treats the symptom at the reasoning layer while the wound sits at the collection layer. Adding models to an empty input only produces empty predictions, better presented.

The deeper blind spot lies elsewhere: this profession rewards output volume and does not reward refusal. A data room reporting “we blocked forty pieces this month for lack of sourcing” gets asked why output fell. A data room reporting “we published forty pieces this month” gets praised. That incentive structure leads to a predictable outcome: tables full of words and hollow inside, and worse, they always agree with what the coaching staff already believed — because people tend to hunt for metrics that confirm rather than metrics that contradict.

A system that tracks everything and verifies nothing will produce exactly the kind of output readers cannot distinguish from real analysis. That is the largest risk, and it is quieter than any technical error.

The verification threshold for the next cycle, written down and measurable: before any analysis table is distributed, it must pass a gate — at least three atomic facts with source and date, at least one named entity. Fail the gate and it is flagged blocked and not distributed, with a specific reason.

For readers, one sentence is enough to carry: next time a metric stands beside a player's name, ask two things — how many contacts it was computed from, and who wrote it down. A system willing to say it does not know is still working. Our job is to keep that gate closed, even when someone pushes.