A Complete Volleyball Analysis Framework With Empty Data: The Limits of Analytical Honesty

On a Tuesday evening, I opened an analysis file sent from the data desk. Nine major sections, each with a table, a rating box, a risk line. Skimming it, it looked exactly like every report I had received after a round of the national volleyball league. I scrolled down to find the first number. There was none. I scrolled up to find the match name. None. Team names, player names, competition names — all blank. The only field filled in was a single short line naming the domain: volleyball.

I read it three times, out of habit. On the third pass I realised I was not reading an analysis. I was reading an empty frame, carefully formatted to look like one. That feeling is colder than reading a wrong report. A wrong report can be fixed. An empty frame presented as complete has nothing to fix — only the risk of being mistaken for fact.

My work in recent years has been tied to two-tier analysis systems. Tier one extracts text: it reads an article, a bulletin, a match record, and pulls out "information points" — event-shaped claims that can be verified. Tier two takes that list as raw material and runs it through nine dimensions: tactics and technique, data, competition system and schedule, team landscape and positioning, rules and governance, squad building and personnel, risk surface, public narrative and expectations, and the sport's industry transmission chain.

The relationship between the tiers is causal, not decorative. Tier two does not invent events. It only verifies, compares and contextualises. When tier one returns an empty list, tier two has nothing to verify. It can still produce form, but that form is not anchored to any reality. And a form not anchored to reality is, in my trade, the most dangerous thing there is: it clothes emptiness in a professional look.

The problem is not the idea behind the two-tier system. The idea is right: extract first, analyse second. The problem is that the extraction tier can fail in two very different ways, and if we cannot tell them apart, we will handle it wrongly. The first is under-extraction: the source article is rich but the system captured only part of it. The second is structural failure: the system returns fields pointing at content that does not exist. The first is fixed by re-reading the source. The second requires fixing the pipeline before re-running it.

I watched 17 matches just to find the gap a player left behind his back. If I did that for a small hypothesis, I must do it even more for a document claiming to be a conclusion. Before publishing any model, I try to break it first. Here, no breaking is needed: the frame collapses at its first anchor.

The first anchor is the "information points" field. In this entire architecture, it is the only evidentiary field. Every dimension behind it — from attack efficiency, blocks per set and ace-to-error ratio to perfect-pass rate — draws its data from there. That field is empty. Not short of a few items, but entirely blank. That does not mean "we know little"; it means "we know nothing". An honest analysis must say so, rather than filling the gap with speculation presented as conclusion.

There is a distinction I always stress to younger colleagues, because it is the single most common distortion in volleyball reporting: the difference between attack success rate and attack efficiency. Success rate is attack points divided by total attempts. Efficiency is points minus attack errors and times blocked, then divided by total attempts. The two are often merged into one in news copy, and when they are merged, a hitter with a 45% success rate and eight errors in a match looks far better than he actually was. Only raw data can settle that difference. The analysis in my hands has not a single number to settle it, so it can say nothing about anyone.

The sample matters too. A metric means something only when you know how many touches it was computed over, against which opponents, and across what period. The same blocks-per-set figure can be impressive over three matches against weak sides and ordinary across a full season. Without a denominator and without opponent-strength adjustment, every comparison is a comparison between two numbers measured on different rulers.

But the fault here is heavier than a blank list. The entities field reads "identify from the information points above". That is no longer missing data; it is a structural defect: the field points at content that does not exist. A system handed a pointer to nowhere is not under-extracting, it is broken at the extraction tier.

I once tracked one month, 64 matches, and logged every dead-ball situation. Not because I like numbers, but because a dead ball is a window into the whole defensive system: it shows how a team places its players, how it divides zones, how it substitutes. One dead ball says nothing, but nine repeating in the same pattern start to speak. One month, 64 matches, and every dead ball logged; that work has value only because it rests on what actually happened on court. Remove the data and the whole edifice collapses.

Every analytical dimension has a minimum evidentiary requirement, and in this analysis none of them can run. The tactical dimension needs at least a lineup, an attacking scheme or a substitution; there is none, so we cannot even tell whether this is indoor or beach volleyball, men's or women's, club or national team. The data dimension needs at least one quantified metric with a source and scope; all five core metric groups sit at blank. The competition-system dimension needs a competition name, a season and a stage, because the same statement means something entirely different in an Olympic year and in a mid-cycle adjustment year. The landscape dimension needs at least one named team and one opponent to compare; without team names, no contender, medal or quarter-final tier can be assigned. The rules-and-governance dimension needs a specific decision, rule or dispute, and here I decline to infer any compliance issue merely from the existence of a document. The squad-building dimension needs a coach or player name with role and age; without names, age curves, injury risk and public pressure cannot be assessed. The risk dimension needs a subject to attach risk to; without one, the only confirmable risk is analytical-integrity risk. The public-narrative dimension needs a headline and an author stance, and has neither. The industry-transmission dimension needs an industry event — a transfer, a policy change, a rights deal — and there is none.

That is why I cannot, and should not, fill the gap in that analysis with intuition. My intuition is built from hundreds of matches watched, but it has value only when attached to a specific match, a specific team, a specific moment. Without those three, intuition is just prejudice spoken confidently.

There was a time I read a model claiming that high press was thirty percent more effective when stadiums were empty, because players could hear each other call. I doubted the figure, so I watched 15 matches before believing it. I did not believe it because the number was big, but because after 15 matches I could find no way to break it. A number placed in the right spot carries more weight than a whole page of description — but only when the number survives the test. The analysis in my hands has no number to test, so it carries no weight.

But the paradox lies elsewhere, and this is what I want to say.

In sports media, what gets rewarded is not the honesty of a conclusion but the confidence of the person giving it. A piece saying "I don't know, because the data doesn't tell me" has almost no place. It has no catchy headline, no prediction to argue over, no hook to share. A piece confidently asserting that team A will win because their block is better spreads many times faster. That creates a quiet pressure: the analyst is pushed toward having an opinion, toward having a conclusion, even when the data foundation is insufficient.

And when that pressure meets a pre-formatted frame — nine sections, tables, rating boxes — the result is something more dangerous than ignorance: an ignorance that looks organised. A blank page is blank to everyone. A report full of headings but empty inside can slip past a skimming reader. The biggest blind spot in analysis is not analysing wrongly; it is analysing things that do not exist and presenting them well enough that nobody checks.

I do not deny that a ready-made frame is a useful tool. Nine analytical dimensions are a way of not missing the important questions: what is the denominator, is the opponent strong or weak, where is the team's age curve, which risk has not been counted. But a frame only helps you pose questions; it does not answer for you. When a frame is filled with form instead of evidence, it shifts from tool to mask.

I have seen the same at match level. A full statistics table is usually trusted more than an admission of missing data, even though that table may be built from different definitions of the same metric. Perfect-pass rate, for instance, is defined differently by the international federation than by a national league, and differently again by the way a broadcaster summarises it. When the three definitions are mixed, the number still looks complete, but it has lost its anchor. Completeness of form is not the same as completeness of evidence.

One further point deserves to be said plainly: the biggest risk of such a document is not that it is wrong, but that it can be archived, cited and reused as a real assessment. Once it sits in a repository, it is no longer a technical error; it becomes a source. And sources are trusted.

So the lesson I draw is not about that empty analysis, but about our need for a validation gate before any document goes out. An empty information-points list, an entities field pointing at itself — these are clear signals, detectable automatically, and blockable before they dress themselves up as a conclusion.

If the evidence does not permit a conclusion, do we have the courage to say we do not know — or will we keep issuing a beautiful, complete, empty analysis?