When the Scouting Dossier Comes Back Empty: Data Discipline in the Italian Volleyball Transfer Market

6:47 a.m., Milan time. A twelve-page PDF sat in my inbox, its subject line naming an outside hitter currently playing in the SuperLega. I opened page one: the attack-metric table was empty. Page three: the block-metric table was empty. Page seven: the tactical assessment contained a single English line, saying roughly that there was insufficient information to draw a conclusion. Twelve pages, not one figure.

Forty minutes from then, a club would lock its roster for the European cup. They needed me to confirm a proposed salary for that contract. In my hands was an empty dossier.

I sat still in front of the screen for a while. Not out of confusion. I sat still because I knew exactly what I was about to do, and I knew it would annoy a few people further down the chain. Eventually I picked up the phone, called the scouting department, and said one sentence: this dossier is unusable.

My job in Milan is to read player dossiers and price them for the Italian market. Volleyball is the main beat; I still follow football out of professional habit. A scouting dossier passes through two stages. Stage one deconstructs: it establishes the article title, the source, the information points, the entities named, and the time sensitivity. Stage two analyses: tactics, data, schedule, team positioning, rules and governance, roster structure, risk, public narrative, and the industry transmission chain.

When stage one comes back empty, stage two can do nothing but return an empty report. That is a different category entirely, not a low rating. A player rated two out of five is still a player with data to argue about. A player with no data has nothing to argue about at all.

I learned that distinction fairly late. In 2026, while I was a sports-journalism student in Milan, I wrote a blog analysing advanced metrics in Serie B. I tracked fourteen matches of Patrick Cutrone, noticed that his expected-goals rate per minute of playing time sat well off the baseline, and predicted he would score at least eight goals the following season despite being nineteen. Cutrone scored ten goals in Serie A 2026-18. An editor at a local sports paper read the blog and invited me to contribute.

The point is not the ten goals. The point is that I stated the sample size in the opening paragraph: fourteen matches. Had I hidden the sample size, the prediction would still have been right, but it would have become a cheap prophecy. The 2026 World Cup taught me one lesson: a model does not need to be big, it needs to be correct.

In the summer of 2026, when European football returned behind closed doors, I collected data from 412 matches and set it against 412 matches from the same period in 2026. Home-win rate fell from 46% to 36%, and average goals per match dropped by 0.4. I carried that controlled-comparison method across into volleyball: data never lies, only readers rush. The problem almost always sits with the reader, with whether they bothered to build a control group.

Back to the PDF. That empty dossier carried three very specific messages, and all three were about the data pipeline, not about the player.

First, the person who built the dossier had a source but skipped the deconstruction step. They knew the player's name, the owning club, the position. They could not establish a single verifiable information point. In scouting, a name and a position are not data. They are labels.

Second, no entity was fully identified. No club was recorded with a specific season, no competition with a round, no date appeared anywhere. A report that cannot identify its entities cannot extrapolate. You cannot compare a performance against its own previous season if you do not know which season that was.

Third, there was no assessment of time sensitivity. That is the most expensive error. A club locking its European roster on Friday and a club planning for next season need two different kinds of information. Same player, same metric set, entirely different utility.

At this point I should be explicit about how I check volleyball data, because this is the part no raw metric table will ever do for you.

The first layer is sample size. Spike success rate is the most abused metric in volleyball. A player hitting 52% across the first seven matches of a season can finish at 46%, and nobody calls that a decline. Fifteen matches create a trend; one match creates a memory. If you are going to pay wages against a metric, know how many rallies that metric was computed on.

The second layer is opponent adjustment. An outside hitter at 52% against bottom-half teams who drops to 41% against the top four is an entirely different asset from one who holds 47% evenly across every tier. Identical sample sizes, opposite conclusions. This is where cheap scouting reports collapse.

The third layer is the timestamp. Blocks per set from two seasons ago carries almost no pricing power. In the SuperLega, the scouting department's decision speed outruns the publication speed of many statistics systems. Anyone pricing with stale data is paying for the past.

Now the part I consider most important. Blocks per set does not measure blocking ability. It measures the final outcome of a dependent chain: the serving quality of a teammate, the opponent's distribution, and the positioning of the blocker next to you. A middle blocker with a high block count is sometimes just standing beside a better middle blocker.

The ace-to-error ratio works the same way. A server with 12 aces and 25 errors is not better than one with 6 aces and 6 errors, especially in decisive sets. Perfect-pass rate is different: it determines whether a coach can run a middle attack at all, and therefore it determines the value of the setter. That is why the market pays enormous sums for names like Wilfredo León or Earvin Ngapeth, while single metrics never fully explain those price tags.

Based on my experience watching matches in the SuperLega and the European cups, I work by one rule: every metric must come with an answer to what it means on court. A metric with no on-court meaning is decoration for a report.

Inside those seventy-two hours I rebuilt the dataset by hand: fourteen matches, tagged by opponent strength tier, with match dates recorded. A small sample, and I said clearly that it was small. I do not argue with emotion, I argue with sample size. The result was not enough to price the player, but enough to define conditions. I sent back a single page: no recommended salary, only three conditions that would have to be met before the conversation could continue.

The club signed the player anyway, but restructured the deal: a lower fixed salary than first proposed, higher performance bonuses. They did not lose the player. They lost a little time, and I consider that cheap.

This is where the counter-intuitive part belongs, because it is the part most easily skipped.

An empty dossier does not prove the player is weak. It proves the person who built the dossier has not done the work. Those two conclusions drive opposite actions: the first sends you looking for another player, the second sends you fixing the process. Clubs usually choose the first because it is faster, then three months later receive exactly the same empty PDF from a different provider.

The second risk is metric FOMO. Volleyball has no expected-goals metric yet, but it is producing copies of one: numbers that look thoroughly modern, computed by models, cross-checked against video by nobody. I once saw a report ranking an outside hitter on a composite index whose author had not watched three sets of that player. That is not analysis. That is decoration with spreadsheets.

And the third risk, the most dangerous one for the analyst: being willing to throw away an attractive argument for lack of data. More than once I have written a passage that read beautifully, found it persuasive on rereading, and deleted it because the sample size would not carry it. Error is not the enemy; it is the silent teacher of every model. But error only teaches if you are willing to write it down instead of hiding it behind an adjective.

Correlation is not causation, and in a small market like Italian volleyball, where only a few dozen meaningful contracts are signed each season, any correlation can look like causation if you only look at one season. A middle blocker coming from the team with the best reception line in the league arrives with a beautiful metric set. Put him on a team with weak reception and that set evaporates within six sets.

The direction I believe is right over the next two seasons is not buying another expensive data system. It is accepting that “not enough data to conclude” is a legitimate answer, and paying for the person willing to give that answer instead of the person who always has a number ready. In the SuperLega, clubs that build their own deconstruction layer will price players one step faster and cheaper than clubs that outsource the whole pipeline.

That is why I still keep that twelve-page PDF in a separate folder. Not as evidence against anyone. I keep it to remind myself that a formally beautiful report can contain exactly zero information, and that data discipline starts at the humblest point: state the sample size before you state the conclusion.

Next transfer window, the first question I will ask the scouting department is not who they found, but what they managed to deconstruct.