From 10 Million to 70 Million Euros: Transfer Multiples and the Data-Labelling Trap

In July 2026, Napoli signed Khvicha Kvaratskhelia from Dinamo Batumi for a reported fee of around 11.5 million euros. Two and a half years later, in January 2026, Paris Saint-Germain paid roughly 70 million euros to take him out of Serie A. A multiple of nearly six.

That number ran through every news bulletin. By the time it reached Southeast Asia it had become a different question: if Napoli can do it, why can't a V.League club? I sat in that regional analysts' call some time later, listening to four people argue about academy costs, wage bills and agent relationships. The division was not wrong. The label stuck onto it was.

That is why this piece starts with a mislabelled data column rather than a match.

Value multiple: a tool borrowed from corporate finance

A transfer value multiple divides exit value by entry cost. Buy at 10 million, sell at 70 million, and the multiple is seven. A player who comes through an academy, signs a first professional deal internally and is later sold for 20 million produces a paper multiple close to infinite, because the denominator is close to zero.

The concept entered European football at scale once investment funds began buying club shares, which is to say once football had to speak the language its investors understood. It spread through Southeast Asian sports journalism very fast, because a multiple is perfect headline material: short, numeric, and carrying a feeling of objectivity.

The problem is that a multiple does not travel with its own definition. Every time a data column is named value, the reader at the other end fills in whatever they already want to believe.

In 2026, a story from the film industry exposed the technique. A film with a production budget of about 10 million dollars returned more than 400 million dollars in global box office, a multiple of roughly 40 times on production cost alone. Sports media jumped straight in: the creator economy will reshape football, internet-native talent will change how clubs scout, teams must learn to produce content. Then an automated content pipeline tagged that article as football. The original text contained no player, no club, no competition. The classifier simply saw the words record, result and market share, and filed it under sport.

I bring that up because it is the cheapest possible test for anyone working with football data: the most serious error is usually not inside the model, but in the label applied before the model runs. Mislabeled data is worse than missing data, because missing data stays silent while mislabeled data is very confident.

The evidence chain: where multiples actually come from

Rasmus Højlund left Copenhagen for Sturm Graz in January 2026 for a reported fee of about 1.8 million euros. Six months later Atalanta paid around 17 million euros. In August 2026, Manchester United paid roughly 72 million euros, add-ons excluded. Against that first fee, the multiple is close to 40 times inside 19 months.

Julián Álvarez joined Manchester City from River Plate in January 2026 for about 14 million pounds, then moved to Atlético Madrid in August 2026 for around 75 million euros plus add-ons. A multiple of roughly 5.5 times.

Jadon Sancho left Manchester City's academy for Borussia Dortmund in 2026 for around 8 million pounds and returned to Manchester United in 2026 for 73 million pounds. Nearly nine times.

Victor Osimhen went from Wolfsburg to Charleroi, then Lille, then Napoli. Each step multiplied the price; the final one reached about 70 million euros in 2026.

And the ceiling is still Neymar: 222 million euros, Paris Saint-Germain, 2026. No transfer in football history has ever touched 400 million euros on the fee line.

That last detail matters. The 40-times multiple the film industry can advertise has never appeared in football at absolute scale. Football has high multiples, but in small number pairs: 1.8 million turning into 72 million. Once a deal enters nine figures, the multiple has usually already run dry.

The V.League is different at the root. Most clubs carry no meaningful transfer revenue line at all. Academies such as Hoang Anh Gia Lai, PVF and NutiFood develop players and promote them internally, and their economic value sits in the wage bill they save rather than in fees they collect. The overseas moves of Nguyễn Công Phượng, Nguyễn Quang Hải and Đoàn Văn Hậu have mostly taken the form of loans, short contracts or free transfers, where the parent club records almost nothing on the balance sheet.

In other words, the denominator in Vietnamese football's formula is usually zero, which makes the paper multiple look good in a meaningless way. Sell an academy player who cost no transfer fee for 200,000 dollars to a Thai League side and the division returns infinity. An infinity sitting in a report is a sign of a data-entry error, not of a transfer genius.

In Vietnam, the player-export argument is usually built on a single sample. One player moves to Japan, starts a few matches, and analysis pieces immediately declare that domestic academies have reached export standard and that the next generation will keep the pipeline flowing. One case does not create a rule. It creates one data point, and one data point is not enough to draw a line.

There is one more layer the bulletins skip: hidden cost. The transfer fee is only the visible part. Wages, signing fees, agent commissions, medical costs and adaptation time are the submerged part, and no public statistics column records all of them. Divide an output number by an incomplete input number and you do not have a multiple. You have a dressed-up estimate.

What the bulletins leave out

A multiple is only calculated after the fact. It is a label applied to survivors, not the output of a scouting model. Napoli did not know Kvaratskhelia would produce a six-times multiple; they knew the maximum loss was 11.5 million euros and the upside had no ceiling. That is an option payoff, not a forecast.

For every Højlund there are dozens of academy players who never generate a fee at all. I have spent long enough in data rooms to have seen academy tracking sheets where the conversion rate into the first team sits at one in fifty, and the rate of selling abroad is lower still. All models are wrong, but a few are usefully wrong. The useful kind of wrong here is accepting that the average multiple across an entire academy intake is far lower than the multiple of the one player chosen to tell the story.

The real risk in a buy-low-sell-high model is single-point dependency. The film story illustrates it better than any league table: a distributor's market-share record is concentrated in one single title, and a sequel is unconfirmed. Football has a replica of that structure: a V.League club funding next season's budget by selling one player to a K-League or J-League side. When that flow jams, the club does not weaken gradually. It stops.

The counter-intuitive angle

People like to say the multiple proves the data model is winning. Correlation is not causation. A player's multiple depends on minutes played, on price inflation across the entire transfer market, on years left on the contract, on whether the selling club needs cash, and on an agent who happens to make the right call at the right time. That list of confounders is far longer than the model itself. xG does not score goals, but it generates more argument than the ball itself, and the transfer multiple behaves the same way: it does not score, it just records what already happened and then styles the story.

So I keep one rule: until the confounders are excluded, the word random is not allowed. Football stopped rolling in 2026, but randomness has never taken a lunch break, and precisely for that reason, every time I am about to blame luck for a result, I have to answer how many items on that list I actually controlled.

There is one failure I still use as a lesson. At the 2026 World Cup, my model was right when it predicted South Korea beating Germany, and I said so far too loudly. In the next round the model picked Brazil over Belgium, I said so loudly again on live television, and Brazil lost 1-2. The lesson was not that the model was weak. The lesson was that I let a probability turn into a promise. Since then every analysis of mine carries one line: a model is a probability, not a prophecy.

One more blind spot few people will say out loud: the buyer at the top of the multiple is always the one who absorbs the loss. When a club pays 72 million euros for a player once valued at 1.8 million, it is buying the previous seller's expectation plus a risk premium. The seller's beautiful multiple is the buyer's deferred expectation debt. The bulletins only print one half of the story, and it is always the more comfortable half.

Signals for the next cycle

What is worth tracking in Vietnamese football is not whether one more player goes abroad, but whether clubs begin recording transfer revenue as its own line in their financial statements. Once that line exists, the multiple finally has a real denominator to divide by. Vanishing data is not the loss of data; it is a kind of data, and in the V.League, data on actual transfer fees is vanishing in exactly that sense. Until then, every beautiful number is just a column named too generously.

Every spreadsheet is a meditation, except that when the meditation ends you have lost money. If you are about to use a multiple to persuade someone to put money in, ask yourself first: that column you are dividing by, what exactly is it labelled?