Fans discuss transfer fees against published valuation figures as though those figures were an appraisal. A club pays 40 million for a player valued at 25 and it's an overpay. Sells one at 30 who was valued at 45 and it's a bargain for the buyer.
Those valuations are estimates produced by models with specific assumptions, and the assumptions are frequently not what people think they are.
What the models are actually estimating
Broadly there are two families and they answer different questions.
Market-comparable models. These predict what a player would sell for, based on historical transfers of similar players. Inputs are age, position, contract length, league, minutes played, output metrics, and sometimes club reputation. The output is a price prediction.
Value-contribution models. These estimate how much a player is worth to a team in performance terms — how many points he adds relative to a replacement — and convert that into money using the financial value of a league position.
These two produce very different numbers for the same player, and neither is wrong. They're answering "what will he cost" and "what is he worth to us," which are separate questions that only coincide in an efficient market.
Why market-comparable models mislead
The most widely quoted figures are market-comparable, and they have a structural issue: they're trained on completed transfers.
Completed transfers are a biased sample. They're the deals where a buyer and seller agreed, which means they systematically exclude the players nobody wanted at the asking price and the players whose clubs refused to sell. The model learns the shape of the transactions that happened, not the distribution of value.
They also inherit market inefficiencies. If clubs systematically overpay for young forwards from certain leagues, the model learns that overpaying is the correct price, and then reports future overpayments as fair value.
And they're heavily driven by a small number of variables — age and recent minutes at a high level do most of the work. Which is why the valuations sometimes look absurd for players whose value comes from something the model can't see.
The contract term effect
The single most underappreciated driver of transfer fees, and the one most misunderstood in public discussion.
A player with four years remaining and a player with one year remaining can be identical footballers with wildly different prices. The one-year player can leave for nothing shortly, so the selling club's leverage is nearly gone.
Good models include contract length. Casual discussion mostly doesn't, which produces a lot of confused commentary about a club "selling too cheap" when what actually happened was that the club had no negotiating position.
The corollary is that a lot of apparently brilliant recruitment is really just contract timing. Buying good players in the last eighteen months of their deals is a repeatable strategy and several clubs have built reputations on it.
What models systematically miss
A handful of things that matter and don't fit into the input vector.
Tactical fit. A player who is excellent in one system can be ordinary in another. The same set of statistics can represent completely different footballers depending on what their team asked them to do.
Wages. The transfer fee is often the smaller part of the total cost. A free transfer on enormous wages can be a worse deal than an expensive purchase on modest ones, and public valuations almost never account for this.
Sell-on potential. Clubs increasingly buy with resale in mind. A 22-year-old at 20 million who might sell at 45 in three years is a different asset to a 29-year-old at the same price, even if their current contribution is identical.
Agent and intermediary costs. Real money, frequently substantial, essentially never modelled publicly.
Amortisation. Transfer fees are spread across the contract term in club accounts, which changes how a deal interacts with financial regulations. This drives a lot of deal structuring and it's invisible in a headline fee.
The reflexivity problem
Here's the one I find most interesting. These models are now widely available, which means selling clubs read them.
If a public valuation says a player is worth 30 million, the selling club now has a reference point for negotiations, and buyers find it harder to argue for less. The model's estimate becomes an input into the price it was trying to predict.
Over time that pushes the market towards the model rather than the model towards the market. Which makes the model look more accurate while making it less informative — it's measuring its own influence.
This isn't unique to football. Any widely-adopted valuation methodology in any market does this. But football's transfer market is small, opaque and heavily influenced by public narrative, so the effect is unusually strong.
How to read a valuation
Practical suggestions if you're going to use these numbers.
Treat them as a rough distribution rather than a point estimate. A figure of 30 million means something like "similar players have sold for somewhere between 18 and 45 million," and the range is the useful part.
Always check the contract situation before concluding anything about whether a fee was good.
Ignore any comparison between fees paid in different financial environments. Broadcast revenue and regulation change what clubs can spend, and a fee from six years ago isn't comparable to one today in any straightforward way.
And remember that a transfer fee is the price of an asset, not an assessment of a footballer. A club overpaying doesn't mean the player is bad. It usually means somebody else was also bidding.