The Transfer Window Mis prices Strikers — and the Data Sheet Warned Us Three Seasons Ago

In this transfer window, a striker's price is set not by the quality of his finishing but by the goals he has already scored. The data shows the gap between those two things is where the most money gets burned.

1. The eleventh column

On August 14, when a Premier League club announced a 68 million pound deal for a 24-year-old striker, I did what I have done for nine years: opened last season's data file, scrolled to the eleventh column, and read the number no reporter asked about at the unveiling press conference.

His non-penalty xG was 13.8. His actual goals were 22. The difference: plus 8.2.

In a sample of 118 strikers I have tracked across Europe's five major leagues from 2026/20 to 2026/24, only nine sustained an overperformance above plus 3 goals across two consecutive seasons. Seventy percent of players who posted an overperformance above plus 5 in one season fell below plus 1.5 the next. The sheet does not say the player is bad. It says the market just paid for an event, not a process.

Data does not lie. The listener just has not been patient enough.

2. Context: a market priced by three groups with different incentives

The transfer window is the only sports market where the listed price is set neither by the asset's user nor by its seller, but pushed along by three intermediary groups with entirely different motives.

First, agents. Their commission is a percentage of contract value, not a percentage of goals in season three. Selling a player for 68 million instead of 42 million creates a commission gap that can reach seven figures, and that gap does not depend on whether the player scores 20 the following year.

Second, sporting directors. This role has a far shorter average lifespan than the contracts they sign. A deal announced today produces articles, engagement and a line in the quarterly report. Whether it works out gets answered in 24 months, by which time the person who signed it may be at another club.

Third, head coaches. They are the only group under pure results pressure, and the group with the least say in pricing. Coaches are sacked on league tables, not on xG. Which means: the person making the football decision is not the person paying, and the person paying is not the person absorbing the consequence.

When three incentives drift out of phase, the market still runs smoothly — but it runs on the logic of whoever holds pricing power, not the logic of whoever holds the table. That is why every summer produces a group of strikers bought at prices their own buyers cannot explain three months later.

The transfer window is a chess game in which most people only see the pawns.

3. Four layers of data that decide real value

I build my striker pricing model on four layers, and the lowest layer in the spreadsheet is always the one that matters most.

Layer one: shot volume, not conversion.

A striker taking 3.4 shots per 90 and one taking 1.6 shots per 90 can finish the season on the same goal tally. In my data, shot volume correlates with next-season goals far more strongly than conversion rate does. Put differently: the player who creates more chances is more likely to score again. The better finisher is simply sitting at the peak of a curve every player must descend.

This is why Erling Haaland is the exception that forces every regression model to accept its own error margin. He is not in the high overperformance group. He is in the high shot volume, high shot quality, high box-arrival group. Those three are what repeat.

Layer two: the quality of the position where the ball is received.

I do not care how many times a player shoots. I care where he receives the ball before shooting. In my dataset, a shot taken after receiving the ball inside 12 metres of goal is worth many times a shot from outside the box, regardless of who the player is.

So the metric I track most closely is not xG but touches inside the box per 90. Few people scroll to that column. It does not appear on the scoreboard. It generates no highlight clip. But it is the single best predictor of next-season goals I have ever validated.

Layer three: team context.

A striker scoring 18 goals for a side with 62 percent possession that constantly feeds the box is not equivalent to a striker scoring 18 for a side with 41 percent possession that counterattacks and generates 0.9 box entries per match.

When I normalise goals for the chances a team creates, the price list reorders sharply. Some names sit in the top ten scorers of a league yet drop to the thirties once divided by supplied chances. Others score only nine but rank top ten on the same adjustment — and those are typically the players sold below their true value.

This is the point most transfer reports miss: they read the numerator of a goal tally, I read its denominator.

Layer four: minutes load and the age curve.

A 24-year-old playing 3,100 minutes a season for three straight seasons sits in a completely different risk zone from a 24-year-old playing 2,100. Minutes load appears in no transfer bulletin, yet it predicts injury better than any interview about physical foundations.

I wrote about Jesse Lingard in 2026, when global football was suspended and I had just taken a 30 percent pay cut eight months into the job. The piece showed he covered 11.2 km per match but contributed only 0.2 goals and assists per match, and concluded the problem was the system, not the player. The following season he scored nine goals in 16 games for West Ham.

What I learned from that was not that the call was right. It was the method: when a player is mispriced, the reason is usually that his most important metric sits in a column the market is not reading.

4. The columns nobody scrolls to

Most people watch the score. I watch the rest of the sheet.

The four metrics below are the columns I believe will determine transfer value over the next 24 months, and right now they are almost entirely unpriced.

PPDA (passes allowed per defensive action). I used this to analyse Croatia's first five matches at the 2026 World Cup, when their average PPDA was just 9.2 — meaning opponents had almost no time on the ball before being closed down. When Croatia reached the final, many called it a surprise. To me it was data published before the tournament.

Ball recoveries in the opponent's third. This column measures the ability to generate chances from opposition errors, and it does not depend on whether teammates pass to the player. A striker with a high score for this at a weak club is a systematically underpriced asset.

xGA (expected goals against) by zone. Analysing Morocco at the 2026 World Cup, I found they held an average xGA of 0.3 per match, the lowest in the tournament, alongside 14.2 successful central tackles per match. Spain held 78 percent possession against them. I wrote that they would be powerless. The match ended in a penalty shootout, and I do not treat that as luck — I treat it as the output of a defensive block organised around data.

Share of goal involvement excluding set pieces. A striker with 14 goals, six from penalties and four from corners, has actually contributed four goals from open play. That column never appears in the golden boot table, which is exactly why it matters.

One number is an accident. A cluster of numbers is a confession.

5. When correlation gets read as causation

Most transfer pricing errors do not come from misreading data. They come from reading the right data and assigning the wrong causal relationship.

A club sees Striker A score 24 goals and concludes: buy A, get 24 goals. But the data only says A scored 24 inside a specific system, with a specific midfield, against a specific fixture list. Goals are an output of a system, not a property of an individual. Sell the system and keep the individual, and the club has not bought 24 goals. It has bought a probability.

I have made this mistake in reverse, too.

In 2026, as a second-year student in Binh Duong, I collected Long An's numbers across the first 20 rounds of V-League. They generated 2.1 xG per match but scored only 0.8, while opponents with less possession converted better. I wrote that they would survive if they kept their coaching staff. The club sacked the coach before the return fixtures. They were relegated with 21 points.

That article was right about the model and wrong about the system. I calculated xG correctly but failed to calculate that the sacking decision was never inside my model. Since then, every spreadsheet of mine carries an extra column labelled the human variable — a column with no data, but with a weight.

This is where I think sports analytics still falls short: we are getting better at measuring the process of play, and still poor at measuring the process of decision-making. A model that predicts a match result correctly but cannot predict an ownership losing patience remains an incomplete model.

Two positions I keep defending in my work both flow from that principle.

First, the Saudi Pro League is not developing its football by buying stars past their peak. The contract structures, durations and the way the deals are announced show the objective is not club-level performance. Those players are brought in as tourism ambassadors, not to raise the standard of the league. Playing-time data and the volume of accompanying commercial activity move in clear inverse proportion.

Second, an amateur side reaching a major final usually does not prove its academy system works. It proves it drew a favourable bracket and peaked for exactly one match. I checked the sample of surprise deep runs in cup competitions over ten years: most of those clubs returned to their prior level within two seasons. One match does not make a system. Three seasons is worth examining.

6. My own margin of error

I do not write to be agreed with. I write to be checked.

That means accepting that part of my model will be wrong, and that I must be the one to publish it before someone else finds it.

My pricing model has at least three known weaknesses.

The first is dependence on input data quality. In leagues without standard positional tracking, I have to use manually coded event data, and error can reach 15 percent. Southeast Asian leagues, V-League included, fall into that bracket. Every conclusion I draw about regional football has to be read with that warning attached.

The second is that the model cannot yet quantify the value of off-ball pressure. A striker whose run drags a centre-back out of position can create a goal without touching the ball. Current tracking systems record the movement but cannot assign goal value to it. I am testing a variable based on space created, but it is not stable enough to publish.

The third is that the model does not measure psychological pressure. A striker fighting relegation in round 30 carries a different load from one at a club already assured of European football. I once underrated a player simply because his season unfolded inside a squad disintegrating internally. Spreadsheets do not display dressing rooms.

This is why I tell young editors: before you criticise a player, go back and check your own database.

7. Signals for the next transfer cycle

Three signal groups are worth watching in the coming window.

The first is players with high box-touch rates but low goal totals. This is the most underpriced group on the market, because their price is computed from the goals column rather than the chances column. When a side with a creative midfield signs one of them, it is a sign that club is reading the right sheet.

The second is players whose minutes load spiked sharply over the last two seasons. This is the highest injury-risk group, and the group whose medical reports are least fully disclosed in transfer dossiers. A big deal for a player in this bracket usually carries an insurance clause nobody mentions at the press conference.

The third is release clause structure. The transfer fee is the published number, but the payment structure decides who actually carries the risk. A 60 million pound deal paid over four years means something entirely different from a 45 million pound deal paid up front, and a club's balance sheet reflects that more accurately than any bulletin.

Across thirteen years watching this industry, one thing holds: every transfer cycle contains a group of people who believe they are buying a player, when in fact they are buying a season that has already happened.

My sheet will keep updating. And if three months from now these numbers prove me wrong, I will be the first to write about it.

The numbers say the opposite. Believe it or not, that is your call.