The 12-Metric Spreadsheet: What Brentford Read Before the Market Priced Ollie Watkins

1.8 million pounds, 28 million pounds, and three years without new information

In July 2026, Brentford paid Exeter City 1.8 million pounds for a 21-year-old striker who had just scored 13 goals in League Two. In September 2026, Aston Villa paid 28 million pounds for the same player, plus up to 5 million pounds in performance-related add-ons. Between those two dates, no goal made the price rise fifteenfold. There was no surgery, no positional conversion, no recorded physical breakthrough. The numbers used to justify the 28 million pound fee had been sitting intact in the data since May 2026.

The only thing that changed across those three years was the number of people willing to sit down and read them.

I once believed that prices reflect information. If a player is undervalued, there must be a reason: an undisclosed injury, an attitude problem in the dressing room, or a technical ceiling the cameras do not capture. Three months spent tracking Brentford in 2026 destroyed that belief. Brentford were not smarter than the rest of England. They simply read the same category of data with a different decision rule.

Data is never in a hurry, but people always are. The transfer market is a match in which whoever prices correctly wins, and in that match most of the money is not spent by the most careful reader but by the one afraid of being left behind.

Three months, 1,247 players, 15 leagues

In 2026, at the age of 51, I was working as a transfer market administrator for a sports consultancy based in London. My job was to value players for client clubs, which meant that every week I had to answer the same question: what is this player worth, and at what price does buying him stop making economic sense.

I spent three consecutive months on Brentford, then still in the Championship. This was the club whose owner Matthew Benham, founder of the betting data company Smartodds, also owned FC Midtjylland in Denmark, and who had publicly said he bought the club because he believed in a model rather than in inspiration. The sporting leadership at the time included Rasmus Ankersen and Phil Giles, two men running recruitment on a budget smaller than almost every rival in the division.

I built a dataset of 1,247 players from 15 European leagues, restricted to those born between 2026 and 2026, the cohort the market had not yet fully priced. From that dataset I filtered 38 potential targets. Not the 38 best players. Thirty-eight players with the largest gap between measurable ability and the price the market was quoting.

When Brentford signed Ollie Watkins from Exeter City for 1.8 million pounds, and later sold him to Aston Villa for 28 million pounds, I understood something I had never written in any previous report: data is not a supporting tool for recruitment. Data is a strategic weapon, and the market only discovers that after the invoice has been paid.

From then on I never wrote a transfer judgment based on a player's reputation or a feeling from the stands. I start with a data table, cross-check three independent sources, and only then allow myself to write a conclusion.

Twelve metrics, and why there had to be exactly twelve

After those three months I built my own analytical framework of 12 metrics. Not because the number 12 carries any mystical meaning, but because at metric 13 I began adding variables that did not change the conclusion. A good framework is not the one with the most variables. It is the one where removing any single variable changes the output.

The first group covers chance creation. I use xG per 90 minutes, normalised by actual minutes played rather than appearances, because a player who comes off the bench 20 times is not a player who started 20 matches. Alongside it sits xG per shot, the metric that separates the player who creates many chances from the player who creates good ones. A striker with 2.4 shots per match at 0.08 xG per shot is a player shooting from outside the box because nobody is passing to him. A striker with 1.6 shots at 0.19 xG per shot is a player being fed the ball exactly where he needs it.

The second group covers ball progression: progressive carries per 90, passes into the box per 90, and involvement in shot-ending sequences. Together these three answer one question: is this player inside his team's attacking flow, or merely standing near it.

The third group covers pressure. Here I use PPDA, the number of passes an opponent is allowed before each defensive action, alongside regains within five seconds of losing the ball in the opponent's third. These two must travel together. PPDA on its own is a meaningless figure, and I will return to that point later.

The fourth group covers aerial duels and wide defending: aerial duel win rate in central zones and defensive actions in wide corridors. This is the group that immature data models routinely skip, because it does not generate highlights.

The fifth group, and the most important one for Brentford, covers valuation. It includes age adjusted by the minutes-played curve, senior professional minutes before the age of 21, and finally percentile-based price discounting.

Value does not lie in a player's absolute ability. It lies in the gap between his ability percentile and his price percentile.

A player in the top 15 percent of his league for progressive carries, but paid at the 45th percentile, is a mispriced asset. A player in the top 5 percent, already paid at the top 2 percent of the market, is correctly or over-priced. The market looks at these two players and sees the second as better. The spreadsheet looks at them and sees the first as more profitable.

That is the whole secret. There was no miraculous algorithm at Brentford. There was a different decision rule.

The valuation table: what seven transfers actually say

I always want readers to see the logic for themselves rather than trust my prose. Below are seven representative Brentford transfers between 2026 and 2026, with figures reported by the English media at the time of each deal. Transfer fees always carry a margin of error, because contracts include performance add-ons and no club publishes the full picture. But even after stripping out every possible discrepancy, the gap is far too large to be explained by luck.

Ollie Watkins, from Exeter City for 1.8 million pounds in July 2026, to Aston Villa for 28 million pounds in September 2026. In his final Championship season he scored 25 goals in 46 matches. But the important metric was not the 25 goals. It was his xG per 90 at Exeter, and a progressive carry profile showing he received the ball in the left channel more often than a conventional centre-forward. Brentford bought an unfinished player. Aston Villa bought a finished one. The 26.2 million pound difference is the price of patience.

Neal Maupay, from Saint-Etienne for around 1.6 million pounds in July 2026, to Brighton for around 20 million pounds in August 2026. He scored 37 goals in 85 Championship matches across two seasons. The notable figure lies elsewhere: in his first season he scored only 12, and in his second he scored 25. Had Brentford assessed him after one season, they would have sold at five million pounds. They assessed him on chance-creation metrics, which are more stable than goals, and waited.

Said Benrahma, from Nice for around 1.5 million pounds in July 2026, to West Ham for around 25 million pounds in 2026 after a loan year. His profile on arrival was unremarkable except for progressive carries per 90, which sat in Ligue 1's highest bracket despite his almost complete absence from the starting eleven.

Ezri Konsa, from Charlton for around 2.5 million pounds in July 2026, to Aston Villa for around 12 million pounds in July 2026. One season at Brentford. This was the fastest profitable exit in the group, and it showed the model worked even when a player did not stay long enough to develop.

Chris Mepham, a Brentford academy product, to Bournemouth for around 12 million pounds in January 2026. Purchase cost: zero. This is evidence that the model was used not only to buy outside the club but to decide who deserved promotion to the first team.

David Raya, from Blackburn for around 3 million pounds in July 2026, to Arsenal for around 27 million pounds in 2026 after a loan season. For goalkeepers, the metric I use is not saves, but the accuracy rate of long passes under pressure and positioning while the team is in possession.

Bryan Mbeumo, from Troyes for around 5.8 million pounds in August 2026, and Ivan Toney, from Peterborough for around 5 million pounds plus add-ons in August 2026. Toney scored 31 Championship goals in 2026-21, a divisional record for a single season, before moving to Al-Ahli in 2026 for a reported fee of around 40 million pounds.

In total: under 16 million pounds of purchase fees for the group above, and more than 160 million pounds in sales for those who left. A ratio of roughly ten to one, across six years, at a club that never had the budget of a leading group.

But here is the part the media routinely skips. Brentford were promoted to the Premier League on 29 May 2026, beating Swansea 2-0 in the play-off final at Wembley. The squad that day was not composed only of data signings. It included players recruited by conventional methods and, more importantly, it was coached by Thomas Frank, who built a system in which the metrics of the bought players could actually convert into points.

The Mbappe case: the market prices what happened, data prices what is repeating

In June 2026, the World Cup in Russia took place while I was 52. I did not go to Moscow. I stayed in London, rented a small flat, and set up four screens tracking twenty matches simultaneously through movement data.

After the group stage I published a 4,000-word analysis on my personal blog. The central argument was not that Kylian Mbappe would shine, which anyone watching for ten minutes could guess. The central argument was that France would win not because of a famous attack, but because of the space Mbappe stretched open.

The numbers at the time: a top speed of 38 km/h, the highest at the tournament. But the more important figure was acceleration: from a standing start to 30 km/h in 4.5 seconds. In football, top speed only matters if the player is already running. Acceleration from a standing start is what creates an un-defendable break, because it happens inside the window in which a centre-back cannot turn his body.

When France won, the article was shared more than 12,000 times. An editor from The Athletic got in touch to offer a collaboration. But what I took from it was not that I had predicted correctly. It was that the data on Mbappe's acceleration had existed since the 2026-17 season, when he was still at Monaco, and it was in the hands of at least three major European clubs. The market only paid for that profile after the world had seen it on television.

The market prices what has happened. Data prices what is repeating.

Mbappe's speed is not frightening. What is frightening is the speed at which data recognised him beforehand. The same logic applies to every player in every league. On television, people see the ball. In the data, people see the space. The ball only arrives in that space if someone created it three seconds earlier.

PPDA and the trap of the pressure metric

No metric is more abused than PPDA. It is the number of passes an opponent is allowed before each defensive action. The lower the figure, the more aggressively a team presses. A team with a PPDA of 6.8 gets described by the media as pressing high.

The problem is that PPDA cannot distinguish two opposite situations. Situation one: the team presses proactively, regains the ball in the opponent's third, and a low PPDA is a sign of strength. Situation two: the team is being played through repeatedly and is chasing the ball in a structurally broken shape, and a low PPDA is a sign of panic.

Two teams with the same PPDA of 6.8 can sit at opposite ends of tactical quality. In my tracking work in the Championship, I learned to always read PPDA alongside three things: the location of the regain, the scoreline at the moment the metric was generated, and the number of passes the opponent completed before being stopped. If the opponent loses the ball in midfield, that is pressing. If they lose it in defence, that is counter-attacking. If they do not lose it at all and are merely slowed down, that is a pretty number on a report and meaningless on the pitch.

This is where I set a rule for myself: every metric must answer the question of who benefits when the number improves. PPDA improves when a team runs more. But running more over 90 minutes, at the tempo of modern football, is not a free advantage. It is a loan.

The five-substitution rule and the war of attrition in the final twenty minutes

When five substitutions were made a permanent fixture from the 2026-23 season, most analysis focused on the benefit to big clubs with deeper squads. That is correct, but it is not the whole story.

What actually changed was the time structure of the match. Previously, the third substitution was a strategic decision that had to be saved. Now the final twenty minutes become an organised war of attrition, in which both teams keep introducing players with different physical profiles to attack the same weak point.

The transfer consequence is clear. The value of the twelfth to sixteenth players in a squad rises, because they are no longer contingency options but second-half weapons. For a model like Brentford's, this is an underpriced opportunity: the market pays for the best eleven players, while the points of a season are decided by sixteen.

I do not need a complex model to see this. I only need to place side by side the minutes played by the group of players who come on after the 70th minute across three consecutive seasons, and the points won in that period.

The heat map is the new astrology

There is one metric I deliberately left out of the twelve-metric framework. That metric is the heat map.

Heat maps appear more and more in coverage, as glowing red trails spread across half the pitch. They convey a sense of quantification, a sense of science, and they contain almost no conclusive value.

A heat map describes where a player stood, not what the player did there.

Two central midfielders can have almost identical heat maps and perform opposite roles. The first receives between the lines, turns toward the opponent's goal, and breaks the defensive structure with a line-splitting pass. The second drops between the centre-backs to receive, only to move the ball wide and set the tempo. Both have a red patch in the same zone. One creates goals. The other creates safe passes.

The heat map is not bad data. It is data read in the wrong place. It answers a positional question, while the tactical question is always about function. In my work, a player is only assessed once I know where he receives the ball relative to the opponent's defensive line, and what he then does with it.

When a metric becomes popular enough that everyone cites it, it stops providing an advantage and starts providing comfort. Astrology was once that popular. It was used to explain wars and marriages. The problem is that those who use it cannot tell the difference between observation and interpretation.

Brentford do not read the future

This is the part I want to write most clearly, because I know it will upset people who sell models.

Brentford do not read the future. They merely read data more carefully than everyone else.

That distinction is enormous, because if they could read the future, they would be able to reproduce the success in any environment, with any coach, under any budget constraint. They cannot. What they have is a consistent decision rule in a market where most participants decide by reaction.

And here is the point I must state plainly: there is a survivorship bias in every analysis of Brentford, including the ones I write. We all remember Watkins, Maupay, Benrahma, Toney, Raya. None of us can name three players from my 2026 list of 38 targets who never reached the Championship. I keep only the successful names myself, because those are the names with public records to cross-check.

I have asked myself that question directly: if the model's hit rate is only one in three, where does its value come from. The answer lies in the cost structure. When the average purchase price of a cohort is 1.8 million pounds and the average sale price of the successful ones is 25 million pounds, the model does not need to be right half the time. It only needs to be right one time in five to be profitable, provided the club has enough patience not to sell unfinished players at a low price.

This is why I am always wary of stories about the person who predicted correctly. They do not describe a mechanism. They describe an outcome. Every football cycle imitates the data of the previous cycle, but nobody learns.

The imitator's trap

After 2026, as the Brentford story became widely known, the price of players labelled as data players rose. A 21-year-old full-back in the lower divisions with strong progressive carry numbers no longer cost 500,000 pounds. He cost four million, because every club in the Championship and League One had heard that the metric was valuable.

This is the nature of arbitrage: it exists precisely because few people see it. The moment a model becomes common knowledge, it does not disappear, but it is compressed. Those who arrive later do not buy the asymmetry. They buy a process.

And a process does not automatically generate profit. It generates profit when there is a coach willing to build a style of play that fits the profile of the players that process identifies. Brentford did not only select players by data. They chose a playing style suited to the type of player the data indicated was underpriced. Those are two separate decisions, and most imitating clubs make only the first.

Based on my experience watching matches in the Championship across several consecutive seasons, I see the same script repeat. A club hires an analyst, buys three players with good metrics, places them in a system that creates no space for those metrics, and then concludes that data does not work in the lower divisions. Data did not fail in those cases. The integration failed.

Correlation is not causation, and that matters more than it appears

There is an implicit assumption in most Brentford content that I cannot prove: that the data itself produced the success. What we actually know is a correlation. The club used more data than average, and the club achieved better results than average over the same period.

At least four variables sit between those two sides, and nobody has separated them.

The first is coaching quality. A player undervalued because he was in an unfitting system still needs to be placed in a fitting one for his metrics to convert into points. The second is leadership stability. Brentford kept the same sporting department for years, which most lower-division clubs do not, and a model only works if it is not changed every time the team loses three matches. The third is a willingness to accept one losing year in exchange for three profitable ones, which a club with a patient owner can do but a club under promotion pressure cannot. The fourth is luck, and I will not pretend it does not exist.

At 60, I no longer believe in luck, only in numbers that have not yet had time to speak. But believing in numbers that have not yet spoken does not mean denying that luck exists. It means building a system in which luck is not the only condition for survival.

The empty stadiums of 2026 and what lost its mask

There was a natural experiment that football could never have planned, and it happened. When matches returned in 2026 in stadiums without crowds, aggregate research published afterwards showed home advantage fell markedly across major leagues.

For my work, that was one of the most valuable pieces of data in twenty years, not because it says something about crowds, but because it says something about what we had been calling character.

The empty stadiums of 2026 exposed something: much of what we called character was just noise.

A significant share of what was called home spirit, second-half comeback ability, and psychological strength in front of goal turned out to depend on a variable that appears in no player profile. When that variable was removed, metrics shifted enough to force a rewrite of several old assumptions.

This is not an observation about sports psychology. It is an observation about valuation. If part of a player's value comes from an environmental factor that can disappear, that part should not be priced as a fixed attribute. The transfer market still prices players believed to have leadership and fighting spirit based on performances delivered in conditions with crowds, without discounting the share of value that depends on those conditions.

Very few buyers think that way. That is why the gap remains.

What will happen in the next cycle

I do not write predictions for entertainment. Each judgment below comes with a metric I will track, so that it can be proven wrong.

First, the price of ball-carrying players in wide areas will keep rising, but not for attacking full-backs, rather for wide centre-backs and holding midfielders capable of passing through the first line. The metric to track is the number of first-line-breaking passes per 90 minutes.

Second, as the number of matches in a season continues to grow, the value of players aged 21 to 23 with strong physical profiles and managed minutes will rise faster than the value of players at their peak. The metric to track is the correlation between a player's minutes across three consecutive seasons and the number of days missed through injury.

Third, the market will keep paying high prices for goals and keep underpricing the ability to create goals from the third position onwards in a move sequence. That is the largest remaining gap, and it exists simply because goals are the only thing written into the match record.

And finally, something I will never stop repeating to everyone in this profession: if your model explains only the successes and not the failures, you do not have a model. You have a collection.

What I am still waiting to see

I have followed professional football for 44 years and reported on it for most of that time. I have watched models praised and then forgotten, players valued at 20 million pounds and then sold for two, and correct spreadsheets filed away in drawers because nobody had the patience to let them prove themselves.

What I am still waiting for is a club that applies the percentile valuation principle fully to the selling side, not just the buying side. Brentford sold at the right moment in most cases, but that was still an art practised by instinct. Nobody has published a model that decides the moment to sell systematically, based on the slope of a player's development curve rather than on an offer arriving on a particular day.

When that happens, the transfer market will have a new war, and this time it will not be about who buys cheaper, but about who knows precisely when an asset peaks in value. The question facing everyone in this profession is an uncomfortable one: if you knew exactly the day to sell a player, would you have the courage to sell, while the whole stadium is singing his name?