# Data-Driven Scouting · Football Level 8: Expert

Football · Level 8: Expert · Insporta Book (https://book.insporta.app/football/level-8/data-driven-scouting/)

Data-Driven Scouting in Football (Level 8: Expert) on Insporta: 100 sourced facts behind 10 tests. Covers: the recruitment funnel; and more.

How clubs use data to find players: the recruitment funnel, similar-player searches, league strength, age curves, several seasons of data, radars, market values, data-led clubs, famous finds and the limits of the numbers.

## Test 1: The recruitment funnel

- Clubs start broad: data filters thousands of players down to a list worth watching. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A handful of scouts can watch a few hundred players a season; data can look at every player in dozens of leagues. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The long list comes out of the data screen and is narrowed by video and scouting into a short list. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A 'ball-playing centre-back who can defend a high line' asks for different numbers than a 'centre-back who wins headers'. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- How a player reacts to mistakes, talks to team-mates and works without the ball is best seen in the stadium. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A player who misses many matches gives the club less for the same fee and wages. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A good season can be a peak; several seasons show whether it is the level he really plays at. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Each step is more expensive in time than the one before, so it is applied to fewer players. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Every stage removes players, so the funnel narrows from thousands to one. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The long list is cut to a shortlist that scouts and analysts study on video and in person before a target is chosen. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))

## Test 2: Similar-player searches and role profiles

- It compares statistical profiles across a database and ranks players by how close they are to a reference player. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A player built into the team's way of playing is best replaced by one who does the same things. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A winger and a centre-back will never look alike in the numbers, so comparisons are made within a role. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A 'deep-lying playmaker' profile, for example, weights progressive passes and passes under pressure over aerial duels. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Similar numbers against weak opposition are not the same as similar ability, which is why league strength is adjusted. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Being in the 90th percentile for his position says more than a raw count, which varies with league and style. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Adaptation, language, personality and ambition sit outside the event data a similarity score is built on. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The ranking is a starting point; the video check decides whether the closest numbers mean the closest player. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Each filter keeps only part of the players left by the one before it. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The passing figures are close, 7.9 and 7.4, but the candidate wins three times as many aerial duels. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))

## Test 3: Adjusting for league strength

- The same finishing produces more goals against poorer defenders and goalkeepers, so raw totals overstate him. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- They give every club a rating on one scale, which lets analysts compare leagues and adjust players' numbers. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Where event data is full and consistent, a player's figures are reliable; thin coverage leaves more guesswork. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The bigger the step up in opposition, the more a player's output tends to fall. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- If strikers moving from one league to another usually lose a third of their goals, the next one probably will too. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Defenders who close faster and block better take away both the volume and the quality of his chances. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Those leagues develop and showcase talent at lower prices than the biggest leagues, and the data shows who stands out. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- UEFA's coefficients place Portugal well above Scotland and England at or near the top. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The adjustment borrows the experience of earlier transfers between the same two leagues. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Adjusting for the stronger opposition he would face lowers the goals he can be expected to score. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))

## Test 4: Age curves

- Output usually rises through the early twenties, levels out in the mid to late twenties and then declines. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A goalkeeper's reading of the game keeps improving into his thirties, while outfielders depend more on repeated running. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A club buying before the peak gets the player's best seasons and can still sell him for a large fee. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Past his peak, a player's output is more likely to fall than to repeat, so the last season overstates the next. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- It is built from many players' careers and shows the typical rise, peak and fall of output by age. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Two 18-year-olds with the same numbers can end up years apart, depending on growth, coaching and playing time. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Buying young and cheap, developing them and selling at 24 to 26 is a business model built on the age curve. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Roles built on pace peak earliest; defending on reading and positioning later; goalkeeping latest of all. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The shape is typical, not fixed; recruitment models use it to judge what a player of a given age is likely to do next. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Past the peak the index drops each season, so the later years of a long contract pay for less and less. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))

## Test 5: Seasons, minutes and availability

- Several seasons smooth out lucky and unlucky spells and show the level the player really sustains. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A player who starts every match for a demanding coach has passed a test that raw numbers cannot show. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A player who misses a third of each season gives two-thirds of what the club pays for. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Players develop and decline, so last season's numbers are a better guide to next season than those from three years ago. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A move from full-back to winger, or into a more attacking system, can transform the numbers without the player changing. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Dominating his own age group is common among academy players; doing it against adults is much rarer. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Late cameos come when games are open and opponents are tired, which can inflate a per-90 rate. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The more minutes and the more seasons, the less luck and context drive the numbers. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Cleaning and weighting the data comes before any judgement against other players in the role. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A drop from 2,800 minutes to 900 and back suggests an injury, a loan or a falling-out, each of which matters to a buyer. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))

## Test 6: Percentile radars

- Each spoke is a metric, and its length is his percentile rank against players in the comparison group. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Knutson's radars, first published on his blog and then at StatsBomb, became a standard way to show a player's profile. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Each slice is a metric, and how far it is filled shows the player's percentile rank. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The 90th percentile among League Two midfielders is a very different claim from the 90th among Premier League ones. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Change the peer group or swap a metric and the same player's shape changes, which is why the small print matters. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A percentile ranks him against peers; it says nothing directly about his success rate or total. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Placing strong metrics next to each other makes a bigger filled shape than spreading them out, though the numbers are the same. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The comparison group and the metrics must be fixed first, because every percentile depends on them. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The narrower the group, the more like-for-like each percentile becomes. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- He is in the 92nd percentile for passing and the 88th for tackles, but only the 20th for aerial duels. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))

## Test 7: Market-value models

- The values are crowd-sourced estimates discussed by the site's users and set by its editors, not official prices. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The site began in 2000 and grew into the most widely quoted source of player market values. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The CIES Football Observatory estimates values with a statistical model built on past transfer fees. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Published estimates move slowly and use less data than a club has, so clubs model fees themselves. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The selling club holds the player's registration for longer, so it can ask for more. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Buyers know they can wait and sign him for nothing, so the selling club's bargaining power shrinks. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A club in financial trouble sells cheaply; one under no pressure can hold out far above any published value. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The more years of registration the seller can offer, the higher the fee it can demand. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The model learns what drove past fees and applies that to the player the club wants. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- With the same form, the drop from 40 to 15 million follows the years left on his deal, from three to one. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))

## Test 8: Data-led clubs beyond Brentford and Liverpool

- Bloom, a Brighton supporter, made his fortune in betting and brought a data-led approach to the club. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Starlizard's modelling of football matches sits behind the analytical culture at Bloom's clubs. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Toulouse beat Nantes 5-1 in the final, a first major trophy under owners known for a model-driven approach. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Beane, who ran the Oakland Athletics on statistical principles, took an advisory role at the Dutch club in 2015. (Source: Club statements and match reports (BBC Sport, The Athletic))
- The deal, around £100 million plus add-ons, was the biggest fee a British club had paid at the time. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Buying undervalued players and selling them at full price is the model; the fee pays for the next round of finds. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Mac Allister came through at Argentinos Juniors and later won the World Cup with Argentina while a Brighton player. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Mac Allister signed in January 2019, Caicedo in February 2021 and Mitoma in August 2021. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Bloom became chairman in 2009, Benham bought Midtjylland in 2014 and RedBird bought Toulouse in 2020. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Mitoma signed from Kawasaki Frontale in 2021 and spent his first season on loan in Belgium. (Source: Club statements and match reports (BBC Sport, The Athletic))

## Test 9: Famous data-assisted signings

- At the University of Tsukuba he studied dribbling, filming himself with a head-mounted camera. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Without enough international caps for a permit, he spent 2021-22 on loan in Belgium and returned the next summer. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Independiente del Valle's academy has produced several players sold to Europe. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Mahrez cost Leicester a small fee from Le Havre in Ligue 2 and became the PFA Player of the Year in 2016. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Walsh's scouting brought in both players cheaply before the 2015-16 title, and he then left for Everton. (Source: Club statements and match reports (BBC Sport, The Athletic))
- When few clubs have spotted a player, the seller cannot start a bidding war, and the first buyer pays less. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The numbers that attracted the club came from a different context, and adapting to the new one is part of the risk. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- He arrived in 2021, spent part of that year on loan at Beerschot, broke into Brighton's team and left in 2023. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Vardy arrived from Fleetwood in 2012, Mahrez from Le Havre in 2014 and Kanté from Caen in 2015. (Source: Club statements and match reports (BBC Sport, The Athletic))
- Mitoma spent 2021-22 at Union Saint-Gilloise, the Belgian club also owned by Tony Bloom. (Source: Club statements and match reports (BBC Sport, The Athletic))

## Test 10: What data cannot tell a scout

- Effort without the ball, reactions to mistakes and conduct in training leave no trace in event data. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A player's mentality, motivation and how he will fit the group decide whether his numbers carry over. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A player who struggles off the pitch in a new place may not reproduce the numbers that got him signed. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A wing-back in a back three or a free roamer may put up numbers that the buyer's own system cannot recreate. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A scout can explain why the numbers look as they do: a role, an injury, a team-mate, or an opponent's tactics. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Analysts or scouts who like a player can pick the metrics that support him and ignore those that do not. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- A skinny 17-year-old may be well behind physically today and far ahead of his peers in five years. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- Event data logs passes directly, tracking data can show positioning, and leadership barely shows up anywhere. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- The checks move from what the data implies to the person, and the medical comes last, once a deal is close. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))
- How a player handles pressure is not logged as an event; the other options are standard event-data counts. (Source: Football analytics literature (Opta, StatsBomb, FBref definitions))

Sources: IFAB Laws of the Game / FIFA & UEFA technical reports / Opta, StatsBomb and FBref definitions / club and analyst publications / peer-reviewed research

Take the tests: https://insporta.app/try?utm_source=book&utm_medium=referral