Assessment of the Market Value of Football Player in the European

League

Yao Liu

Shanghai Normal University, school of mathematics and science, Shanghai, 200000, China

Keywords: Market Value, Multi Linear Regression, Random Forest.

Abstract: In recent years, more and more money has been spent on the football transfer market. To utilize funds

efficiently, assessment of the market value is indispensable. After collecting the performance of the top 5

European Leagues (Premier League, Ligue 1, Bundesliga, Serie A and La Liga) football players in 2022-2023

season from Kaggle and Market Value from the transfer-market, this paper uses Multi Linear regression and

Random Forest model to give an insight into the correlation between Market Value and the performance

variables. This paper not only compares the Multi Linear Regression model and Random Forest model to

determine which one is better but investigates the important variables of the Market Value. Based on the

results of this paper, it can be inferred that minutes that player played is the most important variables. In

summary, with assessing the performance of the football player, the article gains insights into market value

of the football player and help the football club utilize the funds more reasonably.

1 INTRODUCTION

With the popularity of the transfer market, the costs

in the transferring market also have seen a significant

increasing. In the past ten years, the general transfer

fees for football player have inflated over 110%

compared to their market value (Poli, 2023).

However, according to Gerhards et al., the high

market value player plays an indispensable role for a

team to win the championship (Gerhards, 2017).

Hence, more and more professional clubs and

scholars started constructing their own assessment

system, aiming to choose the suitable player in a

reasonable price.

The assessment system of football players is

complex with many indexes have taken into

consideration. Therefore, Metelski investigated from

the single football league, Ekstraklasa and used

descriptive statistics and statistical tests. He pointed

out the importance of age. The top football clubs tend

to choose the talented young players as they can

easily adapt to new style and tactics (Metelski 2021).

Rodríguez selected three European Football Leagues

(Premier League, Bundesliga and Serie A) for

comparison. He explored that different Leagues

usually focus on different factors. The Premier

https://orcid.org/0009-0000-1860-4695

League pays more attention to experience,

performance in the previous season and assistances

matches. In Germany, people care about goals and

assists. However, in Italy, age and experience and

substitutions are the top priority (Serna, 2021). In

order to simplify and quantify the performance of the

football player, He et al. performed the Least

Absolute Shrinkage and Selection Operator (LASSO)

regression to determine whether the market value of

the player was overrated or underrated. The better a

plyer performed, the higher market value they would

get (He et al., 2015). Majewski utilized ordinary least

square (OLS) and feasible generalized least square

(FGLS) methods to figure out the economic potential

of players. He also emphasized the importance of

econometric modelling in assessing player value

(Majewski, 2021). By using the Chi-square test,

Mario et al. identified six factors (nationality,

chronological age, laterality, playing position,

international player condition and sports training

condition) in the top five European Leagues which

would influence the market value (Sánchez et al.,

2022).

Lots of researchers were devoted to studying the

influencing factors of market values of the football

player and forecasting the market value. Müller et al.

226

Liu, Y.

Assessment of the Market Value of Football Player in the European League.

DOI: 10.5220/0013822600004708

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd International Conference on Innovations in Applied Mathematics, Physics, and Astronomy (IAMPA 2025), pages 226-232

ISBN: 978-989-758-774-0

built a multilevel regression model and exploited the

data-driven approach to identify several factors to

estimate the market value. The model provides

accurate prediction for the low-to- medium- priced

player. The study demonstrated the data-driven

estimation of market value has wider application

(Müller et al., 2017). AL-ASADI et al. applied

supervised machine learning to analyze different

variables that reflect the skills and characteristics of

football players. With combined Train and Test Split,

the model contributed to the evaluation of the market

value of players (Al-Asadi et al., 2022). Sun et al.

focused on prediction of Football Players’ Value by

presenting Decision Tree Regression (DTR), Random

Forest Regression (RFR) and Rhizostoma

Optimization Algorithm (ROA). The research

modelled the relationship between the performance

and market value of players across different European

Football Leagues. By employing a robust statistical

methodology, the study improved the accuracy of

market value predictions (Sun et al., 2024). Patnaik et

al. explored the influence of goals and assists to the

market value, by comparing three different models

(Crowd-based Estimation, Multilevel Regression and

Option based). By integrating advanced statistical

methods, the study provided a robust model for

predicting the transfer market value of football

players (Patnaik et al., 2019).

In light of the diverse perspectives and

methodologies employed in previous research, this

research conducted a study of the correlation between

the performance and the market value of the player.

By comparing Multi Linear Regression and Random

Forest Regression, this paper seeks to investigate the

key factors influencing a player’s market value.

2 METHODOLOGY

2.1 Data Source

The dataset used in this study is fetched from the

Kaggle website. The dataset includes 2,451 players

and 124 indicators measuring various aspects of

player performance in the Big Five European Leagues

(Premier League, Ligue 1, Bundesliga, Serie A and

La Liga) throughout the 2022–2023 season, which

was collected by Vinco from the ‘Football Reference’.

Furthermore, the current value on the market of each

player was extracted from Transfer-market, a widely

recognized platform for football valuation.

2.2 Variable Selection

The original dataset consists of over 100 variables,

encompassing technical skills, playing time, passing

accuracy, defensive actions, and more. However, a

substantial portion of these variables contained

missing values, which could adversely affect model

performance and interpretability. To enhance both the

efficiency and effectiveness of modelling process,

this paper applied correlation analysis to identify the

most informative predictors. As a result, this paper

selected the top 12 most relevant variables based on

their statistical significance.

2.3 Method Introduction

This article is going to use the multi linear regression

and Random Forest Regression to analyse the

relationship of the player and their market value. By

applying the log- transformation, this paper will

compare the significance of the two models and the

accuracy of the results. Eventually, it will enable the

selection of the optimized processing of models.

Multiple Linear Regression is a linear regression

model used to reveal the relationship between a

dependent variable and two or more independent

variables. The general form of the multiple linear

regression model is:

𝑦=𝛽



+𝛽



𝑥



+𝛽



𝑥



+⋯+𝛽



𝑥



+𝜀 (1)

In this paper, y represents the market value of the

player and 𝑥



, 𝑥



are independent variables, each

contributing to the linear relationship with 𝑦. The last

term of this formula represents the error term or the

residual, capturing the unobserved factors that affect

𝑦 but are not accounted for by the independent

variables in the model.

Random Forest Regression is a machine learning

algorithm that uses decision trees to predict

continuous outcomes. Through this application, the

article will uncover insights into various player

performance.

3 RESLUTS AND DISCUSSION

3.1 Descriptive Analysis

Before analysing the model, this paper conducts the

descriptive analysis of the dataset.

Assessment of the Market Value of Football Player in the European League

227

Figure 1: Distribution of Market Value (Picture credit: Original)

Table 1: Top 12 most relevant variables

Variable Coefficient

Goals 0.452

Starts 0.328

Min 0.326

X90s 0.326

MP 0.288

Rec 0.215

PasShoCmp 0.197

PasLonCmp. 0.193

PasShoAtt 0.192

TouAttPen 0.190

CPA 0.184

G.SoT 0.182

The bar chart in Figure 1 presents the distribution of

player market values in euros. It shows that the

majority of players have market values under 10

million. As market values increases, the number of

players has significantly decreased. However, a small

number of players have exceptionally high market

values, creating a long tail stretching towards the right

(up to and beyond 100 million).

IAMPA 2025 - The International Conference on Innovations in Applied Mathematics, Physics, and Astronomy

228

3.2 Correlation Results

The output of the correlation shows in the Table 1.

Therefore, in the following study, there will be 12

variables (Goals, Starts, Min, 90s, MP, Rec,

PasShoCmp, PasLonCmp%, PasShoAtt, TouAttPen,

CPA, G/Sot) and one dependent variable (Market

Value). The specific description of this dataset is

shown in Table 2.

3.3 Multi Linear Regression

The paper started to establish the Multi Linear model.

Based on the results of the model analysis in Table 3,

several variables showed statistically significant

relationships with the dependent variable, which

would be interpreted as follows:

It is indicated in Table 3 that Goals and CPA have

a strong positive impact, with each additional goal

and each carry into the 18-yard box are associated

with an estimated increase of over 3 million units and

almost 2 million units respectively. While the

PasShoCmp also have a significant impact, with each

additional completed pass between 5 and 15 yards can

increase approximately 661290 units in the Market

Value. By contrast, the G.SoT has a large negative

coefficient, over 5 million units. The MP also has an

insignificant impact, each additional match played

lead to 685300 decreasing for the Market Value.

From the above descriptive analysis, the

distribution of the Market Value is heavily right-

skewed. This uneven distribution justifies the use of

log transformation:

𝑦=log 1+𝑀𝑎𝑟𝑘𝑒𝑡 𝑉𝑎𝑙𝑢𝑒 (2)

In the regression model, this transformation helps to

stabilize variance and normalize the data. Hence, a

log-transformed was applied to the dependent

variable (Market Value). To assess the accuracy of

the transform, the article compared the residuals

between the log-transformed and the untransformed

model.

The study would analysis the plots in Figure 2 and

Figure 3:

Figure 2: Residuals of untransformed (Picture credit: Original)

Assessment of the Market Value of Football Player in the European League

229

Figure 3： Residuals of Log-transformation (Picture credit: Original)

Residuals vs Fitted: The log-transformation made the

residuals more random which effectively mitigates

heteroscedasticity.

Normal Q-Q plot: Although the log-

transformation has deviation at the tails, most of the

standardized residuals align closer to the diagonal.

But the untransformed indicated non-normality of

both tails.

Scale-Location: The log-transformation has more

random scattering of residual points - but still a

cluster in the middle compared with the

untransformed model.

Residual vs Leverage: The log-transformation has

fewer high-leverage points, while the untransformed

has some high-leverage points with large residuals.

Hence, the log-transformation model significantly

improves the accuracy of the evaluation.

Table 2: List of Variables

Logogram Meaning

Goals

𝑥



Goals scored or allowed

Starts

𝑥



Matches started

Min

𝑥



Minutes played

X90s

𝑥



Minutes played divided by 90

𝑥



Matches playe

Rec

𝑥



Number of times a player successfully received a pass

PasShoCmp

𝑥



Passes completed (Passes between 5 and 15 yards)

PasLonCmp.

𝑥



Pass completion percentage (Passes longer than 30 yards)

PasShoAtt

𝑥



Passes attempted (Passes between 5 and 15 yards)

TouAttPen

𝑥



Touches in attacking penalty area

CPA

𝑥



Carries into the 18-yard box

G.SoT

𝑥



Goals per shot on target (Does not include penalty kicks)

Market Value

𝑌

Market Value from Transfermarket

IAMPA 2025 - The International Conference on Innovations in Applied Mathematics, Physics, and Astronomy

230

Table 3: Multi Linear Model analysis results

Coefficient SE t

Constant -3117942 913976 -3.411 0.000657***

Goals 3201562 176591 18.13 < 0.001***

Starts -587654 312116 -1.883 0.059846

Min -76819 108045 -0.711 0.47716

X90s 8621678 9713056 0.888 0.374823

MP -685300 104864 -6.535 < 0.001***

Rec 66373 34393 1.93 0.053744

PasShoCmp 661290 180829 3.657 0.000261***

PasLonCm

. 40551 12903 3.143 0.001694**

PasShoAtt -342529 169739 -2.018 0.043703*

TouAttPen 422413 162880 2.593 0.00956**

CPA 1897253 435146 4.36 < 0.001***

G.SoT -5317837 1217966 -4.366 < 0.001***

0.3162

ust R² 0.3129

Dependent variable: Market Value

<0.05 **

<0.01***

<0.001

Table 4: Regression Model Evaluation Results

MAE RMSE R

Multi linear regression

1.365600 2.706930 0.097

Random Forest 1.327004 2.685046 0.111

3.4 Model Comparison

After dealing with the Multi Linear Regression model,

the study started building the Random Forest Model.

The dataset was split into 70% training and 30%

testing to evaluate the model.

The article would compare the Multi Linear

Regression model and the Random Forest model to

determine which model is more accurate. In order to

choose the best model, this paper compared Mean

Absolute Error (MAE), Root Mean Absolute Error

(RMSE) and R-squared(R

), where the lower MAE

and RMAE values indicate better predictive accuracy

and the higher R

value indicates the higher

interpretation of the variables.

It can be drawn from Table 4 that Random Forest

outperforms the Multiple Linear Regression model

across all three metrics, though the differences are

relatively small. The Random Forest model has lower

MAE and RMSE values and higher R

value. Hence,

this paper will choose the Random Forest model to

evaluate the football player market value.

3.5 Model Results

After choosing the best model, the analysis of the

Random Forest model is reckoned, including the

accuracy of the training and testing set and variables

importance.

Figure 4: Training vs Testing Set (Picture credit:

Original)

In Figure 4, the bar chart compares the classification

accuracy of the random forest model on the training

and testing datasets. It can be noticed from Figure 4

that the training accuracy is 0.997, while the test

accuracy is only 0.521. The training accuracy is

nearly perfect, which indicates that the model has

learned the training data extremely well. However,

the testing accuracy drops significantly to around

52%, indicating a substantial decline in performance

when the model is applied to unseen, new data.

Assessment of the Market Value of Football Player in the European League

231

Figure 5: Variable Importance (Picture credit: Original)

The figure 5 used Mean Decrease in Gini, which

measures how much each variable contributes to

reducing the impurity in the decision trees that make

up the model. It is indicated in figure 5 that among all

predictors, Min (minutes played) shows the highest

importance score, exceeding a value of 110. This

suggests that playing time is the most influential

factor in the model’s decision-making process. The

other indispensable variables are the TouAttPen,

PasShoAtt and PasShoCmp with over 100 mean

decreases in Gini. These variables play a critical role

in reducing impurity across decision trees. By

contrast, the Goals and G.SoT appear at the bottom of

the chart, with values below 40.

4 CONCLUSION

Overall, the regression model evaluation results

indicate the Random Forest Regression has better

performance. Although the Random Forest model

slightly outperformed the linear regression model in

terms of predictive accuracy, it also exhibited signs of

overfitting, as evidenced by a high training accuracy

(0.997) and a relatively low testing accuracy (0.521).

By analysing the importance of the variables of the

Random Forest model, the paper identified key

performance metrics that significantly influence

market value. The Min (Minutes played),

TouAttPen(Touches in attacking penalty area) and

PasShoAtt(Passes attempted between 5 and 15 yards)

have a strong positive impact on market value, while

Goals (Goals scored or allowed) and G.SoT (Goals

per Shot on Target) show significant negative effects.

Therefore, the football club may prioritize the key

player who consistently receive substantial playing

time and they can focus on the attack player as they

touch more in attacking penalty area compared with

other player. However, one thing that cannot be

ignored is that the dataset used in this study only

contains one season which means it is difficult to

apply to other time series. Further research could be

implemented in the future, aiming to use different

methods to analyze several seasons to improve the

applicability.

REFERENCES

Al-Asadi, M. A., Tasdemır, S., 2022. Predict the value of

football players using FIFA video game data and

machine learning techniques. IEEE access, 10, 22631-

22645.

Gerhards, J., Mutz, M., 2017. Who wins the championship?

Market value and team composition as predictors of

success in the top European football leagues. European

Societies, 19(3), 223-242.

He, M., Cachucho, R., Knobbe, A. J., 2015. Football

Player's Performance and Market Value. MLSA, 87-95.

Majewski, S., 2021. Football players' brand as a factor in

performance rights valuation. Journal of Physical

Education and Sport, 21(4), 1751-1760.

Metelski, A., 2021. Factors affecting the value of football

players in the transfer market. Journal of Physical

Education and Sport, 21, 1150-1155.

Müller, O., Simons, A., Weinmann, M., 2017. Beyond

crowd judgments: Data-driven estimation of market

value in association football. European Journal of

Operational Research, 263(2), 611-624.

Patnaik, D., Praharaj, H., Prakash, K., Samdani, K., 2019.

A study of Prediction models for football player

valuations by quantifying statistical and economic

attributes for the global transfer market. In 2019 IEEE

International Conference on System, Computation,

Automation and Networking (ICSCAN), 1-7.

Poli, R., Besson, R., Ravenel, L., 2023. Inflation in the

football players’ transfer market. CIES Football

Observatory Monthly Report, 82.

Sánchez, M., Orgaz, B., Ramirez Campillo, R., Nakamura,

F. Y., Luis Pereira, J. M., Carretero, M., Sánchez

Sánchez, J., 2022. Factors associated to the market

value of professional soccer players. CIES.

Serna Rodríguez, M., 2021. Factor analysis of the market

value of high-performance players for three major

European association football leagues. Managing Sport

and Leisure, 26(6), 484-507.

Sun, Y., Gu, K., 2024. Prediction of Football Players’ Value

in the Transfer Market of Well-known European

Leagues based on FIFA 19 and Real-world Data.

International ARAB journal of information technology,

21(4), 723-740.

IAMPA 2025 - The International Conference on Innovations in Applied Mathematics, Physics, and Astronomy

232