Predicting Soccer Player Market Values to Identify Undervalued Talents

This project leverages machine learning to estimate the market value of soccer players based on their attributes such as skills, physical traits, and club information. The goal is to help football clubs make informed and cost-effective transfer decisions by identifying players who are potentially undervalued.

---

Project Objectives

• Build predictive models to estimate player market values.
• Identify undervalued players who offer high value for their cost.
• Provide actionable insights into which attributes influence player valuation.

---

Dataset

• Size: ~18,000 players
• Attributes: 40+ features including:
  • Age, Height, Weight
  • Overall Rating, Potential
  • Preferred Foot, Team, Contract
  • Market Value (€)

---

Data Preprocessing

Key steps in data preparation:

• Missing Values:

  • Numerical: Filled with column mean.
  • Categorical: Filled with mode.

• Standardization:

  • Height and weight converted to metric units.

• Normalization:

  • Applied to numerical features for select models.

• Encoding & Parsing:

  • Encoded categorical features.
  • Parsed contract data to extract start/end dates.

---

Models Used

1. Linear Regression

• A baseline model to understand linear dependencies.
• Easy to interpret the influence of features.

2. Random Forest Regressor

• Captures complex, non-linear relationships.
• Provides improved predictive accuracy.
• Hyperparameters tuned using GridSearchCV.

---

Evaluation Metrics

• Root Mean Square Error (RMSE) – Measures average prediction error.
• R² Score – Indicates how well the model explains variance in the data.

Results

Model	RMSE	R² Score
Linear Regression	€9,986,093.81	Moderate
Random Forest	€4,126,979.45	High

---

Key Findings

• Skill-related attributes (e.g., Overall Rating, Potential) are the strongest predictors of market value.
• Physical attributes (e.g., Height, Weight) had less impact.
• Random Forest significantly outperformed Linear Regression, highlighting the importance of non-linear modeling.

---

Lessons Learned

• Preprocessing is critical for accurate predictions.
• Model selection impacts both accuracy and interpretability.
• Regular updates are needed due to the dynamic nature of the football market.

---

Future Work

• Integrate time-series data to monitor market trends.
• Experiment with advanced models such as neural networks.
• Incorporate additional factors like injuries, morale, and club economics.

---

Team

• Mustafa Sallat
• Mohammed Khaled
• Amro Altawashi
• Abdulmalik Altwaijry

---