PlaWorDet-WARP24 - FIFA Player Worth Determination Model

PlaWorDet-WARP24 is an AI/ML project designed to calculate the value of football players based on a range of metrics, including performance statistics, team chemistry, and positional dynamics. This model is unique in its ability to offer club-specific valuations by considering existing team dynamics, enabling more informed bidding decisions in both in-game and real-life FIFA auctions.

Objective

The primary goal of this project is to provide an accurate player worth estimation by incorporating the following factors:

1. Player Statistics: Detailed analysis of player performance metrics.
2. Chemistry with Team: Evaluation of the player’s fit into the current team setup.
3. Positional Analysis: Assessment of how well the player matches the team's positional requirements.

Features

1. Club-Specific Valuation: The model calculates a player’s worth relative to the specific club, considering the chemistry with existing players.
2. Bid Recommendation: Suggests optimal bidding strategies based on player value and team dynamics.
3. Interactive Interface: A user-friendly interface built using Streamlit for easy input and output of player valuations.
4. Scalability: The project is designed to work with large football datasets and can be scaled across different leagues and clubs.

Detailed Pipeline

1. Input: The user inputs new player statistics, preferred position, and club name.
2. Base Price Calculation: The player's base price is predicted by passing their statistics through an MLP model.
3. Preferred Position Prediction: The optimal position for the player is calculated using another MLP model.
4. Team Similarity Score: The chemistry score is derived by comparing the new player with existing team members.
5. Club-Specific Value: Mathematical transformations are applied to the base price and similarity score to generate a club-specific valuation.
6. Position Factor: A binary tree is created where nodes represent positions, and the distance between the preferred and required position is calculated. This is used to adjust the player’s worth.
7. Final Valuation: Club-specific value is adjusted based on the position factor to provide the final player worth.

Flowchart

Installation

1. Clone the repository:

   git clone https://github.com/Laksh-Mendpara/PlaWorDet-WARP24.git

2. Navigate to the project directory:

   cd PlaWorDet-WARP24

Usage

1. Install Requirements: Install the necessary dependencies using requirements.txt.

   pip install -r requirements.txt

2. Running the App: Start the Streamlit app by running:

   streamlit run Src/app.py

Files and Directories

├── model
│   ├── app2_f1.pth          # Saved model for the player worth prediction
│   ├── best_params.json      # Best model parameters
│   ├── label_encoder.pkl      # Label encoder for categorical data
│   ├── poscalerfin.pkl        # Position scaling model
│   └── pos_model_file.pth     # Position prediction model
├── plots
│   ├── accuracy_plot.png      # Accuracy plot
│   ├── loss_plot.png          # Loss plot
│   ├── pipeline.png           # Pipeline flowchart
│   └── plot_2_f1.png         # Additional plots
├── Src
│   ├── app.py                # Main app file for Streamlit
│   ├── price_predictor.py     # Final inference pipeline
│   ├── analysis.py           # Analyzes the dataset for insights
│   ├── analysis_similarity.py  # Examines price variations with similarity scores
│   ├── FIFA01_merge.py       # Merges datasets from FIFA 17 to FIFA 23
│   ├── position_tree.py       # Constructs a binary tree for positional representation
│   ├── PositionPredictor.py   # Predicts the best preferred position for a player
│   ├── Similarity_Score.py    # Calculates similarity scores between players
│   ├── preprocess_data.py      # Preprocesses input data for the model
│   ├── worthpredictionbasic.py  # Calculates the base price of a player
│   └── __pycache__            # Compiled Python files

File Descriptions

• app.py: The main application file for running the project using Streamlit, providing a user interface for input and displaying player valuations.

• analysis.py: Contains code for analyzing the dataset, focusing on extracting insights from player statistics.

• analysis_similarity.py: Implements functionality to examine how a player's worth varies with their similarity to existing team members.

• FIFA01_merge.py: Responsible for merging the datasets from FIFA 17 to FIFA 23 into a comprehensive dataset for analysis.

• position_tree.py: Constructs a binary tree structure representing player positions, where nodes contain positions with similar characteristics.

• PositionPredictor.py: Contains the logic for predicting the best position for a player based on their skills and team needs.

• price_predictor.py: The core inference pipeline that combines all previous steps to determine the final worth of a player.

• Similarity_Score.py: Calculates the similarity score between the new player and existing players on the team to inform the valuation process.

• preprocess_data.py: Responsible for preprocessing the raw input data to ensure quality and consistency before it is fed into the model.

• worthpredictionbasic.py: Contains the code to find the base price of a player based on their statistics using a multi-layer perceptron (MLP).

Future Scope

1. Enhanced Data Sources: Incorporating additional datasets, such as player injuries, transfer history, and external performance metrics, could improve the accuracy of valuations.
2. Real-Time Analytics: Developing a live feed that updates player statistics in real-time during games could enhance the model’s utility.
3. Expansion to Other Sports: Adapting the model for use in other sports could broaden its application and impact.
4. User Personalization: Implementing user-specific recommendations based on their bidding history or team strategies could make the model more effective.

License

This project is licensed under the MIT License. For more details, see the LICENSE file.