Football-Data-Predictions ⚽🔍

Python script that shows statistics and predictions about different European soccer leagues using pandas and some AI techniques.

Run it 🚀

First, run git clone or dowload the project in any directory of your machine. The data files are extracted from https://www.football-data.co.uk/data.php and are updated automatically.

Before you start, you should install some dependencies via pip, in the case that you haven't installed them yet:

pip install pandas
pip install tkinter
pip install xlsxwriter
pip install sklearn

Once this is done, run the following command and follow the instructions of the terminal according to your wishes:

python main.py

In Linux or MacOS, you might have to type instead:

python3 main.py

Warning: This project has been fully developed in python 3.9, so its operation for previous versions is not guaranteed.

For developers 💻

Most of the models used are based on the same pandas dataframe. This dataframe is made up of a series of rows, each with a series of attributes (columns). These attributes can be divided into two types:

• Attributes of the current match: Used by the models to measure errors and check their effectiveness. They will be the attributes to predict. They are located in the first columns of the dataframe.
• Attributes before the game: Data available from the teams before the start of the game. Used to try to predict the attributes of the previous type described. For example, if a team has many goals in favor before a match, it is likely that in that same match (same row) the attribute indicating the number of goals scored by that team in that match is also high.

Thus, dataframes are made up of this combination of matches and attributes. For the best visualization of the dataframes, they are exported in an external folder called trainingData. The absolute data appears in the output.csv file, however, for all the models a conversion of the attributes is carried out, taking the average of them as a function of the number of games played.

In addition, the first matches of the dataframe have to be eliminated, because the averages still give uncertain values. As in any automatic learning process, the dataframe will be divided into test and training files, being able to select the percentage of matches dedicated to the test.

The Algorithms 🧠

Random Forest

A number N of decision trees is generated. These trees are formed by a series of leaves or rules, and in their final leaf, they will launch a prediction. The final prediction will consist of the arithmetic mean of the predictions made by the N trees.

Multilayer Preceptron

It is one of the simplest neural network models, made up of an input layer, an output layer, and N hidden layers or intermediate layers. Each layer will have a certain number of neurons and connections, each of which has a specific weight.

Next updates (To do list) 🔜

• Improve the flow of elections through the interface
• More artificial intelligence algorithms
• More options to print the statistics and sort the classifications by attributes such as goals, corners ...
• Allow users to select the parameters of certain AI models
• Allow users to select what they want to predict, and not only predict goals (predict shots, corners...)
• Who knows, maybe a graphical interface?

Want to collaborate? 🙋🏻

Feel free to improve and optimize the existing code. To contribute to the project, read the previous points carefully and do the next steps with the project:

1. Fork it
2. Create your feature branch (git checkout -b feature/fooBar)
3. Commit your changes (git commit -am 'Add some fooBar')
4. Push to the branch (git push origin feature/fooBar)
5. Create a new Pull Request

Need help ❓

Feel free to contact the developer if you have any questions or suggestions about the project or how you can help with it.