Predicting football match results of the Bundesliga. I called it DieMaschine (TheMachine).
This repository contains the code for predicting for the popular football result prediction game kicktipp (https://www.kicktipp.com). The model has proven to perform at super-human abilities by achieving more points than any human in a kicktipp round with more than 20 humans.
We use an evolutionary algorithm to determine the optimal parameters for a domain-specific algorithm determining winning, draw and losing probabilities of different matches. The evolutionary algorithm determines the optimal parameters by being "trained" on past Bundesliga seasons.
It contains a node.js backend which loads Bundesliga match from OpenLigaDB (https://www.openligadb.de/). The frontend shows the predicted results and contains code for learning algorithms.
This is one of my earliest programming projects (developed between around 2015 and 2019). I am proud of this project as it is . Therefore, it should be obvious that the code quality is low and the methods utilized would not be my first choice today...