Ιδρυματικό Αποθετήριο
Πολυτεχνείο Κρήτης
EN  |  EL

# Forecasting the outcome of Greek football games using mathematical models and power rankings

#### Paliatsa Dimitra

Πλήρης Εγγραφή

 URI: http://purl.tuc.gr/dl/dias/24944DCC-2DA6-46B7-AB8D-AABE1BE69A1E Έτος 2014 Τύπος Διπλωματική Εργασία Άδεια Χρήσης Βιβλιογραφική Αναφορά Δήμητρα Παλιάτσα, "Forecasting the outcome of Greek football games using mathematical models and power rankings", Διπλωματική Εργασία, Σχολή Ηλεκτρονικών Μηχανικών και Μηχανικών Υπολογιστών, Πολυτεχνείο Κρήτης, Χανιά, Ελλάς, 2014 https://doi.org/10.26233/heallink.tuc.22812 Εμφανίζεται στις Συλλογές Διπλωματικές Εργασίες στην Κοινότητα Σχολή Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών

## Περίληψη

The subject of this thesis is the prediction of the results of football matches and the estimation of the final points gathered by the teams at the end of the football year. In particular, we are interested in the prediction of the outcome of the Greek Superleague games for the season 2007-2008 to 2013-2014. For this purpose it was necessary to record the relevant data for all teams after each game (goal difference, current points, number of wins, number of draws, number of defeats, etc.). Then, using the data gathered and utilizing various techniques we try to predict the final outcome of the matches. Our first approach falls into the sport rating systems category, where we rank all the teams based on their power scores (power ranking procedure). Our second approach is based on clustering and is implemented using the k-means algorithm. In the third approach we use two independent Poisson distributions for describing the goals scored by the home and the away team, respectively. Moreover, we use two naive methods for extracting baseline results: in the first we assume that the home team is always the winner and in the second we assume that the last year's result will be repeated for all teams that compete in two consecutive years. Finally, we use the moving average, the weighted moving average and the exponentially weighted moving average to estimate the final points that each Superleague team will gather at the end of the football year. The results of our study are promising, demonstrating the usefulness of the proposed models and making the further study of this problem interesting for future improvements.