THIS PROJECT IS FOR EDUCATIONAL PURPOSES ONLY. IT WAS PERFORMED FOR COURSE IE256(Statistics for Industrial Engineers) AT BOĞAZIÇI UNIVERSITY.
AIM OF THE PROJECT:
This project is about understanding the behaviour of different betting companies and Premiere League with the use of available information from different sources (odds from different betting companies, team status and etc.).
METHOD:
Based on the data of matches and odds,Premiere League teams' matches are filtered in date,hour,season and other necessary categories. In different steps of the project, home,away and total goals were studied and questioned if there were any known distribution for these variables. Also hypothesis testing was applied in order to evaluate the claims in process. To visualize data and support the claims, histograms and plots are highly used. In final part of the project, the relation between total goals and selected predictors is studied and multiple linear regression is performed. In order to perform predictions, the data is split into test(2018-2019 season) and train data(seasons starting from 2010 until 2018). After modeling and predicting, the success of the model is tested by betting for spesific odd type for various matches.
CONCLUSION:
With the model built with train data, 73% of the total variance in the number of total goals in the dataset is explained. The predictions on the 2018-2019 season were highly successful. Based on our model obtained with train data, we decided a bet rule on predicted total goals.According to that bet rule, we bet on the predicted values of 3.0+ and 2.3- . In that region we obtained a total profit of 105.6097 for 1 $ for every match we bet on. Average profit overtime for season 2018-2019 is analyzed cumulatively. We did this by dividing total profit with number of matches we bet till that time, and observed the change in average profit over time. We observed that average profit increased over time.