Data mining v oblasti kurzového sázení 3. anglické fotbalové ligy

Thesis "Data Mining in the Field of English Football League Third Division's Betting Odds" deals with data mining referring to acquiring knowledge from data. The main objective of this work is to develop data models for prediction of match results and to compare these predictions with a chosen strategy of betting. The selected betting strategy is based on betting single bets with odds belonging to chosen intervals, which generate a profit. These odds intervals were discovered by analyzing 2006-2009 football matches in a created simulator. On the basis of these odds ranges data models were constructed. Each data model contains a hypothesis which is generated by SD4ft procedure of LispMiner based on all football matches played in seasons 2001-2008. Developed data models are tested afterwards using 2006-2009 football matches data. Results show that all derived data models are profitable in all four seasons under consideration. More than half of them successfully predicted 2009 matches as well. The analysis showed that betting agencies offer mostly odds which make it almost impossible to be profitable while betting on matches according to their odds. In spite of this fact I identified some odds intervals with which you can success while betting single bets on home-team, draw or visitor-team with odds falling within these intervals. Association rules with reasonable confidence and support can generate high profitability. It is important to realize that there are no data models which guarantee a certain profit. Most of developed data models are not applicable in the real world, some of them can actually generate a loss. Nevertheless there are data models to be found that could generate a profit in the real world.