Vurdering af antal mål i fodboldkampe
Translated title
Assessing the number of goals in soccer matches
Author
Olesen, Rasmus
Term
2. term
Education
Publication year
2008
Pages
75
Abstract
Dette projekt undersøger, hvordan man kan forudsige, hvor mange mål der bliver scoret i en fodboldkamp, med henblik på at automatisere beregningen af sandsynligheder og fastsættelse af odds. Modellerne bruger historiske kampdata og statistiske værktøjer som empiriske sandsynligheder og Poisson-fordelinger (en model for antal hændelser), samt begreber som offensiv styrke og defensiv svaghed til at beskrive holdenes niveau. Tilgangene er optimeret, implementeret og sammenlignet ved hjælp af en logaritmisk scoringsfunktion, som vurderer, hvor gode de beregnede sandsynligheder er. Resultaterne er yderligere vurderet med hypotesetests, herunder Wilcoxons Signed-Rank Test (en ikke-parametrisk test til at sammenligne parvise resultater). Undersøgelsen viser, at det er svært at matche bookmakeres vurderinger fuldt ud med en automatiseret model. Alligevel præsterer de fremlagte metoder tæt på bookmakerens niveau og peger på, at automatiseret odds-sætning kan være mulig og lovende.
This project examines how to predict the number of goals in a football (soccer) match, aiming to automate probability estimation and odds setting. The models draw on historical match data and statistical tools such as empirical probabilities and Poisson distributions (a model for counts), along with concepts like offensive strength and defensive weakness to describe team performance. The approaches were optimized, implemented, and compared using a logarithmic scoring function, which evaluates the quality of probability forecasts. Results were assessed with hypothesis tests, including the Wilcoxon Signed-Rank Test (a non-parametric test for paired comparisons). The findings show that fully matching bookmaker assessments with an automated model is challenging. However, the methods presented perform close to bookmaker levels and indicate promising potential for automated odds setting.
[This abstract was generated with the help of AI]
Documents