Authors: Athanasios Anastasiou, Giannis Salogiannis

Title: A linear regression on NBA game results to test fatigue. Is this factor possible to predict the outcome of a match?

Abstract

Basketball is one of the most popular sports in the world and the National Basketball Association (NBA) is the greatest league on planet. Many issues related to NBA basketball have been studied. Forecasting the outcomes of future matches has been being investigating extensively and different statistical models have been proposed.

The methods of statistics and especially linear regression find application in many fields of science, either as evidence of the relationship between factors in a phenomenon, or as a first attempt to predict an event, based on obvious factors that influence this phenomenon.

In particular, sports betting was and is an area of interest and application of statistical tools, as with the advent of the internet it has evolved into a global business with a turnover of billions. In addition, the methods developed in this area can be very easily adapted to models for various sectors of the stock market: Trading stocks, forex, derivatives and others.

In this study we will test these methods. Choosing the NBA basketball league, for which there is abundant information, both present and past, to feed our models, we will test through linear regression and various related methods and we will reveal how various factors, such as fatigue, location of the game, effectiveness of scoring and several others, affect the outcome of the final result. We try to prove that with the linear regression can produce a positive return on investment for the bettor. Besides, our research has an application in monitoring the betting market, which may be useful to bettors or betting companies.

The results of the research revealed the existence of significant relationships of all these factors to the outcome of the match. Also, we try to measure the relationship of each factor, both to significance and efficiency of prediction.