Prediction of Future NBA Games' Point Difference: A Statistical Modeling Approach

As the NBA (National Basketball Association) is becoming more and more popular, people are trying to predict score differences of each game for many reasons, including for gambling. There are a lot of factors that can affect the game result. The aim of this article is to establish a convincing but simple model to predict the point difference for each game using previous game results, as well as deriving team ability and separate home advantages based on data from five seasons. By analyzing the games from 2012 to 2016, the model makes reasonable predictions by analyzing some crucial factors, such as team ability and home advantage, even as compared to the bookie's point spread. Although there are still some factors that are not concerned by the data, such as injuries, fouls and trades, the prediction of this model is still convincing. This paper applies the "least square fit" model to create a best-fit model. By comparing the standard error of the results, the best linear regression model is created and team strengths for each year is derived, providing good predictions.