The aim of this study was to predict final score difference between home and away NBA teams for a regular season. Separate linear regression models depending on total over/under were fit to data before the all-star break, and checked for adequacy. Consistently important effects include
the closing line, rest days, non-college players, and age. Finally models were checked with data after the all-star break. We assigned a monetary value to each out-of-sample bet on a game and calculated the expected profit for each model.