NBA Win Prediction Model

Over the off-season, I’ve developed a win-loss projection model based on point differential for the upcoming season. This model will be used to analyse all 30 teams and project their final records for the season, but first, I’ll give a quick explanation as to how it works in this article. The model basically works under the premise that players are the most influential thing when it comes to scoring and that team values implemented through coaching and training (positioning, switching etc.) are the most influential thing when it comes to other aspects of the game. As such, it uses players minutes, two-pointers made, two-pointers attempted, three-pointers made, three-pointers attempted, free throws made and free throws attempted to calculate how many points per game the team will achieve. Once these values were inputted into the spreadsheet, a multiplier for the minutes was calculated.
E.g. Total minutes for the 15 man San Antonio roster was 315.2, but there are only 240 minutes in a game (5 players * 12 minute quarters * 4 quarters), so San Antonio’s minutes multiplier will be 315.2/240 (0.7614 to four decimal places).
The next multipliers I needed to put into place were regarding Field Goal and Free Throw attempts. Using the Spurs as an example again, they attempted 83.6 Field Goals a game last season, but even after applying the minutes multiplier the model still projected them to take 91.1 shots a game. This model relies on the assumption of a relatively similar Field Goal attempts number from year to year (historically the only alarming increases come with radical coaching change), so a multiplier must be used to lessen the amount of FGA. In the Spurs case, it was 83.6/91.1 (0.9180 to four decimal places). The same process occurred for Free Throws Attempted. This resulted in an offensive points per game, which subtracted from defensive points allowed gave the Point Differential per game.

Next, to find how many wins each team is projected to get, the relationship between Point Differential and Wins each year was graphed. This produced a linear function for both the Eastern and Western Conference and a formula for finding wins from point differential was also produced. These are shown in the graphs below.
Graph 1

Graph 2

Using this formula, the wins for each team were calculated, however, these wins totaled just 1190, when there are 1230 wins per season. As such, each teams wins were multiplied by 1.0338 to give a grand total of 1230 wins.
Clearly with this model evenly distributing wins in such a manner will cause some middling teams to record more wins than one may project (e.g. Dallas 45 wins) and will cause some better teams to project less wins than expected (e.g. Houston 51). Unfortunately, that is the nature of the formula and there will always be out-liars, however, it still serves as a relatively accurate model for win projection.
There you have it, enjoy the articles that will follow!

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s