I’ve touched briefly on Dean Oliver’s Four Factors. It’s an interesting concept, that four fairly simple statistics can be good indicators of how a team is playing. They can be tracked for both offense and defense (so really eight factors) to give a sense of how a team is playing on both ends of the court. As a brief refresher, or for those who are not familiar, the four factors are as follows:
Effective Field Goal Percentage: A measure of shooting percentage that takes into account the added values of three-pointers. A high EFG% signals that a team is making more of their shots, translating to more points.
Free Throw Attempt Rate: The ratio of free throws made to field goals attempted. A higher FTA means that a team is getting to the line more often and converting on those opportunities, which, like a high EFG, means more points are being scored.
Offensive Rebound Percentage: The percentage of rebounds that a team gets on their own missed shots. A higher OREB% means that a team is getting more possessions and, thus, more opportunities to score.
Turnover Percentage: Simply put, this is the ratio of turnovers a team has to the number of possessions that team has. Fewer turnovers means more scoring opportunities.
These numbers can offer a snapshot of how a team is playing, and you can see the interesting ways in which these factors interact.
This raises some interesting questions. Historically, how good of indicators have the Four Factors been of team success (i.e. winning)? How good of predictors can they be for the future? And how do those projections change over the course of a season?
Let’s start with the first question. I began by performing a regression analysis of the eight factors (four on offense and four on defense) versus a team’s net rating. I chose net rating because it is a measure of the difference between points scored and points allowed by a team per 100 possessions. Looking at the five seasons prior to the current one (2012-2016), I calculated an equation with an R-squared value of 0.9821, indicating a very high correlation.
I then performed a second regression analysis of net rating versus wins and calculated an R-squared of 0.9420, also a very strong correlation. These two numbers seem to tell us that the four factors have historically been a strong indicator of overall team success, just as Dean Oliver predicted.
Ok, great. That was step one. Still with me? Good.
From there, I applied these equations to the current NBA season in an attempt to predict the number of wins each team would finish with. Being so early in the season, there was a lot of variability in the numbers, as one might expect. The Cavaliers, for example, are projected to win just 28 games at the moment. Spoiler alert: that isn’t going to happen.
So I came up with an idea, one that will answer both the second and third questions I posed above. I have created a program to track each team’s cumulative four factors for both offense and defense over the course of the season. Using that program, I am able to project both a team’s net rating and season win total after each game of the season. My hope is that I will be able to a) track each team’s projection over the course of this season and b) see how well these projections match each team’s actual performance.
For example, here is the chart for the Boston Celtics.
You can see that things are already starting to even out, as my model was predicting them as a 30 win team based on their net-rating through 2 games but is now projecting them as a 62 win team after their 10-game win streak.
Another interesting team is the Orlando Magic.
Their hot start had them projected at an unfathomable 76 wins through five games. But they’ve cooled off, and just as their net-rating has plummeted, so, too, has their wins prediction.
I plan to check back in on this project once-per-month to see how things are shaping up and, by the end of the season, have a better understanding of how well this model stacks up to what actually happened. Then we can do a better analysis of what needs to change in the model moving forward.
If you would like to see the rest of the teams, you can find them in this google drive folder:
https://drive.google.com/drive/folders/1kjXHcabeOfkOfuvPWZCUGppea_w13FoO?usp=sharing
Or on my personal blog:
technicalsports.wordpress.com/2017/11/09/using-four-factors-to-predict-success/
