Categories

# An Equation to Predict NBA Playoff Probabilities

Based on a previous paper I wrote that used machine learning to determine the most relevant factors for teams making the NBA playoffs, I did some further analysis in an attempt to come up with an equation that outputs the probability of an NBA team making the playoffs in a given season.

From the aforementioned paper, one concludes that the two most important factors in determining whether a team makes the playoffs or not is its opponent assists per game and opponent two-point shots made per game. Based on that, I came up with the following equation: $\boxed{P(playoffs) = 0.49 \left[ \frac{1}{1 + \exp\left(-7.6683 +0.2489 o2P \right) } \right] + 0.51 \left[ \frac{1}{1 + \exp\left(-9.1835 +0.4211 oAST \right) } \right]}$

A plot of this equation is as follows: A contour plot is perhaps more illuminating: One can see from this contour plot that teams have the highest probabilities of making the playoffs when their opponent 2-point shots and opponent assists are both around 20. In general, we also see that while a team can allow more opponent 2-point shots, having a low number of opponent assists per game is evidently the most important factor.

Using this equation, I was able to classify 71% of playoff teams correctly from the last 16 years of NBA data. Even though the playoff classifier developed in the paper mentioned above is more accurate in general, those methods are non-parametric, so, it is difficult to obtain an equation. To get an equation as we have done here, can be extremely useful for modelling purposes and understanding the nature of probabilities in deciding whether a certain team will make the playoffs in a given season. (Also: note that we are using the convention of using 0.50 as the threshold probability, so a probability output of >0.5, is classified as a team making the playoffs.) ## By Dr. Ikjyot Singh Kohli

Sikh, Theoretical and Mathematical Physicist, main research in the structure and dynamics of Einstein's field equations.