For the statisticians

[KHD]

Here is a paper I came across on the arXiv where they've done machine learning, neural network analysis in order to determine what the key statistical features of playoff and championship teams are.

http://arxiv.org/abs/1604.05266

There are almost certainly a bunch of issues with these methods and of course they are all counting stat-driven. But here's some of their conclusions regardless:


For championship teams: offensively, the 3 most important predictors are 2p%, opponent personal fouls, and assists.

Interestingly FTA/FTM is not a strong predictor, but opponent fouls are. There's probably a subtlety in the analysis that I'm missing to interpret that one.

overall FGA and opponent FGA are not strong predictors of championship teams either (as close as i can get to "pace").

Anyways, the conclusion in the paper at least is that offensively you need to have an offence that generates good 2 point shots. Would like to see a model incorporating more detailed stats like points in the paint and fast break points.


From their conclusion:

Using an artificial neural network structure, we concluded that championship teams must be able to have very strong defensive characteristics, in particular, strong perimeter defense characteristics in combination with an effective half-court offense that generates high-percentage two-point shots. A key part of this offensive strategy must also be the ability to draw fouls. This analysis will hopefully dispel the rising notion that an offense geared towards shooting many three point shots is a sufficient and necessary condition for an NBA team to be successful in qualifying for the playoffs and winning a championship

[Fully]

Schwing!

[Raptor Jesus]

Mid Range Game confirmed GOAT

DD Finals MVP

[KHD]

Mid Range Game confirmed GOAT

DD Finals MVP

eh, the predictor stat is 2p fg percentage, not attempts. :P

[DogeLover1234]

Cool, thanks for sharing. I should read the whole thing before commenting but I wonder how many years of data they used. I would imagine the the rise in three point attempts over the past few years makes that 2 point % indicator less important than it was a decade ago, although efficient scoring is always needed.

[Puffer]

eh, the predictor stat is 2p fg percentage, not attempts. :P

HaHaHaHahaHaHa.

Love it.

[KHD]

Cool, thanks for sharing. I should read the whole thing before commenting but I wonder how many years of data they used. I would imagine the the rise in three point attempts over the past few years makes that 2 point % indicator less important than it was a decade ago, although efficient scoring is always needed.

should've said in my post.

they took the past 16 years.

your point is valid too. Like i said i'm sure there's a lot of issues with the methods as in any paper. But it's interesting to think about regardless