APBRmetrics

Hi everybody,

I have a relatively simple idea how the "Live and die by the Three" saying could be tested - at least for a person that is used to parsing stuff from Basketball-Reference.com
In my opinion, the basic mathematical idea behind "Live and die by the Three" is: 'teams that shoot a lot of threes have a higher variance in their game to game outcome and are thus more relying on chance."
This is a hypothesis that can be easily tested in a few steps (again, if you know better than me how to parse data). What the person would need to do is:
1. extract on these pages (or wherever you like) http://www.basketball-reference.com/tea ... 4/gamelog/ the ORtg game logs
2. calculate the variance or std
3. You need as well the mean ORtg of a team and the mean 3PAr. But if you are able to complete step 1, this should be easy.
4. For most values in this world, there is usually a correlation between std and mean (signal-to-noise-ratio). And there might be a correlation between ORtg and 3PAr.
So you would need to use noise instead of std. Noise is usually variance/mean^2 or variance/mean. It's a theoretical thing, so take whatever between variance or noise does not/barely correlate with your mean.
5. Look for correlation between noise/variance and 3PAr

I guess this would not be the end of the story, but it would be a way better way to show anything than 'Well, the Spurs shot a lot of threes las year!'. Just in case someone is interested to look into this

Cheers,
Hannes (@SportsTribution)

Ideally I think one would want to look at offensive rebound trends too because they affect the level of riskiness / success compared to 2 pt shooting and maybe turnover rates getting to those shot types too. But yeah it would be good to see analysis of team shot strategy along the lines you suggest and more.

Sure, especially offensive rebounds would be another good factor, as you could expect 'crashing the glass' to be of high variance.
Once the Variance/Noise part is figured out the rest becomes pretty much gravy