Page 1 of 1

P values in your basketball stats

Posted: Fri Mar 22, 2019 1:45 pm
by Snedecor

With the new number of ‘The American Statisticia’ treating about the use of pvalues I would like to know if is frequent or sometimes and when do you use pvalues, with high numbers of repeations like in NBA is possible to use pvalues in some cases.

What do you think?

Article link:


Re: P values in your basketball stats

Posted: Sat Mar 23, 2019 3:37 pm
by xkonk
I use p values in my job because we're expected to but would much rather not. I'm not on top of basketball research as I used to be but I don't think p values are nearly as prominent there. And why would they? What hypothesis are you testing? If someone's RPM is above 0? I don't think that's an important or interesting question.

I do think the lack of description of variability is an issue in basketball. If someone runs an analysis about someone's rim protection ability or impact on pick and roll defense, I might see what percentile they are or the points per possession. But how tightly packed are players? If they reduce opponent points per possession by .1, is that a lot? What's the range? If it's a small range, how many players are in, say, .08 to .12? If it's a big range, maybe .02 to .18? A player at 95th percentile at rim protection is obviously near the top of the league but how different are they from 90th percentile, or 75th? I rarely see anything like that.

Re: P values in your basketball stats

Posted: Mon Mar 25, 2019 9:26 pm
by Snedecor
Maybe this is not usual, but for me make sense do some hypothesis contrasts with a group of players and another, or maybe can work to hired some players and compare to others, is pure statistics and maybe are not worthy. NBA have a lot of data per player and per team, tracking, and doing some analysis maybe is posible.