Turnover worthy throws???? WTF is that and when did that become a thing? Some people seem pretty desperate when they need to use stats of what COULD have happened to make their point.
[
Boldface emphasis is mine]
To be fair, much of frequentist statistics (the Fisher-Neyman-Pearson stuff most people get in basic statistics classes) does that. If you think about what a
p-value is, it involves integrating (or summing) over a region of the sample space that is at least as unfavorable to the null hypothesis as the actual observations were. So any inference coming from
p-values depends on observations that theoretically
could have been made, but weren't.
And that's not even the worst problem
p-values have. Because the inferential cutoffs are fixed, any statistical test involving a
p-value can be fooled by a big-enough sample. Yes, despite the intuitive (and correct!) idea that a larger sample should give us better inferences, that just isn't how it works with
p-values. And when you consider that most medical research approved and published in your lifetime used
p-values as statistical evidence that a given medical treatment was effective (and some of it involved misuse of the statistical tools, including various forms of "
p-hacking"), it's pretty horrifying.
ObSmith: I think it makes some sense to talk about lucky and unlucky outcomes. I'd want to see exactly how they're defining turnover-worthy plays, and I'd want to see historical analysis of how likely players with a lot of turnover-worthy plays are to have their turnover rates increase, but
if turnover-worthy-play percentage, according to whatever definition is used, turns out to be a decent leading indicator of turnover rates, I think it makes all the sense in the world to talk about it.
