Sample Size

[NYK Dolemite]

Hey there. Kind of a newb to hoops metrics. I've done quite a bit with them relating to baseball over the past ten years. It seems pretty intuitive to me, by now, as to when a sample size for either a team or individual becomes large enough to mine for data and have that data yield predictive value. I was just wondering if there is any rule of thumb amongst basketball stat heads as to when a player's improvement or decline (or even that of a team) stops being an outlier and starts being recognized as the new level at which a team or player has begun to perform. 40 games? Full season -- I understand that it's contextual, but it seems to me as if 40 games within a season is a number at which (with reasonable minutes) that any sort of deviation from a player's norm can cease to be looked at as fluctuations due to luck.

Sorry this question is so broad -- just really trying to get a basic handle on when something really means what it appears to mean, in a basketball sense.

Appreciate any guidance.