Obviously I’m picking and choosing my samples. That’s what a sample is.

No, actually, a sample is a single data point. Picking and choosing your samples is cheating.

Look, you made a decent judgement on your samples by including only ones where he spent a decent amount of time on the court, and you should be commended for doing that instead of saying “Well, when he plays exactly 19 minutes, he shoots the lights out…” or “if you look at the games where he scores more than 10 points, he usually scores a lot”. Good job for being objective.

However, when a single data point completely changes the scientific meaning of the data the validity of the entire data set comes into question. Either:
1) that data point is an anomaly that should be thrown out
or
2) the data set is just too small to mean anything

I was giving Belinelli the benefit of the doubt by saying that it’s just too small of a sample size to be properly judged. If you throw that data point out, the numbers are pretty grim.

I even think that his last game should mean the most because he was actually playing in a game where he had an opportunity to effect the outcome. Maybe it means he steps up to the challenge. And if you look at the overall production (counting everything), he shot 39% from beyond the arc, which is pretty good and hard to fake.

But it’s still way too small of a sample size to mean anything. Let’s reevaluate after next year if he plays 1000 minutes and takes 400 shots. That would mean something.