Post#278 » by runtmc » Mon May 6, 2024 12:27 am
Since theres a lot of talk about what the math says without anyone actually doing it, I figured doing the math might help a bit:
A 95% CI on Ant's playoff FG% is (43.7%, 54.3%), which includes his career RS FG%, so as of now, it is a bit too early to say there's something there beyond luck/streakiness. That said, its relatively close. He's currently sitting at 335 FGA for his playoffs career, for a 1-tailed test he needs to get to about ~505FGA, or a little over 8 more games at the same rate for it to be a significant result. You can do a similar analysis for his TS%/3p% etc.
OTOH, if you do a two-proportion z-test with a two tailed hypothesis that p1!=p2, you end up with p = ~.12, and need somewhere around 540 shots, or about 10 more games.
Essentially, there is some good evidence so far that Ant does better in the playoffs, but its not yet significant, and you'd need somewhere between 8-10 games at the same level for it to become a significant result (meaning there's only a 5% or less chance its due to normal random noise/luck).
Edit: also just to point out, games played is really only a very rough guide for this, as usage will heavily influence things. A guy taking 20 shots per game will converge much faster than someone taking 5 or 10 shots a game, and Ant does have relatively high usage in the playoffs, which is why it might seem odd a relatively small sample size of ~25 games would be enough for a 4.4% FG% difference to show significance. For Gobert for example, you'd need closer to 70 games for a 4.4% gap to show significance.