Who is the most statistically average player in the NBA?
Not the 250th-best player. Not the most mediocre player. I wanted the player who sits closest to the actual center of the NBA — physically, statistically, stylistically, and everything in between.
So naturally, I built an unnecessarily complicated system called B.A.S.I.C. — the Basketball Averageness Similarity Index Calculation.
We start with a horribly broken definition of “average,” then progressively make it worse by adding:
- Player size and age
- Playing time and role
- Per-possession production
- Efficiency
- Shot profile and play style
- Offensive and defensive impact
- Correlation adjustments
- Mean vs. median
- Thousands of randomized definitions of “average”
Until eventually, one player refuses to stop being normal.
And because apparently that wasn’t enough, I also used the results to construct the literally average NBA player: John Basketball.
All of this work to scientifically discover the role player.
Fantastic.
Fun modeling exercise. @SimpleStacker and I feel your pain.
Enlighten us—who is this player that fits those statistics? He’s surely not someone who gets much media attention.
#1594815
I just noticed "position index", though. I don't know if it makes sense to have that be on a continuous scale from 0 to 4.
Earlier iterations attempted to use the share of time spent at each position. This is probably highly correlated with the physique metrics, which wasn't addressed in the video.
I like the idea of share of time spent at each position. Why did he give up on that approach?
Sure, it might be correlated with physique, but from my skimming of the video he had some techniques to deal with heavily correlated variables.
I didn't catch his reason for switching. Maybe it was too noisy.
400 data points is a pretty small sample size for this kind of high dimensional analysis
I didn't catch the sample size. Power was my other thought. If those five categorical variables can be closely approximated by one, then collapsing them is pretty valuable.
LOL! Even his name is average (though an unusual spelling of an average name). I didn't watch the whole video but I'm guessing name wasn't used as a metric?
No, you should contact the creator and inform him of the oversight.
Nah, the name coming out average, despite not being used as a regressor, is an interesting bit of external validation.
One thing he didn't include that I would want to at least toss in the mix is salary.