# Is there a standard (advanced?) stat in basketball for measuring player “consistency”?

Is there an accepted stat that analysts use to measure how consistent a basketball player is?

Take points per game (or points per 48 min): two players with the same average scoring rates could have very different game distributions. Of course standard deviation comes right to mind (after you standardize it to account for high vs low scorers), and it makes sense when I try it.

But...

• Is another measure even more informative? I've looked at some data and am intrigued by other possible stats, but with only small datasets (and somewhat limited personal impressions of specific player [in]consistency in the NBA this year), the patterns I'm seeing could just be red herrings. :)
• Even if you use standard deviation, would any adjustments be needed? I'm not sure for example if low scorers would need to be handled differently, or if that all comes out in the wash once you standardize the data. And do you weight right skewed data differently, so it hurts your consistency rating less when you score more (vs less) than your average?
• What scale is most intuitive to use for the "consistency" measure? If you use standardized SD, for example, would it be %, where 100% is the player who always scores 12 every night? (what a machine!) Is there a scaling that would make comparisons between two realistic players' consistencies more meaningful than 25% vs 35%?

So I wonder if consistency is a "thing of interest" in sports analytics, and if so how pro analysts handle questions such as those.

Thanks in advance for any insights!

• Per 36 Minutes normalizes production. Chris "Birdman" Andersen typically had respectable Per 36 numbers, but never averaged 36 minutes per game (ostensibly due to lack of stamina). This stat, however, shows things like "12 fouls Per 36" because the player administered two fouls in six minutes of play. – user16112 Jan 28 at 13:55
• @user16112 Cool thank you! Makes sense that per 36 would be a key step to the consistency calculation. I take it per 36 is preferred because per 48 is less of a realistic game, though either way it's normalized to enable better comparisons. – Chad Parmet Jan 29 at 14:09