APBRmetrics

Mike G

Among players who were ages 24-27 last year and 25-28 this year, from the top 100 minutes played (675+) in 2009-10, there are 85 who also have also played at least 385 min. this season.

From b-r.com, I got their 'advanced' stats -- and also inserted some of my own -- to see which statistics seem to be most and least stable from one year to the next.
http://bkref.com/tiny/zSS8I

'delta' is the average absolute change of 85 players, not weighted by minutes or any other quantity.
%change is delta/2010

m indicates one of my stats; in some cases just per36 with some adjustment for starters' minutes, etc.

DSMok1 · Posted: Tue Jan 25, 2011 10:23 am Post subject:

For each, you need to weight by the attempts/opportunities. For shooting, in particular, this would be important. That way we get a better picture of how much 3Pt% actually changes (the players with only 1 opportunity would have no impact). Actually, come to think of it, it would be best to weight by sqrt(opportunities), right?
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1

Mike G · Posted: Tue Jan 25, 2011 11:27 am Post subject:

This survey isn't actually about 3pt%. You may ignore that one, or see it as a novelty outlier, since not everyone even has a 3pt%. Even with it's intrinsic volatility, it's not the least stable of these.

What surprises me is that players in this group seem to be in decline, since last year. Are players now aging more quickly? Earlier studies indicated no population-wide decline before age 28.

The group includes LeBron, Dwight, Amare, Bosh, Melo, Iguodala, CP3, Roy, Nene, GWallace, DLee, DWilliams, Josh Smith, Aldridge, MGasol, Varejao, Landry, Bogut, Millsap, Mo Williams, Felton, Granger, Noah, Deng, Okafor, Nelson, KevMartin, TParker, AJefferson, . ..

Rebounding might be expected to drop off -- OffReb, as it turns out -- but should be made up for by better TS%, Ast%, TO%. Are veterans suddenly being crowded and outcompeted by an unusual number of good, younger players?
_________________
`
36% of all statistics are wrong

DSMok1 · Posted: Tue Jan 25, 2011 11:47 am Post subject:

kjb · Joined: 03 Jan 2005 Posts: 865 Location: Washington, DC

TrueLAFan over at RealGM did a study showing (if I recollect correctly) that total minutes played was a better predictor of decline than age. The study got flushed when they redid their message boards, so I can't link to it or refer to it.

Crow · Joined: 20 Jan 2009 Posts: 816

Mike G,

Would you be willing to report the averages and change in your scoring rate in complete form except for the counterpart scoring adjustment to see how each piece is changing relative to the other part? I am wondering if your counterpart scoring adjustment generates advantage on average for this age band or not (or the ones above and below) and wonder how far the player's own unadjusted scoring term is falling. That might be one useful addition to your extensive breakout.

DSMok1,

Would you have any interest in applying a somewhat counterpart scoring adjustment to your ASPM? It would be easier to implement than other choices.

Both of you,

I think the main differences between your metrics are the counterpart scoring adjustment and rebounding. Anything else that is big that I am missing?

Would you have any interest in some side by side complete and component comparisons of hypothetical players or a few real players?

I'd be particularly interested in how the metrics differ on the scoring element. While they are different in detail, they also somewhat similar in style. The complex, dynamic construction of the scoring element standouts in comparison to other main linear metrics where each stat is a separate piece. How similar and why different are questions that come to mind that I think might be useful to highlight a bit further.

bbstats · Joined: 25 Apr 2010 Posts: 45

I think you should standardize the individual scores before finding their standard deviation, since TS% and eFG% aren't really comparable to other stats (TS% and eFG% are likely skewed in comparison due to the fact that they're not actual percentages).
_________________
http://thebasketballdistribution.blogspot.com

http://twitter.com/bbstats

Mike G · Posted: Thu Jan 27, 2011 6:56 am Post subject:

Eh, is this what you really mean to say?
Whether a TS% has changed from .549 to .516
... or from 54.9 to 51.6
... it's still a drop of 6.1% (.061*.549) from the first figure to the 2nd.

These are all simple ratios, and without converting to standard deviations, they retain their actual magnitude.

Yes, we fully expect shooting% to fluctuate less (as a fraction of total) than some other stats. They're included mostly as a frame of reference.

Meanwhile, we can see how close the %change is for ORb% and DRb%, for TRb% and Reb, for MPG vs everything else.
_________________
`
36% of all statistics are wrong

Mike G · Posted: Thu Jan 27, 2011 9:23 am Post subject:

Up in the opening post, PER changes relatively little from last year to this, compared to WS/48 and eW/484.

But when we subtract 'replacement level' from PER, it takes the lead in volatility.
In this thread --
viewtopic.php?t=2153
-- at least 2 years ago, JohnH had RL at about PER = 10.8 (between 10.5 and 11.5, depending on position).

Here are the averages and %change for these summary metrics:

bbstats · Joined: 25 Apr 2010 Posts: 45

Ah, ignore my prior post.
_________________
http://thebasketballdistribution.blogspot.com

http://twitter.com/bbstats