APBRmetrics

Harold Almonte · Joined: 04 Aug 2006 Posts: 616

I think we're trying to discover two different things at a time with one only metric:

The "home bookeeping privileges-in stats subject to interpretations without any league revisions" as a part of the "home advantage", and..
The way a team play (regarding passing-assists) when being at home environment, or away. Which could be, aside bookeeping, a matter of some kind of different chemistry with environment. Very rare in this "home advantaged" sport, but this could be the case for SAC, NYK, etc.. providing that the schedules are not playing a role here.

Then one thing is to look at each arenas (historical home/opponent-s differential of assists, totals or %) and another thing is to look at each team (home/away assist% performance)

**HoopStudies** · Posted: Tue Sep 09, 2008 7:31 pm Post subject:

Harold Almonte · Joined: 04 Aug 2006 Posts: 616

Just that there is a noise produced by team h/a performance that simple formula-s can't take off. The Mike chart gives us more information about bookeepers's judgements (even with noise), because includes (each arena's) treatment to opponents numbers (respect to the league). Then, we can see which of them see more (DEN, CLE), or less (SAC, PHO), assists than league for everybody, which of them see more for home and less for opponents (ALL MINUS ONE), or viceversa (NJN).

But, the only way to know that bkeeping bias exactlly, is a pbp replay revision by more than one people of course, after a previous agreement about the rules.

It could come one day, who knows, when the league attempts to bookeep def. atempts stats, that would need to make revisions of doubtfulness before making boxscores official.

**HoopStudies** · Posted: Wed Sep 10, 2008 7:56 am Post subject:

I should also note that the ASFs I presented are essentially uncorrelated to playing style. I don't know what teams had new scorers in 2008, but the biggest changes in ASF are for New Orleans and Atlanta. And, hmm, New Orleans almost definitely had a new scorer because they were in OKC the year before. I might guess that Atlanta had a new scorer. But teams with the most roster turnover (and probably different style, like Boston and Minnesota) probably did not have new scorers as they were very consistent in ASF.
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

This seems a pretty interesting quick study. Good observation Ben. I was also surprised by how much these bias' varied from team to team, but were consistent from year to year. I noticed initially a few people mentioned individuals they felt had benefited from generous "assists" and I thought it might be worthwhile to look into.

To check this, I wanted to see how MPG, PER and APG impacted this bias. I found home splits from nba.com for the 2008 season and adjusted the them based on the "PF_ast" figures posted by Dean. In order to weigh factors of 5 and .2 similarly, I applied the following formula for my analysis (Ast_h - Ast_r)/Max(Ast_h,Ast_r). Performing regression analysis, I found that this value for players varies according to the following coefficients:

Mike G · Posted: Thu Sep 11, 2008 8:31 am Post subject:

I've often wondered about John Stockton's anomalous 1989, when his assist rate was interrupted from it's steady rise. In 1988, he had become a starter, and Utah's (road) assist rate jumped to the next level.
Factor = (1+AAst%/HAst%)/2 :

**HoopStudies** · Posted: Thu Sep 11, 2008 9:04 am Post subject:

Harold Almonte · Joined: 04 Aug 2006 Posts: 616

**HoopStudies** · Posted: Thu Sep 11, 2008 10:54 am Post subject:

Harold Almonte · Joined: 04 Aug 2006 Posts: 616

OK, my fault. The formula adds the opponents's assists, I was all the road thinking as it was substracting them, like a kind of differential.

It appears that having a lot of teams that give up more assists than they give, even at home (and bkeep-biased against the opponents), and that give less assists at home than on road; that, home and more assists are not correlated, although they should be (from the first factor).

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

Harold Almonte · Joined: 04 Aug 2006 Posts: 616

Good points.
I think that things should be put on easilly (binarial) to bookeepers: assist or no assist. But the conditional towards a consensed rule and (to be further clarified by somebody) should be one, and only one of these: zero move, just one move, or two, to avoid a defender (and the definition of what is an individual shot creation move, of course).

The weight of assists (between 0 and 1 full assist) is a metricians's job (and decision). The "assistfigure" figure, was already invented for this kind of question.

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

I understand that this could be somewhat confusing to track. I was thinking along the lines of how 82games.com tracks close shots, tips, and jumpers and box scores track fade aways and hook shots.

I definately agree that a binary statistics are needed, at least for now. I intended to say that there could be 2 types of assists, like in hockey and a count of passes, which may already exist.

And, certainly there are many ways to approximate the value of the average assist in the current form. I just think that if we can better measure what we are looking at, the variations of our estimates will lower and the accuracy of any models will improve, not to mention the decreases in the occurances of bias' (which would actually have more to do with consistency in measure). For example, adjusted plus-minus measures can approach almost exactly what an individual's point value,above a certain level, is as data and permuations of lineups approach infinity, but, a full season often produces poor measurements due to lack of credibility (or a high coefficient variances). I imagine that over a smaller period this statistic could be nearly useless (such as the NBA Finals or a measure of adjusted plus minus turnovers or something). However, if you improve the accuracy of your measurements, I think you can better estimate true value and better assess value a shorter sample.

That said, I don't really know how feasible or helpful this would be. It is difficult to track statistics to test and there are probably better ways for the many people like me who limited time in their schedule.