Re: [APBR_analysis] Defensive ratings
- McKibbin, Stuart wrote:
> Laid to rest? Okay. Here's a few theoretical problems I have with Ed'sHell, you didn't even mention the two biggest problems, IMO: I had to
> approach, which on first blush sounds reasonable.
estimate both the FGM and OR terms, both in a way which penalizes big men.
> 1. A portion of the offensive ratings are earned in transition, say...
> the first 5 or 6 seconds after a steal or defensive rebound. This
> favors the big men because it's the guards and SFs that do most of
> the damage offensively in those situations in oddman rushes.
> My proposal is that GETTING BACK is a team defense category and noAs you note above, shot clock time does not equal possession time. About 30%
> individual blame should be assessed. Everything that happens in the
> first 5 or 6 seconds after a steal or def reb (provided there isn't
> an interuption in play by timeout or something) should be taken out
> of the equation. With the 82 games database this could be done.
> As it is http://www.82games.com/clock1.htm shows that teams give up
> anywhere from 40-44% of offense in the first 10 seconds of a shot
> clock. This could be further refined by changing the shot clock
> segmentation (which would also eliminate such things as fouling
> immediately to get the ball back, and stupid loose ball fouls by
> attempted offensive rebounds that send the other team to the line
> easily) and subtracting out all off reb putbacks and putting them in
> a separate category. I would propose by removing those two things
> we've isolated half -court offense and defense---which is what we are
> really after, right?
of all missed shots are offensively rebounded, and therefore not subject to
the "getting back" defense (GBD). Looking through a sample of 143 games that
I have here, only 15% of all possessions end in under 10 seconds.
Another thing is that of the offense recorded with 10 seconds or less on
the shot clock, a lot of it comes on fouls at the end of games where one
team fouls quickly to stop the clock. This too is not subject to GBD. Three
percent of all possessions end in under 3 seconds -- I assume that these end
in a player being fouled or a set play coming off an inbound or putbacks.
That said, your points are well taken. I will see what I can do about
analyzing my data more closely following your advice above.
> 2. It doesn't take into account the variable offensive ratings of theAs far as I know, the difference here isn't all that great. Why don't I just
> positions. Centers having lower average Ortgs than SGs and PFs, so the
> centers as a group benefit again.
check that? <sound of ruffling paper and pencil scratching> Hmm. It turns
out that centers actually have *higher* offensive ratings, although the
difference is not significant:
> 3. It doesn't take into account the team's defensive scheme. ForA standout example of a lack of strength-of-opponent factor in my system.
> example, I'm told Boston did a lot of trapping of the PGs, so the
> opposing PGs PER is lower than other opposing positions. I don't
> think it's because Boston's PGs were exceptional defenders. Maybe
> they were, I just doubt it. Or more to the point: Reggie Miller shows
> up as superior to Ron Artest. That a 38-year old rates better than
> the alleged Defensive Player of the Year is curious. In any event, I
> wouldn't know how to account for it, so I'm just complaining without
> a concrete alternative.
Abdul-Rahim shows up on my top-20 list also, presumably for the same reason:
neither of these players are asked to guard tough assignments. DanR's system
is the only one, I think, that begins to address this problem.
> 4. These rating systems are unfair to good perimeter defenders on poorAn interesting idea.
> defensive teams.
>.... It would take critical mass of perimeter
> defenders (2, 3, 4?) to overcome their center's ineptitude.
> So the options might be that only centers can be judged across teams,I never look at my defensive stats in isolation. I usually mix them with a
> and all we can do for the others is judge them relative to their own
> team; or we give recognition for large relative decreases in pts per
> 100 defense to perimeter players.
heavy component of team defense, just to provide a safety net.