#8 Best Defensive Small Forward of all time - The ten best defenders in each position project

[trex_8063]

All of those stats are derived from box scores; right?

Which stats are you referring to? I somewhat get the feeling you're referring to data I presented in one of my above posts; if so, I'll summarize......

All-D Honors "Points" is an arbitrary weighting given to honors received. DPOY Shares is exactly what you think it is, drawn directly from bbref.

DWS and DBPM are box-derived figures (steals, blocks, rebs.....numbers curved toward the team's defensive and overall performance in each given season). You can search in here to get an idea how they're derived.

DRAPM is a pure plus/minus based metric to which box statistics don't figure in at all. DRAPM is the basis for the various Defensive Value Over Replacement figures I've presented (which I've occasionally referred to as DVOR) for most of the players covered. You mentioned DVORP (the figure which can be derived from the VORP numbers on bbref): that is something different. VORP is derived from BPM (which, as stated above, is a box-derived figure); so deducing the defensive component of that would still be box-derived. I intentionally left the "P" off the end of my acronym to distinguish it from VORP or DVORP; but perhaps it's still too close. Sorry for any confusion. But my DVOR is utilizing DRAPM for *most players [more on that below], with replacement level arbitrarily set at -0.75, as hinted at in header.

Here you are saying, but wait a minute trex, you've included players who pre-date RAPM data. You are correct (that I resorted to something different for earlier players is acknowledged in opening paragraph above, btw, along with a hint at potential disparity which is even in bold type-face).
To specify what data is used for what years.......

RAPM goes back as far as the 1996-97 season (which is pretty good: 22 seasons of data there). We then have regular season APM (also a pure on/off metric with scores very similar in proportion or standard deviation to RAPM) for '94-'96; however, it is not broken into offensive/defensive splits. I nonetheless used it, and incorporated some guidance (from Estimated Impact and/or BPM splits) to estimate the defensive split for '94-'96 APM data.

And then for all seasons prior to '94, I simply used the defensive split of Estimated Impact. EI is a box-derived figure which is curved toward the team's efficiency differential (probably somewhat similar to BPM, though per the author it apparently slightly out-performs BPM in predictive performance). A brief description of the metric can be read here.

Not apples to apples where these older players are concerned, so one must view it with that huge grain of salt, but I still felt it worthwhile to include (though you're free to disagree).

[Dr Positivity]

Vote Luol Deng

Very complete resume. DRAPM data and team results, longevity, physical tools/basketball IQ/intensity

[trex_8063]

For anyone paying attention to it, I just made adjustments to the DVOR data I've presented, pro-rating Havlicek's early seasons to 82-game length (to match those of all more recent players). It only created a tiny nudge to his numbers (didn't change his rank anywhere), but just thought I'd mention it.

[SinceGatlingWasARookie]

All of those stats are derived from box scores; right?

Which stats are you referring to? I somewhat get the feeling you're referring to data I presented in one of my above posts; if so, I'll summarize......


DRAPM is a pure plus/minus based metric to which box statistics don't figure in at all. DRAPM is the basis for the various Defensive Value Over Replacement figures I've presented (which I've occasionally referred to as DVOR) for most of the players covered).

Yes I was referring to your stats.

Thanks for the detailed answer.

Is "DVOR" your project? Googling DVOR wasn't producing results in the first few Google pages so I was guessing that DVOR was either a mostly unknown stat or an alternate spelling of DVORP.

I like Plus Minus derived stats despite the difficulty they have with getting rid of teammates contributions. I won't be analyzing the regressions to firm my own opinion of how successfully they eliminate teamate effects.

I am wary of defensive stats overrating teammate contributions, overrating team winning, overrating rebounding or overrating gambling for steals. On the other hand eye test is too subjective. My faded memories of Paul Pressey are subjective but i hope his mediocre rebounding doesn't effect his defensive stats much.

[trex_8063]

All of those stats are derived from box scores; right?

Which stats are you referring to? I somewhat get the feeling you're referring to data I presented in one of my above posts; if so, I'll summarize......


DRAPM is a pure plus/minus based metric to which box statistics don't figure in at all. DRAPM is the basis for the various Defensive Value Over Replacement figures I've presented (which I've occasionally referred to as DVOR) for most of the players covered).

Yes I was referring to your stats.

Thanks for the detailed answer.

Is "DVOR" your project? Googling DVOR wasn't producing results in the first few Google pages so I was guessing that DVOR was either a mostly unknown stat or an alternate spelling of DVORP.

Yeah, DVOR is something I just started compiling during the course of this project (somewhere during the SG's section), although I've seen people post similar information in very small sample comparisons. It just seemed like the natural progression from DRAPM. It makes the assumption that when our star or defensive stud comes out of the game, he's going to be replaced by a roughly replacement level player [not always, but on average]. That being the assumed case, it seemed worthwhile to scrutinize just how much defensive value ABOVE that replacement-level player he provides (per game, per season, per career, etc.....I provided several break-downs).

A league-avg player would naturally have a total RAPM of 0, so I arbitrarily went with -1.5 as definition of a replacement level player (who is presumably a little below average). But that's total RAPM; so I further assumed that the -1.5 overall score is coming by way of -0.75 on offense and -0.75 on defense. If you follow all that, that is how I arrived at -0.75 DRAPM as my arbitrary definition of a replacement level player. Although truthfully, it's probably more appropriate that replacement level DRAPM is NOT the same across all positions (e.g. maybe it's like -1.0 for PG's, but only -0.5 for PF/C's, or whatever, given the application of different positions are often more offense or defense oriented than others). That's gonna take a little more digging on my part to make those types of adjustments.

[trex_8063]

I'm jumping in for OP again to keep things moving. Looks like Marion came out of no where for this one:

Shawn Marion - 4 (kendogg, penbeast0, lebron3-14-3, LA Bird)
Luol Deng - 3 (iggymcfrack, trex_8063, Dr Positivity)
Kawhi Leonard - 2 (cecilthesheep, SinceGatlingWasARookie)

I'll get the next up in a moment.