1. "Bottom-Feeders"
Well, not quite bottom-feeders per se, but the first thing I looked at was the correlation between Game Score (simple linear weighting of box score inputs) and Offensive Box Score Expected Value (OBEV) which uses the expected value of all box offensive box score values. In lay terms, these are two distillations of the classic box score, the first will reward more volume scoring, the second will reward more efficient offensive "results" (including turnovers).
A higher correlation coefficient between opposing defenses and these composite metrics suggests that a player performs better as the defense becomes easier. This doesn't make someone a bottom-feeder necessarily -- his performance could be quite strong against good defenses to begin with -- but it does represent a relationship between a player's box score stats and the quality of the opposing D. A low coefficient means the player is "flat" or defensively agnostic, and produces similarly against either good or bad defenses.
- Regular Season correlation; Game Score and Opposing Defenses
Garnett 99-08 0.04
Duncan 99-08 0.05
Shaq 96-05 0.05
Dirk 01-11 0.06
Olajuwon 86-96 0.13
K Malone 88-98 0.13
Robinson 90-98 0.14
LeBron 08-14 0.14
Jordan 87-98 0.15
Kobe 01-10 0.16
Miller 90-00 0.19
Some of that could be the defensive component of the box score that Game Score includes that was not removed for this study. What happens looking at the same regular season data from the vantage point of OBEV?
- Regular Season correlation; OBEV and Opposing Defenses
Dirk 01-11 0.03
LeBron 08-14 0.10
Shaq 96-05 0.10
Miller 90-00 0.12
Garnett 99-08 0.13
K Malone 88-98 0.14
Duncan 99-08 0.15
Olajuwon 86-96 0.17
Jordan 87-98 0.17
Robinson 90-98 0.18
Kobe 01-10 0.23
The players most likely to improve against bad defenses: Kobe Bryant. Robinson, Jordan and Karl Malone show movement on both lists. Dirk looks very flat -- it doesn't matter what the defense is against him. (We'll look at actual level of production in a moment.) What happens in the PS?
- Post Season correlation; Game Score and Opposing Defenses
Duncan 99-08 -0.01
Garnett 99-08 0.01
Shaq 96-05 0.03
Olajuwon 86-96 0.07
Dirk 01-11 0.09
Miller 90-00 0.14
K Malone 88-98 0.16
LeBron 08-14 0.19
Robinson 90-98 0.24
Jordan 87-98 0.25
Kobe 01-10 0.27
- Post Season correlation; OBEV and Opposing Defenses
Shaq 96-05 0.01
Duncan 99-08 0.06
Dirk 01-11 0.09
Garnett 99-08 0.11
Miller 90-00 0.16
K Malone 88-98 0.17
Olajuwon 86-96 0.17
Jordan 87-98 0.20
LeBron 08-14 0.23
Robinson 90-98 0.24
Kobe 01-10 0.30
The numbers are more pronounced in the PS, where Kobe (surprisingly) shows a more severe trend than even David Robinson (saddled with the reputation of falling off against good defenses). We'll see what exactly this looks like in a moment when we compare the actual numbers. Not far behind, LeBron and Jordan also show a moderate trend. Duncan and Garnett are relatively flat, and Shaq and Dirk essentially don't really seem to care who is on the other side of the court. CAUTION: Before running with a narrative based on this data, let's look at the actual results to gain a more refined perspective of what happens against good and bad defenses.
2. Regular Season: Good vs. Bad Defenses
For this and the remaining sections I used a 107/103 split for good/bad defenses. This is to try and normalize the data as much as possible across different environment, but it also leaves players with smallish samples at times because they either played really good or really easy defenses. That will be noted when applicable.
How did these same 12 guys fare against 107/103s?
You can see how the core elements of the data reflect the correlations from section 1. Kobe, Robinson, Jordan and Karl Malone have fairly large discrepancies between their "good"defense and "bad" defense metrics. While Olajuwon looks completely flat here, note he only played 32 games (!) against sub-103 defenses in these 11 years. Compare this to Garnett, who played more games against good defenses in this period (like Shaq) and a total of 231 games against sub-103 defenses.
As usual, metrics aren't kind to Kobe. He has the lowest TS% in the RS among this group against sub-103's, the second-lowest GmSc and the lowest OBEV. David Robinson -- the inspiration of this study -- does drop off, but he does not look like an outlier in this regard. Finally, it's poetic to see Duncan and Garnett with almost identical pts/36 and TS% numbers against the sub-103's, although Garnett is above Duncan in GmSc and OBEV. The largest drop off in assists from good to bad defense belongs to...Michael Jordan.
3. Post Season: Good vs. Bad Defenses
What happens in the playoffs?
Right off the bat, note that most of these players no longer play bad defenses in the PS. This is perhaps one of the oldest adages in basketball, and this case, it's very, very true. It is rare to encounter bad defenses in the PS, and even rarer to encounter them in the critical 3rd and 4th rounds of the PS where teams are better. Thus, emphasis in this section will be on the sub-103 performance alone.
Dirk looks fantastic: 21/36 and 57% TS, with 16 GmSc and 4.3 OBEV. The best and worst of the group against sub-103s:
- Best vs sub-103 defenses, PS
OBEV: LeBron +4.7
GmSc: Jordan 19.5
Pts/36: Jordan 27.0
TS%: Miller 58.9%
- Worst vs sub-103 defenses, PS
OBEV: Garnett +1.6
GmSc: Jordan 13.9
Pts/36: Robinson 18.2
TS%: Malone 50.7%
All sorts of common beliefs being supported there. First, we see evidence for Miller upholding his impressive postseason reputation as well as his performance against quality defenses. LeBron looks amazing. Garnett has an offensive dip in the PS, Robinson no longer looks very good in the box score, and Karl Malone's impressive scoring efficiency disappears. It should be noted though that a good chunk of this is from FT shooting. In these 36 games, Malone's FT% goes from ~ 74% in the RS to 66.5%. At 7.4 FTA per 36 (down from 9.1 in the RS). 74% FT shooting would put his TS% at 52% and pts/36 at 22.3.
However, look what happens when you compare these numbers to the other players in the group:
Duncan and Garnett have nearly identical volume against good defenses, Duncan trumps him with a 2.4% TS% advantage that, along with a better foul-draw, gains him 1.3 points of value in OBEV. Kobe and Jordan are both around 53% against elite defenses -- but note Kobe played in such a hard defensive environment that his average sub-103 defense is 2 full points better than Jordan's. Nonetheless, Jordan maintains a huge volume advantage. Robinson and Olajuwon...the thing to note there is that neither played many games against hard defenses!
4. Regular Season: Good vs. Bad Defenses
What changed in the PS? The table below reflects the differences between the RS and PS performances against plus-107's and sub-103's. Note that for plus-107's, Shaq, Miller, Kobe, Garnett and Duncan all play under 20 PS games, and for sub-103's Hakeem and Robinson play under 20 PS games.
Karl Malone! Malone, consistent with his well-established overall performance drop in the PS, has a large and comparable drop off against both quality of defenses in the PS. (Whether that's because of flaws in Malone, Utah, or both cannot be determined just from this data.) Compare that to Garnett, who actually sees a relatively small PS drop off against good defenses.
On the opposite end of the spectrum, again...Reggie Miller. The scoring takes a large bump and the efficiency. Olajuwon's improvement is against plus-107's mainly. Shaq has a small drop off, although notice he plays harder sub-103's. David Robinson completely falls off against the easy defenses. Kobe, in 76 games sees his scoring go down with a comparable hold in efficiency, but like Shaq, he played harder sub-103's in the PS.
5. Conclusion
This data reinforces other slices of the box score that show players performance declining in the PS. Among these 11 stars, there's an average playoff drop of about 0.6 pts/36 and 1% in TS. GmSc goes down by 0.8 pts (interestingly, OBEV is almost the same). Of course, defenses are slightly harder in the PS as well.
As far as "good"/"bad" defensive split go, it's also clear that it is, in general, harder to perform well against better defenses. In the RS sample, the group had the following averages:
- vs. 107+: 25 pts/36 | 59.% TS | 20.7 GmSC | +5.0 OBEV
vs. 103-: 23.2 pts/36 | 55.5% TS | 17.9 GmSc | +3.4 OBEV
The average defense faced was over 8.5 points better in DRtg in the sub-103 group. All told, this information is simply a report of the box score, and the key box score metrics. It must be noted that this does not entirely map to offensive goodness for a number of reasons -- team context, role and creation are completely ignored.
