If you missed the last post, it was an overview of Expected Value (EV). And while that approach is not a novel concept — check out this similar method — from what I gather, incorporating a large defensive component is. Most of the defensive numbers used are from my stat-tracking. As a refresher, the defensive component of EV includes:
- FG% against (including blocks)
- Defensive Errors
- Forced Turnovers
- Shooting Fouls
- Defensive Rebounds
So which individual players fare the best in this metric? Below are the top defensive players in EV from the 2010 playoffs, with defensive usage included as a reference for the size of a player’s role (minimum 30 defensive plays “used”):
Dwight Howard, not surprisingly, had the best playoffs on the defensive end according to this. It’s good to be cautious of how small-sampled the playoffs are, given that one or two games against a hot or cold shooting opponent could skew these numbers. Then again, half the all-defensive team is represented on the list above, and that doesn’t include reputable defenders like Joakim Noah, Luc Richard Mbah a Moute and Tony Allen.
Because the playoffs are not only small sampled in games, but in opponents, it’s always important to consider matchups. Which makes Allen’s performance — mostly versus Dwyane Wade, LeBron James and Kobe Bryant — that much more impressive.
For those wondering about Kevin Garnett and Tim Duncan, they both just missed the cut. Garnett, to me, emphasizes the single greatest challenge in measuring individual defense causally: his greatest strength is probably communicating where to be and what is coming at all times to those around him. Now that’s difficult to quantify.
Finally, here is the complete list of defensive EV from the 2010 playoffs for all qualifying players (min 30 defensive possessions “used”).
Author’s Note: All EV values are relative to league average.
ElGee, I use counterpart data for tracking defense, but I’m sure it’s not as good as charting games:
http://thecity2.com/2011/01/31/ezpm-v-2-0-incorporating-counterpart-defense-or-the-biggie/
I can run my program on the 2010 playoffs, and see how the data compares to yours.
[…] « FG% Against Expected Value Leaders – Defense, 2010 Playoffs […]
Here’s what I get with ezpm (>3 GS):
Dwight Howard 6.45
Joakim Noah 5.55
Jermaine O’Neal 5.49
Shawn Marion 5.10
Kendrick Perkins 4.41
LeBron James 4.40
Manu Ginobili 3.59
Stephen Jackson 3.28
Carmelo Anthony 2.92
Thabo Sefolosha 2.87
Marcus Camby 2.79
Luc Mbah a Moute 2.69
Rashard Lewis 2.66
LaMarcus Aldridge 2.61
Kobe Bryant 2.50
Josh Smith 2.37
Pau Gasol 2.36
Richard Jefferson 2.33
Kevin Durant 2.33
Andrew Bynum 2.25
John Salmons 2.09
Tim Duncan 2.05
Kevin Garnett 1.79
Anthony Parker 1.67
Ron Artest 1.64
Dwyane Wade 1.64
Antawn Jamison 1.27
Al Horford 1.22
Caron Butler 1.11
Nenad Krstic 1.08
A fairly high correlation between our two lists, as I would expect.
Cool Evan – you could actually test the correlation using the PDF file I have at the bottom…some of the numbers look quite similar.
Indeed, I think I will do that soon.
Sweet. Whenever you introduce a stat, I’m always clicking refresh hoping to see the actual data.
I feel better now. 🙂
Thinking through what you have here, you’ve previously compared Defensive EV to DRtg, and now you’re pairing it with your Defensive Usage a la how Dean Oliver designed his ORtg/Usage combo. However, if I’m understanding what you’ve done, your Defensive EV isn’t an efficiency metric like ORtg/DRtg, which makes the pairing with usage perhaps a bit misleading.
Would you say that you could call DEV/DUsg “Defensive Efficiency” or something along those lines? Thoughts on the usefulness of that?
Also, while I’m totally aware that the notion of VORP in basketball is loaded with issues, it would be fun to see a defensive VORP based on multiplying some manner of zero-normalized defensive efficiency by your usage.
Ah, I was hoping that wouldn’t elicit an Oliver combo-reaction (that sounds like an order at a fast-food restaurant). You are correct in thinking that I’m not “pairing” them here. I included D Usg and possessions to illustrate sample size of events. (multiplying them together gives the number of “used” defensive possessions).
For eg, Wade’s DEV is really good, but it’s nice to see it in context of only doing a few good things well per game.
re VORP: I realized that I didn’t note that all the EV values are already relative to league average. Going to add an edit to make that clear here. As far as an actually VORP-like stat, I’ll have to think more about it – it would probably have to be position-specific for starters…
[…] Comments « Expected Value Leaders – Defense, 2010 Playoffs […]