Rim Protection

[knarsu3]

using Vantage stats from the last 2 years: http://blog.cacvantage.com/2014/01/who- ... asket.html

The big question for me is which of the two metrics among rim protection rate (measures how often a player contests, alters, blocks shots) and rim protection FG% (FG% on those contested/altered/blocked shots) is a better predictor of defensive efficiency. I plan to look at this on a team wide level and also it's variance over time but for now, I don't have the answer. I would hypothesize that rim protection rate is the better stat since if you are leaving a player open, his FG% is going to have nothing to do with your defense. Alternatively, if you can contest the shot, the player shooting will have a worse % no matter who is contesting the shot. Still, I'm sure there are different effects of say Brook Lopez contesting a shot versus Larry Sanders.

[shoddypippen]

Hey, thanks for the interesting numbers. A few questions:

1. Do you find it suspect that a player like Andre Bargnani comes out near the top in your rim protection rate metric?
2. Do you think including blocking % in the metric skews the results toward weak-side helpers, like Ibaka, and away from defenders who deter shots near the basket?

[knarsu3]

Hey, thanks for the interesting numbers. A few questions:

1. Do you find it suspect that a player like Andre Bargnani comes out near the top in your rim protection rate metric?
2. Do you think including blocking % in the metric skews the results toward weak-side helpers, like Ibaka, and away from defenders who deter shots near the basket?

It is a little surprising to see Bargnani come out near the top. However as I mentioned in the article, we still don't really know which one of the metrics correlates better with defensive efficiency. And he was roughly (slightly better) average in rim protection FG%. Still, that was certainly surprising.

Are you referencing an "intimidation" type factor? For example, it certainly is possible that a player like Hibbert, who has a very low rim protection FG%, is going to deter players from shooting at the rim against him. That's also something Close% may not necessarily pick up on if the player just simply doesn't have a lot of shots attempted against him.

Also, I have a new article out that explores expected points per shot using Vantage's data. Specifically, you can see the points per shot on contested shots at various shot locations:
http://blog.cacvantage.com/2014/01/is-l ... fense.html

[sethypooh21]

I've played around with the SportVU rim protection #'s a bit. My underlying assumption is that contesting the shot at all is roughly as important as the effectiveness of the contest - the average shot <5th contested by a big man is right around 50%, with the range of most bigs being in the 40-60% area, whereas an uncontested shot <5 feet is about 75%, so I scaled the "points saved" from a big man's contests to what a "league average" big would do playing the same minutes for the same team (based on OFGA/gm <5ft) to capture some of pace, deterrence and most of all to not give guys a "bonus" for playing on teams with crap wing defenders. The "points saved over average" is somewhat sensitive to how wide a net I cast in terms of counting guys as bigs, which is a slight problem given the increasing prevalence of small ball 4s. That said, the integer values are less important than the rankings, which I feel pretty decent about.

Note that I've completely ignored shot blocking as that would essentially be double counting those shots as the miss is already factored into OFG%

Unsurprisingly, Hibbert is one of the most impressive anyway you slice it, though he's only 4th on a per minute basis to Chris Kaman (only guy in the sample holding opponents <40% on a lot of contests/minute even factoring in team effects), Biyombo, and Ian Mahinmi.

Data through earlier this week is here: https://docs.google.com/spreadsheet/ccc ... _web#gid=0
explanation of the method is here: http://www.whereoffensehappens.com/1/po ... heirs.html

[italia13calcio]

@sethypooh21
I did essentially the same thing in computing my overall stat. The methodology I used is described on the site, the main difference being that I looked at every player. The order of out lists is pretty much the same (except Roy Hibbert has a far more commanding lead in mine) and most big men are in the positives, which makes sense because I'm looking at their protection vs that of an average player, while your comparing them strictly to big men.
If you want to talk more about this method, or about how to implement this method with respect to the other SportVu stats, I would be delighted to. Just pm me if you are interested - I've been looking for someone to collaborate with.

[sethypooh21]

I thought about that, but decided to include only bigmen as I thought the proper estimation of value isn't as compared to nobody contesting the shot, or a PG contesting the shot, but a proxy for the league average big. It's sort of zero sum as the better the league average is, the harder it is for a big man to provide value over the guy he would be replacing.

[knarsu3]

Right, which is why I ended up only looking at big men. That was after looking at the league averages for each position, where it was clear that there's a clear divide for big men vs. SFs and guards. The averages posted in my article are for big men.

In regards to the SportVu data, one big difference IMO, is that SportVu doesn't differentiate between contested shots- defined putting your hand up and within 3 feet and pressured shots- defined within 3 feet but no hand up. In this article, I look at the difference in value of contesting the shot and pressuring the shot. That is partly why the numbers I have will be different from SportVu's and also because, I'm working with an older sample (my data is from the last 2 seasons but not including this year).

For example, in my sample I have Hibbert with a 47.2% FG% at the rim (keep in mind my data doesn't include this year) but where we see some big differences is when we look at Hibbert's FG% on contested/altered shots- 37.5% (if I include blocks, thats where I get the 27.7% FG% from the article) and pressured shots- 75.8%. SportVu combines that all into one, so you get a FG% higher than what I have in my sample. Also, I'm not sure how SportVu treats blocks.

Anyways, that's probably why we see some differences in our data. I do think it's probably a mix of both contesting the shot and how well you contest the shot that is important. But I hope to complete that study at some point.

[knarsu3]

Note that I've completely ignored shot blocking as that would essentially be double counting those shots as the miss is already factored into OFG%

Ah, I understand what you mean by including blocks in my data now. It isn't being double counted in my data because my sample is shot by shot data (so it includes every shot and not summing up players totals). Blocks are labeled separately from pressured shots, contested shots, open shots, etc.

[sethypooh21]

Right, which is why I ended up only looking at big men. That was after looking at the league averages for each position, where it was clear that there's a clear divide for big men vs. SFs and guards. The averages posted in my article are for big men.

In regards to the SportVu data, one big difference IMO, is that SportVu doesn't differentiate between contested shots- defined putting your hand up and within 3 feet and pressured shots- defined within 3 feet but no hand up. In this article, I look at the difference in value of contesting the shot and pressuring the shot. That is partly why the numbers I have will be different from SportVu's and also because, I'm working with an older sample (my data is from the last 2 seasons but not including this year).

For example, in my sample I have Hibbert with a 47.2% FG% at the rim (keep in mind my data doesn't include this year) but where we see some big differences is when we look at Hibbert's FG% on contested/altered shots- 37.5% (if I include blocks, thats where I get the 27.7% FG% from the article) and pressured shots- 75.8%. SportVu combines that all into one, so you get a FG% higher than what I have in my sample. Also, I'm not sure how SportVu treats blocks.

Anyways, that's probably why we see some differences in our data. I do think it's probably a mix of both contesting the shot and how well you contest the shot that is important. But I hope to complete that study at some point.

I'm sort of operating under the assumption that it largely comes out in the wash. Kevin Love standing next to a guy as he lays the ball in is factored into his (really high) OFG% in SV. I do think your method has some value in determining how much of the data is resulting from "skill" (size, timing, etc) and how much from "effort" just getting a hand up at all. As far as blocks, I don't think they affect the underlying SV numbers and seem there mostly as context. For our purposes to we much care if the shot is actually blocked - an attempt with 0% EFG is plenty valuable whether we assign a counting stat to it or not.

[sethypooh21]

Just a quick update, have done several more iterations and have replaced team level adjustment with individualized OFGA<5ft data: https://docs.google.com/spreadsheet/lv? ... =drive_web