Interesting use of rim protection stats

[Errntknght]

I normally use the SportVU rim protection stats for front court players and for them many attempts against is a good thing because one can interpret it as the guy being active in defending the rim. (Of course the best rim protector is the one that opponents fear and refuse to shoot near him. The stats do not indicate there is too much of the latter effect because the Hibberts, Chandlers and Duncans do get challenged quite a lot.)

This not about the frontcourt players however. For perimeter players, shots against at the rim are clearly not a good thing and made shots at the rim are worse. The SportVU data lists makes against per game but makes against per 36 or 48 is a far more useful stat for comparing players (guys that play few mpg will look better in per game stats, while for stats like rebounds low minute players look worse.)

I applied this (per36) to the Suns guards with this rather surprising result:
player OppFGA@rim OppFGM@rim
Bledsoe 2.2 1.39
Dragic 1.9 0.98
Thomas 1.1 0.64
Green 2.0 0.81
Goodwin and Ennis have too few minutes for the stat to mean much of anything but here they are:
Goodwin 2.3 0.96
Ennis .5 0.00

Small forwards - both PJ and Marcus play some at PF so the interpretation isn't so clear:
Marcus 3.6 1.95
Tucker 3.3 1.70
TJ Warren has very scanty minutes, too
Warren 2.7 2.09

How do our PGs stack up against other PGs - among the 59 who have 500+ minutes: Thomas 15th, Dragic 39th, Bledsoe 53rd

Some noteable PGs
Ty Lawson 3rd 0.8 0.38
Chris Paul 14th 1.2 0.61
Derrick Rose 18th 1.6 0.67
Damian Lillard 22nd 1.2 0.70
Tony Parker 28th 1.0 0.79
Westbrook 36th 1.3 0.92
Kyrie Irving 52nd 1.8 1.26

This list validates the stat to me... Lawson, Paul and Irving I have some sense of their defense and I'd put then in that order. The others I'd put in between Paul and Irving though I don't have enough sense of their game to order them.

 

[Phrazbit]

I think you're being a bit too cute here and using a stat the way it is not intended. Eric Bledsoe does a lot of weak side help defense, him showing up with a lot of field-goal attempts at the rim are the result of him playing good team D and attacking guys at the hoop, and on the other hand you have guys show up who might look good in this stat, but all it really shows is that if they get beat on the perimeter they do not follow the man inside.
If there was not a correlation between field goals against for talented big men and this stat then I think it's rather silly to try and use it with guards. Bad defenders typically don't try to challenge guys at the hoop, Thomas showing up as the best on our team in this stat should have been a major red flag.

 

[slinslin]

I am sure Nash "allowed" few field goals at the rim.

 

[Errntknght]

I think you're being a bit too cute here and using a stat the way it is not intended. Eric Bledsoe does a lot of weak side help defense, him showing up with a lot of field-goal attempts at the rim are the result of him playing good team D and attacking guys at the hoop, and on the other hand you have guys show up who might look good in this stat, but all it really shows is that if they get beat on the perimeter they do not follow the man inside.
If there was not a correlation between field goals against for talented big men and this stat then I think it's rather silly to try and use it with guards. Bad defenders typically don't try to challenge guys at the hoop, Thomas showing up as the best on our team in this stat should have been a major red flag.

The stats are intended to show the frequency with which defenders are within 5 feet of an opponent shooting within 5 feet of the basket and the make %, and that is just how I used them - surely that was clear. Yes, big men playing near the basket have much higher frequencies of attempts than PGs so the stat is more useful for them, but I believe it does give us some useful information about PGs defense - not the whole story, of course.

It was a red flag when IT and Eric were flip-flopped from what I expected to see and that led me to looking at the rest of the PGs. It did occur to me that one reason Eric rated poorly on these stats was that he normally guarded the more difficult opponent between
the opp's PG and SG - but I imagine Ty Lawson and Chris Paul do that to a considerable extent, too. You claim it is due to Eric's weak side help defense, and you may be right, but it is not something that has stood out to me - I will watch for it from now on. I watched the LAL game with that in mind and didn't see examples of it. Nor did I see examples of IT backing off when his man penetrated.

IT is ripped on here regularly for his bad defense and I've seen examples of it so I never challenged anyone for ripping him. For the time being, at least, I'll believe that he's probably better than we've giving him credit for and try to watch more objectively. Its well established that people see what they want or expect to see much better than they see what they don't want or don't expect to see, myself included - and stats give us a way being more objective. Heck, there are coaches that have assistants charting the
frequencies with which certain kinds of breakdowns/failures occur during games because they know that general impressions are not reliable. I remember an article I read some years ago in which Pat Riley or someone on the Heat staff talked about the wealth of things Pat had his guys charting for him. My impression was that other coaches did it but Riley took it to extremes.

If you'd like to give me a group of PGs with your rating of their defense in this aspect of the game I'll give you their stats so you can see how the two correlate. (Of course, you could do it yourself but its a pain to find them and convert to per 36.) I don't even have general impressions about the D of most PGs but Ty Lawson and Chris Paul are generally regarded as very good and Irving's regarded as quite poor so I took them as benchmarks and the stat agreed with the consensus.

Edit: BTW, the averages for the 59 PGs were FGAagainst 1.5/36 and FGMagainst .97/36.

Where you got the idea that the stats does not correlate with FG against for talented big men is beyond me - if it was something I said then I did not make myself clear. It was an aside to principal point of the post so I didn't elaborate - but I will in another post.

 

[Phrazbit]

I think the issue here is the premise of the stat ends up punishing guys for hustle and help defense. Bledsoe shows up not because of the tougher assignment but because he does a LOT of help defense near the rim, way more than you'd expect from a guard. Thomas on the other hand looks good in the stat because if he gets beat on a sloppy effort at a steal and his man goes to the hoop... Thomas does not follow him in.

It also rates Derrick Rose, considered a very poor defender, well and Westbrook, considered a great defender, poorly, and it has the Suns guard defenders completely reversed.

Its a stat that IMO cannot be applied with any real conclusions to guards and the nonsense numbers back that up.

If you can find a way to show how often THEIR MAN gets attempts at the rim, regardless of where they end up on the play (IE a stat that reflects a guy like Thomas giving up layups even though he gave up on the play on the perimeter, and does not punish a player for coming to help on the weakside when someone else gets beat) then I think you might be able to draw something from that, but as this stat is applied you're punishing players for hustle.

The part I mentioned about it not correlating to bigs came from this line:

The stats do not indicate there is too much of the latter effect because the Hibberts, Chandlers and Duncans do get challenged quite a lot.

Again, its not because guys are deciding "I want to challenge Chandler at the rim", its because Tyson Chandler and other good interior defenders make the effort to get there and attack the guy coming in for a layup. It reflects their defensive effort more than a willingness for the other team to go at them.

 

[Errntknght]

You don't have to convince me that the stat could punish guys for helping on defense but the question is how much do they actually do it. The same applies to attempts at steals when the guy starts his drive as the defender that attempts the steal almost always get left in the dust. It would nice to have stats that gave that kind of information.

There are other things to consider - for example last year Bled was a very good defender at the rim, opponents scored only 41% of the time against him and he challenged 3.1/36 while this year opps are scoring 64% and he challenges 2.2/36. So your claims about his D would be supported by his numbers last year, but this year not so much.

Consider Westbrook and Rose. Westbrook challenges 1.2 shots while Rose challenges 1.5 so it seems unlikely that Westbrook is doing much helping. Westbrook's % against is 67.9 and Rose's is at 37.5% against. I'm inclined to think that in the past Russell was a great defender but not at present, while Rose has improved - he certainly has a coach that would push him get better on defense, while Westbrook does not. One other point, Westbrook leads all of these PGs in steals/36 at 2.7 while Rose is one of the lowest at 0.8/36. Who is more likely to be moving his feet and who is more likely to be gambling for steals?

I probably should have listed the % against instead of the FGM because few people would be inclined to disregard a shooting percentage. I like the FGM because it combines both stats into one that I can sort on. Tell the truth, if you'd seen that Eric was scored on 64% of the time and Westbrook 68% wouldn't your first thought have been that there is something amiss with these guys, instead of that the stats were misleading?

 

[Phrazbit]

I probably should have listed the % against instead of the FGM because few people would be inclined to disregard a shooting percentage. I like the FGM because it combines both stats into one that I can sort on. Tell the truth, if you'd seen that Eric was scored on 64% of the time and Westbrook 68% wouldn't your first thought have been that there is something amiss with these guys, instead of that the stats were misleading?

No, because I don't think its a stat that translates to guards. You're talking about rim protection here, there are too many variables that are out of a guards control, is he playing help D? Is he the only guy that hustled back to stop a fast break? Is there an interior big in there that is GOOD at challenging shots right next to him or is there a terrible/lazy one? Furthermore, as previously mentioned with a guy like Bledsoe and Westbrook, these are guys who will challenge anyone at the rim, regardless of position or if its their man. You bring up Rose and Westbrook's percentages as evidence that one is getting better and the other sliding, but Rose plays on a team with 3 big guys who are great at protecting the rim, thats not Rose snuffing out percentages, its the Bulls solid bigs and team defense. There are a lot of new and interesting analytical stats that I think are great, and I actually do find the rim protection stats very interesting when it comes to big men, but to try and apply it to guys who by design are not assigned that job, I think is silly.

Like I said before, if you could come up with something that shows how often their man is getting passed them and to the rim, that would be one thing, but this isnt what the stat is showing. There is very little situational context... and on a stat thats out of context for a guard to begin with.

 

[Errntknght]

The only part of your argument that seemed relevant to me was that Rose might well be helped by having much better rim protectors on his team but when I looked I found that OKC's bigs were just as good at rim protection - maybe a little better. (If you check out these numbers they will be somewhat changed because I haven't updated my data for a week.)


Chicago's bigs: rim protection:
FGAagainst FG%against
Pau Gasol 8.5 45.8
Taj Gibson 5.1 44.2
Joachim Noah 6.1 53.5
Nikola Mirotic 5.9 60.8
*Derrick Rose 1.5 37.5

OKC's bigs: rim protection:
Steven Adams 8.6 47.6
Serge Ibaka 7.5 41.9
Kendrick Perkins 7.3 46.4
Nick Collison 6.3 57.6
*Russ Westbrook 1.2 67.9

I think we can look forward to better defensive statistics with things like times beaten off the dribble; shooting percentages against all over the floor; frequencies of giving help. I'd like the number of screens set, too, and I'm sure I could think of much more.

 

[slinslin]

Phrazbit pretty much in detail explained why the stat is useless applied to Point Guards.

It penalizes all PG's that make the attemps to block a layup, or come from the weakside even to challenge shots.

As phrazbit says it is not coincidence that athletic PGs that actually do block shots look "bad" when you try to interpret the stat this way.

No sane person would argue that Goran Dragic or Isaiah Thomas are better at defending the rim than Eric Bledsoe or Russell Westbrook.

 

[Errntknght]

No person, sane or insane, made the claim that Thomas and Dragic are better at defending the rim than Bled or Westbrook. I gave some stats regarding rim protection where both Eric and Russell rated much worse this year than they did last year, with ratings this year that were worse than Dragic and Thomas. What really puzzles me is why such a dramatic change from one year to the next - my first thought in Westbrook's case was that his injuries might have caused it.

I also looked at the overall stats on PGs to see if there was a change - its new technology and they might have changed it - but the average values for the two years were the same.

I did notice something doing that - almost every guy that I thought was condsided a a very good defender, had a very good rating one year and a poor one in the other year. Rondo, Lawson, Paul, Holiday, Beverley, Teague, Tony Allen. Their percentages are what changed while the FGA numbers where about the same year-to-year. (I threw in some other guys - Wall, Lin, Irving, Dragic, Thomas, Rose - and their numbers were much more consistent - about what I expected for everyone.) Injuries probably played a role but that level of variation has to a major fluke. The sample sizes are small - in the range of 50 to 150 - so that could be a contributing factor.

I do draw a few conclusions from these stats - Rose, Jeremy Lin and Isaiah are probably better defenders that they are given credit for.

 

[Phrazbit]

No person, sane or insane, made the claim that Thomas and Dragic are better at defending the rim than Bled or Westbrook. I gave some stats regarding rim protection where both Eric and Russell rated much worse this year than they did last year, with ratings this year that were worse than Dragic and Thomas. What really puzzles me is why such a dramatic change from one year to the next - my first thought in Westbrook's case was that his injuries might have caused it.

I also looked at the overall stats on PGs to see if there was a change - its new technology and they might have changed it - but the average values for the two years were the same.

I did notice something doing that - almost every guy that I thought was condsided a a very good defender, had a very good rating one year and a poor one in the other year. Rondo, Lawson, Paul, Holiday, Beverley, Teague, Tony Allen. Their percentages are what changed while the FGA numbers where about the same year-to-year. (I threw in some other guys - Wall, Lin, Irving, Dragic, Thomas, Rose - and their numbers were much more consistent - about what I expected for everyone.) Injuries probably played a role but that level of variation has to a major fluke. The sample sizes are small - in the range of 50 to 150 - so that could be a contributing factor.

I do draw a few conclusions from these stats - Rose, Jeremy Lin and Isaiah are probably better defenders that they are given credit for.

Your huge variations are IMO because of the variables I brought up earlier that are totally out of control to the guys you're judging this stat against.

If you think Thomas or especially Lin is underrated defensively then I suggest you watch them more closely... or look up other metric defensive stats besides this one which IMO clearly does not remotely translate to guards. Lin rates as a wretched defender in virtually every defensive metric available. This stat you're applying to him rewards him for the very activities that make him such a raggedy defender, he just lets guys blow past him... and does not follow them in. Opposing PGs put up a mind blowing 22 PER when Jeremy Lin is on the court. He is an absolutely horrible defender.

Your stat is giving you wild fluctuations from year to year, its spitting out names of guys that are known to be awful defenders and are reflected as such in virtually every other defensive metric... I mean... I don't know how you can look at those results and still try to apply them with any meaning.

 

[Errntknght]

Your huge variations are IMO because of the variables I brought up earlier that are totally out of control to the guys you're judging this stat against.

If you think Thomas or especially Lin is underrated defensively then I suggest you watch them more closely... or look up other metric defensive stats besides this one which IMO clearly does not remotely translate to guards. Lin rates as a wretched defender in virtually every defensive metric available. This stat you're applying to him rewards him for the very activities that make him such a raggedy defender, he just lets guys blow past him... and does not follow them in. Opposing PGs put up a mind blowing 22 PER when Jeremy Lin is on the court. He is an absolutely horrible defender.

Your stat is giving you wild fluctuations from year to year, its spitting out names of guys that are known to be awful defenders and are reflected as such in virtually every other defensive metric... I mean... I don't know how you can look at those results and still try to apply them with any meaning.

What variables are in play for PGs that are not there for other players? What kind of thing might vary from year to year to cause such dramatic changes in their scored against % but not show up in FGA. Heck they're going up against the same guys and fighting around screens by the same guys - the amount they help out could change for one reason or another but that would show more in the FGA stat, which is relatively constant.

Where do you find a stat site that lists the PER against? I've never seen it anyplace I've looked. I'm not a fan of PER but PER against would better than any defensive stat you can calculate from boxscore data. I guess this isn't quite PER against because the player in question isn't always guarding his counterpart but its close enough.

In fact, I'd like to know about other sites that list defensive data by player...NBA stats lists FG% against at various distances but you have to do that one player at a time which is a pain for searching or comparing several players.

I started out with something I thought would be interesting to pursue and now I'm in the process of finding out if its useful for my purposes - which is most often finding a player who is being overlooked in some way or spotting a promising player early. Now I'm using it to rate established players to see how valid it is and you are my devil's advocate - and doing your job very well. I do appreciate it.

One thing I don't understand is why you insist that it is simply not applicable to PGs. Its pretty simple, its just a count of the number of times three things happen in the span of a few frames of video - a player is within 5 feet of the hoop, he shoots the ball, and the player in question is within 5 feet of him. A fourth thing, the ball going through the hoop may nor may not happen. The biggest general flaw with it is that fouls frequently happen in such circumstances and we don't know how they deal with them. The second biggest flaw is that there may well be players in the video who are not actually involved in the play but are within five feet - that would happen most with the bigs who are naturally in the vicinity. A PG within five feet is almost certainly involved.
I do understand that cases where a PG is beaten so badly that he is not within five feet of the shooter are not counted against him - or for him if the shooter blows the shot. Its a serious flaw and may be frequent enough to render the stat useless for PGs. Thats what I'm trying to determine - my standards are loose because I'm looking for hints of value. I don't think players simply blow by their defender so fast that he can't keep up happens too often, normally that happens because of a good screen - which we may eventually get data on because it isn't that hard to detect in a video.

What I like about PER against is that assists will figure big in that. If it allows searching or filtering of the right kind I'd use that in preference to this for PGs. (SGs may be a different story!)

The thing that worries me the most now is the large year-to-year variation in the FG% against showing for the guys you'd expect to be the most consistent - its going to take a couple more years of data to see how much of a fluke that is. Its a free ride because I really collect the data for use with evaluating bigs.

 

[Phrazbit]

What variables are in play for PGs that are not there for other players? What kind of thing might vary from year to year to cause such dramatic changes in their scored against % but not show up in FGA. Heck they're going up against the same guys and fighting around screens by the same guys - the amount they help out could change for one reason or another but that would show more in the FGA stat, which is relatively constant.

Where do you find a stat site that lists the PER against? I've never seen it anyplace I've looked. I'm not a fan of PER but PER against would better than any defensive stat you can calculate from boxscore data. I guess this isn't quite PER against because the player in question isn't always guarding his counterpart but its close enough.

In fact, I'd like to know about other sites that list defensive data by player...NBA stats lists FG% against at various distances but you have to do that one player at a time which is a pain for searching or comparing several players.

I started out with something I thought would be interesting to pursue and now I'm in the process of finding out if its useful for my purposes - which is most often finding a player who is being overlooked in some way or spotting a promising player early. Now I'm using it to rate established players to see how valid it is and you are my devil's advocate - and doing your job very well. I do appreciate it.

One thing I don't understand is why you insist that it is simply not applicable to PGs. Its pretty simple, its just a count of the number of times three things happen in the span of a few frames of video - a player is within 5 feet of the hoop, he shoots the ball, and the player in question is within 5 feet of him. A fourth thing, the ball going through the hoop may nor may not happen. The biggest general flaw with it is that fouls frequently happen in such circumstances and we don't know how they deal with them. The second biggest flaw is that there may well be players in the video who are not actually involved in the play but are within five feet - that would happen most with the bigs who are naturally in the vicinity. A PG within five feet is almost certainly involved.
I do understand that cases where a PG is beaten so badly that he is not within five feet of the shooter are not counted against him - or for him if the shooter blows the shot. Its a serious flaw and may be frequent enough to render the stat useless for PGs. Thats what I'm trying to determine - my standards are loose because I'm looking for hints of value. I don't think players simply blow by their defender so fast that he can't keep up happens too often, normally that happens because of a good screen - which we may eventually get data on because it isn't that hard to detect in a video.

What I like about PER against is that assists will figure big in that. If it allows searching or filtering of the right kind I'd use that in preference to this for PGs. (SGs may be a different story!)

The thing that worries me the most now is the large year-to-year variation in the FG% against showing for the guys you'd expect to be the most consistent - its going to take a couple more years of data to see how much of a fluke that is. Its a free ride because I really collect the data for use with evaluating bigs.

I went over most of the variables earlier, but the biggest reason is that rim protection is, by nature, not something a guard is supposed to do. I expect the Bulls, by design, don't have Derrick Rose follow penetration in, if a guy is getting to the hoop it becomes the 'bigs' job to deal with it.

Anyway, if we have not come to terms on the stat at this point I don't think we will.

I respect the search for new and different stats to calculate dudes defensively, its a pretty rough task to really gauge how good guys in the NBA are on D because while they might look spectacular for a play, it does not matter if the guy is lax for the majority of the game.

82games.com is where I got PER against. I find that site fascinating when looking at guys splits. And I think PER against on that site is a tricky stat to trust, especially for teams with combo players or that do a lot of switching, because, for example, if Bledsoe and Thomas are in at the same time Bledsoe will be covering the other team's better guard, but his PER against will be reflected by SG because that's his positional counterpart in that lineup, and for a guy like Tucker he is ALWAYS against the other team's top wing player, not only giving him a tough task but sometimes that stat will be attributing to him the shoddy work that Green is doing along side. But with Lin... its too lofty of a number to ignore. And I think anyone who has watched him play would not be surprised by the stat.

But I do find a lot of useful stuff on there, the guys with extreme splits cannot be chalked up to fluke. Such as Green's numbers. His on court off court stats make a really strong case that the Suns would be better without him. A look back at Beasley's numbers in 2013 are high comedy. Going back year by year you can see Kobe's crumble into now become a below replacement level player. It also has really cool and really detailed 5 man rotation numbers.

Its a fascinating site.