96 replies to this topic

[thejohnnygold]

Yeah, this stat is supposed to be used in conjunction with the eye-test. For example, Tony Parker and Conley's have a high defensive RAPM but alot of that has to do with the fact they're on good defensive teams. Obviously they wouldn't be as effective defensively if they were on the Blazers. The same could be said about Rondo though who I think is an overrated defender, he's obviously a good defender for a pg but he's not the best (he gets beat alot but his team mates are always there to help him), Chris Paul and Avery Bradley are definitely better.

 

As long as you're using alot of the "eye-test" (50-80%) and a large sample size (at least 2000 minutes) then there's no reason why you can't say this stat is thought provoking at the very least. 

 

I wrote out a response to this--mostly stating that you aren't telling me anything I don't know and are even looping things I originally pointed out to you back to me (deleted that)....Instead, I will post some links that more effectively (logically and mathematically) explain why this stat is nothing more than a version of a version of a version of a statistic that doesn't work but stat nerds won't let die.  It needs to die.  You can't take raw data, manipulate it to death, and once it gets to the results you want to see then stop and say, "Aha, it works!".  It's ridiculous.  The whole point of RAPM is to try and create an "objective and un-biased" statistic....which is built, based on, and loaded with subjective and biased factors manipulating the data....am I the only one that thinks this is insane!?!?!?  Ugh, the whole thing is slightly infuriating.  Please, read these links...Feel free to use RAPM, but as a supporting statement in any discussion it will hold zero weight for me.

 

This details the adjusted plus/minus statistic and its inherent problems

 

Here is the regularized adjust plus/minus link

 

Here is some yahoo's stats on rapm and the over/under rated players

 

Since I know you like to stay ahead of the curve...check out this guy's "next step" beyond RAPM

 

Sorry, I don't mean to be surly about it....it just makes me irrationally upset.  This is not physics--where universal laws are in place and coefficients remain constant and reliable regardless of variable.  To apply such techniques and principles in a blanket fashion to basketball games (which can best be described as contained chaos) is to forget what it is to be human--we cannot be categorized in such mechanical fashion.  It is a very slippery slope.  It is possible to be too smart for one's own good and I think this is a prime example.  It is much like an overly-engineered piece of machinery--yes, it's design is brilliant....but is it the best way to do it?

[thejohnnygold]

Just found this list of first round play-off scores based off of this guy's RSPM metric...curious what you think as I am pretty sure you have watched a lot of the playoffs so far.  I'm guessing that the small sample size skews these numbers, but there are quite a few that leave me scratching my head.

[Mason Khamvilay]

I've come across every single one of those links before, and I understand what you're saying. 

 

Still, I've seen professional analysts use the on/off court numbers at www.82games.com, and don't those stats have the same flaws RAPM do? 

[Mason Khamvilay]

But these features don't mean the stat is worthless, as some in the APBRmetrics community would have you believe. (APBRmetrics is the term for hoops sabermetrics, derived from the Association for Professional Basketball Research.) Dallas revealed in its title march last season that plus/minus can be repurposed for a far loftier goal than evaluating individual play -- it can be used to identify winning lineups.

http://espn.go.com/n...c-espn-magazine

 

So someone like Millsap for example is always part of a winning line-up. That might not be saying as much as you would like it to, but you can't say it says nothing at all.

 

One example of what RAPM has done for me was intrigue me into watching Millsap more closely, watch the players he get paired with, understand his role ect. to get a better understanding of why his team is so much better when he's on the floor.

 

Conclusion: Millsap plays really smart and hard, he moves the ball like a hot potato, he creates alot of turnovers, runs the floor both ends every time, and his style of play would fit well on the Rockets. 

 

If used in that way, what's so wrong with that?

[thejohnnygold]

Nothing....except that the stat isn't being used anymore....you're just using common sense and watching the games.  That's my point.  If you want to use it to identify players to observe more closely then fine, but I don't think one needs to put so much work into recognizing what should be obvious if you just watch the games.

 

My problem with it is when the stat is used as a validation point.  It is unreliable, vague, and largely contrived.  For the record, I have grown disenchanted with most of these 'advanced" stats for the same reason.  Once you start applying subjective coefficients based on speculative averages you are "shading" the numbers and moving further away from objectivity--which is the whole point of these stats.

 

I get that these guys are trying to take into account the variables of who is on/off the court, who matches up against who, etc.  But there is so much more to it--Without true context how can a stat tell me if a referee did a terrible job in a game and punished one player while rewarding the other?  I could go on, but gotta run....

[Mason Khamvilay]

Without true context how can a stat tell me if a referee did a terrible job in a game and punished one player while rewarding the other? 

Over the course of 2000+ minutes?

[thejohnnygold]

Yes, over 2000 minutes I still find the stat to be crap.  I was only pointing out ONE of it's faults there....and got cut off so I didn't finish with more examples....every single play has irregularities.  NBA defense is a team concept...this isn't 1 on 1....mismatches, coaching, teammate error, chance, nba offensive preference (or should we call it deference?), on and on....the box score merely shows score or no score without telling how or why....

 

Combine that with contrived coefficients designed to generate desired results on a re-tread of a re-tread of a stat that still has huge variance and irregularity problems and you have a statistic that is highly sophisticated....and not much else.

 

If I want to know if someone can play defense all I need to do is watch.  Like I said before, feel free to cite it, but the stat carries no weight with me.

 

Edit: you still haven't addressed the values I mentioned above in post #12--and that's just in the top 20...My eye test says this statistic has no idea what it is talking about.

[Mason Khamvilay]

But these features don't mean the stat is worthless, as some in the APBRmetrics community would have you believe. (APBRmetrics is the term for hoops sabermetrics, derived from the Association for Professional Basketball Research.) Dallas revealed in its title march last season that plus/minus can be repurposed for a far loftier goal than evaluating individual play -- it can be used to identify winning lineups.

http://espn.go.com/n...c-espn-magazine

[Steven]

+/- differential is the best stat in basketball. If a team is better with a player off the floor then on the floor then said player should not play no matter who said player is. Like K-Love had a +9 differential when he was on the court versus off for the Wolves in 2011-12 (Wolves were +2 when on -7 when off).

[Mason Khamvilay]

Yeah, it definitely isn't useless. Here's something Eric Spoelstra said that I randomly came across and thought I'd share:

 

NBA.com: Do players understand plus-minus?

E.S.: I think now they do, because it's much more prevalent than it was a few years ago. You see it on everybody's website. Journalists now talk about it more.

You can argue all you want about this or that, but at the end of the day, the most important statistic is the result on the scoreboard when you're on the floor. I think that resonates with players, so we challenge our guys all the time. Make your minutes a positive. Whatever minutes you're out there, don't let the score go the other way.

After each game, I have a big printout, about 30 pages, of different statistics. I don't look through all the pages, but there are some that I immediately look at. I'll always look at the plus-minus of individual players to start, but I think that can be deceptive. From there, you have to look at the combinations.

NBA.com: Plus-minus should never be dismissed, but it always has to be taken in context.

E.S.: Absolutely. So I never look at just the player. I look at different combinations and all different layers. I probably learned that from Pat and Stan more than anything. You take all the data and you'll gain a lot more information than you had before. But ultimately, the most important thing it will do is get you to ask more questions and seek different answers.

link

 

So as long as we watch the game and understand the different combinations, it's perfectly fine to use +/- or RAPM 

 

"the most important thing it will do is get you to ask more questions and seek different answers." 

 

 One of the recent questions RAPM has helped me ask myself is "Is Deron Williams not as good of a defender as I thought?" because I remember watching him play defense in Utah and thinking he was good at it, so I started paying closer attention to his games this season and I noticed that even though he has great lateral movement he actually gets beat alot, and soon as you get past him he just gives up. But ofcourse I understand that he plays on a bad defensive team so I do take that into account.

[thejohnnygold]

Ah, see...this is more inline with how I feel about this stat...

 

I have said from the beginning that it is a team statistic and is not very useful when applied to an individual. (which is what got this whole ball rolling).

 

Now, in the Dallas example above that is useful.  Discovering an entire line-up that consistently produces a positive +/- and then observing with the eyes to understand why is perfectly valid and smart. 

 

My beef with the +/- and RAPM, etc. is it's inability to discern causality for the score and the fact that certain scores seem to me to be far from accurate--hence, any one individual score may or may not accurately reflect a player's ability and without already knowing the player's abilities how can one know for sure?  Further, those scores are a reflection of the team's performance--even after the adjustments are made--as it is a team game and that matters.

 

I am all for considering team +/- stats and giving credit to players based on that, but these arbitrary individual scores still don't hold water for me.

[Mason Khamvilay]

That's fair. So basically what you're saying is that a good individual RAPM says "plays well with others" but doesn't say much beyond that. 

[Freebird]

I like Love.  I'm OK with Smith.  I'd prefer Aldridge.  Good points have been made about their availability.  What would it take to get Smith here?  Is it worth it in the long run?  The short run?

 

I think I'd rather wait and see what comes out in the summer after this.  If we miss on a guy - okay, we're still who we are now, with a year under our belt.  This is a good thing.

 

And why focus on these 3 and D12?  Honestly, there are a few younger guys who look decent, and we might be able to get at half the price.  Ed Davis would be decent.  Tristan Thompson might be had, but I doubt it.  Perhaps even Ryan Anderson or Speights.  Other than Thompson, these are bench guys that could be available, have enough seasoning to not suck, can let our guys grow for a year or two, and will cost us next nothing. Best of all, they are mostly on teams that are going to be upgrading son via the draft, so might be expendable.

[thejohnnygold]

That's fair. So basically what you're saying is that a good individual RAPM says "plays well with others" but doesn't say much beyond that. 

 

I guess that is a pretty good way of saying it, but again, I still have a problem with the errant scores that don't seem to accurately reflect the player in question.  I mean, am I supposed to believe that Paul George is just barely above average on offense?  The stats say yes, but my eyes say no....lots of numbers like this that I can't explain.....makes me question all of the numbers.

[Mason Khamvilay]

That's a good question, maybe we should watch more Indiana games to find out. 

 

Personally I think Paul George is still emerging as a true force offensively, he has had games this season where he failed to make the correct passes and decisions in general although he has gotten much better as the season progressed. 

[thejohnnygold]

Absolutely.  I haven't gotten to watch him very much either.  Looking at his shot chart he looks pretty solid around the rim and from three, but his mid-range needs work.  Still, he averages about 17 points, 8 rebs, and 4 assists per game on 42% from the field and 36% from three according to vorped.com.

 

That's gotta be better than an offensive score of +.5, right?  Since a score of zero is supposed to be the league average does that mean those numbers are essentially average?  For me, the bottom line is he came in and flat out made Danny Granger expendable...Whatever his +/- score is--that's the key factor.

 

Further, if Indiana offered us a trade straight up: George for Parsons...I think that trade has to go through...which is saying a lot because I think Parsons is going to wind up being very, very good.

 

I wish that guy still posted the full data set so we cold cross reference more players....

[RollingWave]

Well, 42% FG isn't good,  unless you shoot almost nothing but 3s, which George doesn't. 

 

Take a look at his shot chart, he take a bit too much long jumpers and makes them at a rate no better than his 3, that's not good for your efficiency.

 

his True shooting percentage and EFG is essentially the same as Lin. he does it at a better volume (as more plays go through him) however Lin makes far more assists.  

 

So I think slightly above average offesnive player is actually about right, he is a great great defender though. he had a defensive winshare over 6!! this year,  from a wing player that's amazing. that's basically peak Scottie Pippen territory, let that sink in for a second, yes that means it's basically among one of the greatest defensive season from a wing player all time.

[RollingWave]

And really, Indiana suck at scoring despite some pretty solider scorer like West / Hibbert / George / Hill there are several of these awesome defense terrible offense team in the league that purist would love but average fan hates , the Bulls / Pacers / Memphis are basically it, and unsurprisingly they're doing better in the playoffs so far than you'd expect from their regular season record.

I really hope McHale can bring it in the off season and we show up with greatly improved D next season.

[Mason Khamvilay]

Agreed, slightly above average offensively is about right for Paul George for this season, and I expect that it will be higher next season because he has obviously been improving--the MIP award is a good indicator of that. Defensively he is a stud, I won't be surprised if he is in DPoY talks next season.

 

I'm a little worried about our defensive schemes when coach Sampson is no longer with us, but aside from that, all signs do point towards improved defense next year. I remember seeing the Rockets at 21st in defensive efficiency at one point, but we finished the season at 16th indicating significant improvement. 

[RollingWave]

They need to figure out how to defend in transition, I'm no expert on this, but a few obvious things would be...

 

A. falling down and slow to get back will obviously not end well on the other end. Lin's guilty of falling down a lot but Harden's probably the worst in terms of not getting back up in time. which is one thing I have against his excessive attempts to draw fouls, yes it is a good way to free points, but there are some hidden drawbacks like this as well, you know, the refs aren't going to call it all the time.

 

B. offensive rebound, obviously there will be no transition attack if you grab it,  the Rockets seem to have been unable to find any semblance of balance in this though, they either over gamble and have everyone crash (which ends up in plenty of hilariously easy breaks) or no one crash at all and gives it up freely, they started to find something of a balance toward the end but still.. not very ideal.

 

C.  experience and coordination: obviously this part should come with time, though not entirely sure how fast. they need to at least bother the ball handler a bit and keep track of the lead runner,  often they just do one and not the other which obviously never ends well.