30 replies to this topic

[NorEastern]

I completely agree with each of your criticisms. But the alternative that is I start "adjusting" the data. I just cannot do that. Every other team in the analysis suffers from the same benefits/detriments in the analysis. I find it much less "inflammatory" to just go with last years stats. Speculation and statistical analysis should never ever meet. It may be just me and I would welcome an alternative opinion.

I do everything possible to separate statistical analysis from my opinion. If someone is as acute as you are I allow them to ask the "Asik minutes" question and make their own adjustments as to the result of the analysis. I am more than willing to share the underlying data.

I certainly can and do build statistical models that modify the underlying domain values to reflect my opinions. But I refrain from sharing them because the arguments that result can be bloody.

PHI ranking higher than MIA in dRAPM? I spent an entire day scrubbing the data. The top 5 defensive teams from last season ranked in the top 6 this season. Houston was the only new addition among them. I looked at MIA and satisfied myself that their dRAPM ranking was realistic. And I personally never bought into the 17.5 over/under for PHI. That is a team that will hardly miss Jrue from a win/loss perspective. They sold high and may be a force in a couple of years because they did. Hey that new GM is off to a hell of a start. And then the MCW surprise. I just publish the "facts" in statistical analysis posts.

Anyway, I value your opinion. I need someone to keep me on my toes. Thank you.

Edited by NorEastern, 03 November 2013 - 11:58 AM.

[Sir Thursday]

I completely agree with each of your criticisms. But the alternative that is I start "adjusting" the data. I just cannot do that. Every other team in the analysis suffers from the same benefits/detriments in the analysis. I find it much less "inflammatory" to just go with last years stats. Speculation and statistical analysis should never ever meet. It may be just me and I would welcome an alternative opinion.

 

It's an interesting philosophy - personally I don't agree with it, but I can see where you're coming from. I think it is acceptable to have some speculation provided that it is clearly signposted that that is what it is. Any time you are using statistics to predict something you have to make assumptions about the validity of that data in the future, for example. The raw statistics provide a good starting point but any conclusions drawn from them require some decisions about which statistical measures to use and how to interpret the results and most of the time some manipulation to get them into an acceptable form. It is those discretionary choices that add value and meaning to the statistics and allow you to find results of interest.

 

ST

[NorEastern]

It's an interesting philosophy - personally I don't agree with it, but I can see where you're coming from. I think it is acceptable to have some speculation provided that it is clearly signposted that that is what it is. Any time you are using statistics to predict something you have to make assumptions about the validity of that data in the future, for example. The raw statistics provide a good starting point but any conclusions drawn from them require some decisions about which statistical measures to use and how to interpret the results and most of the time some manipulation to get them into an acceptable form. It is those discretionary choices that add value and meaning to the statistics and allow you to find results of interest.
 
ST

Great post! I try to separate my statistical posts from my opinion posts. Statistical posts depend on an underlying data set and analysis that is unassailable. They provide a firm foundation to begin discussions. I personally do not believe that the initial dRAPM graph in this post needed any modification to insert my personal opinion. Any such attempt on my part would be open to valid questions about the impartiality of the data. Right now on this forum I do not have the cred that would allow my data and analysis to be accepted without question. I hope to change that.

That being said, the responses are certainly open season for all. Speculation is appropriate there. IMHO ground a discussion in a reasonable and factual data representation that all can agree with and then let the free for all begin.

[thejohnnygold]

 Right now on this forum I do not have the cred that would allow my data and analysis to be accepted without question. I hope to change that.
 

 

On this forum, that day may never come...

[Sir Thursday]

Great post! I try to separate my statistical posts from my opinion posts. Statistical posts depend on an underlying data set and analysis that is unassailable. They provide a firm foundation to begin discussions. I personally do not believe that the initial dRAPM graph in this post needed any modification to insert my personal opinion. Any such attempt on my part would be open to valid questions about the impartiality of the data. Right now on this forum I do not have the cred that would allow my data and analysis to be accepted without question. I hope to change that.

That being said, the responses are certainly open season for all. Speculation is appropriate there. IMHO ground a discussion in a reasonable and factual data representation that all can agree with and then let the free for all begin.

 

See I think you tend to get a better discussion if you post some data and then provide an interpretation of that data. People might disagree with the interpretation (and then you get an interesting debate), but provided it is clear what are the 'raw statistics' and what are the conclusions you've drawn from those statistics then I don't think it's a problem. I like to think people should be able to separate out the two and not conflate them when debating the results of someone's work.

 

ST

[NorEastern]

See I think you tend to get a better discussion if you post some data and then provide an interpretation of that data. People might disagree with the interpretation (and then you get an interesting debate), but provided it is clear what are the 'raw statistics' and what are the conclusions you've drawn from those statistics then I don't think it's a problem. I like to think people should be able to separate out the two and not conflate them when debating the results of someone's work.
 
ST

You may be correct. I will try it out next time. But the one thing that I do not want is for the discussion to degrade into a cat fight over the data. In this post I expected people to really attack the Bev point I made in my last sentence. It is a good sign that they did not.

On this forum, that day may never come...

Actually come to think about it, that may be a good thing.

[thejohnnygold]

Sir Thursday, I agree with your view on this matter.  One of my biggest qualms with statistics is the manipulation of data.  I understand that it is necessary, but it often seems like a cloak of secrecy is pulled over these manipulations which leaves me in a dubious state about the conclusions.  I would rather see and understand the process to better understand the results.  In this way I feel I can better decide how much I can take away from it all.

 

NorEastern, I understand your concern about the potential catfights, but I think if people understand where you are coming from they will respect the analysis even more--they won't always agree with your choices, but it will give more value to the insights in my opinion.

 

Keep up the good work (you too Mr. Wehr)!

[timetodienow1234567]

I agree. I hate when stats guys say "I made minor/trivial/etc... Adjustments" I automatically think they tailored the data to fit their opinions. I consider myself smart enough to understand data analysis techniques and would appreciate if writers would go into more detail about their process.

[NorEastern]

I agree. I hate when stats guys say "I made minor/trivial/etc... Adjustments" I automatically think they tailored the data to fit their opinions. I consider myself smart enough to understand data analysis techniques and would appreciate if writers would go into more detail about their process.

If I ever fail to explain what I have done please ask and I will be glad to elaborate. From the previous page of comments:

"RAPM is presented as +/- per 100 possessions. I modified each players contribution to the team by their total minutes. I double checked using player possessions and the value was comparable."

[Chichos]

I do everything possible to separate statistical analysis from my opinion. If someone is as acute as you are I allow them to ask the "Asik minutes" question and make their own adjustments as to the result of the analysis. I am more than willing to share the underlying data.

I certainly can and do build statistical models that modify the underlying domain values to reflect my opinions. But I refrain from sharing them because the arguments that result can be bloody.

PHI ranking higher than MIA in dRAPM? I spent an entire day scrubbing the data. The top 5 defensive teams from last season ranked in the top 6 this season. Houston was the only new addition among them. I looked at MIA and satisfied myself that their dRAPM ranking was realistic. And I personally never bought into the 17.5 over/under for PHI. That is a team that will hardly miss Jrue from a win/loss perspective. They sold high and may be a force in a couple of years because they did. Hey that new GM is off to a hell of a start. And then the MCW surprise. I just publish the "facts" in statistical analysis posts.

Anyway, I value your opinion. I need someone to keep me on my toes. Thank you.

I think people saw your data and assumed you were forecasting since Parson's quote was about the rest of the year.  I think once you dip your toes in the forecasting pool making logical assumptions about changes in variables comes with the territory.  Honestly, it would be interesting to see what happens if 2009-10 dwight were manning the middle, or what happens if Dwight puts up last years number's while Asik's decrease across the board by 10%.  If you clearly define each situation we can get an idea of possible outcomes and what specific things happened to make them come about.  Creating best and worst case scenarios to present really helps support your underlying argument.

[rockets best fan]

Sir Thursday, I agree with your view on this matter.  One of my biggest qualms with statistics is the manipulation of data.  I understand that it is necessary, but it often seems like a cloak of secrecy is pulled over these manipulations which leaves me in a dubious state about the conclusions.  I would rather see and understand the process to better understand the results.  In this way I feel I can better decide how much I can take away from it all.

 

NorEastern, I understand your concern about the potential catfights, but I think if people understand where you are coming from they will respect the analysis even more--they won't always agree with your choices, but it will give more value to the insights in my opinion.

 

Keep up the good work (you too Mr. Wehr)!

I agree also. I don't comment on much of the data.......I am still an infant in my stat geekiness  but when the parameters are explained at least the angle of approach is better understood