Loading ...
Sorry, an error occurred while loading the content.
 

Re: 97% correlation between Team PER-diff and winning percentage!!!!!!!!!!!!!!!!

Expand Messages
  • mrintp2000
    I m not sure what you re asking Is that not a 97% correlation? Please delineate your objections more clearly. ... dividuals is tenuous.
    Message 1 of 30 , Jul 20, 2004
      I'm not sure what you're asking Is that not a 97% correlation? Please
      delineate your objections more clearly.


      --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
      > --- In APBR_analysis@yahoogroups.com, "mrintp2000" <shzys@n...>
      > wrote:
      > > ... If this doesn't help my
      > > proselytizing efforts, I don't know what will.
      > >...
      > > San Antonio 15.4 0.695
      > > Sacramento [7.0] 0.671
      > ...
      >
      > You might start by telling us how you came up with your alleged "97%
      > correlation". Some of the data are quite disparate, in fact.
      >
      > If one focusses on the statistical anomalies (as isolated above),
      > one might find the "unmeasured" factors that lead one team to more
      > success.
      >
      > And ultimately, if one "corrects" by discounting uncorrelated stats,
      > one arrives at the suggestion that it's just Point-differential that
      > correlates to winning, on a team level.
      >
      > Breaking those down by position, of course, doesn't change the
      > total. It might suggest positional strength; but assigning these
      > strengths to in

      dividuals is tenuous.
    • Mike G
      ... Please ... You ll have to excuse me. I am a simple cave man, unfrozen in your modern world, of which I have no grasp. I was just wondering how you arrived
      Message 2 of 30 , Jul 21, 2004
        --- In APBR_analysis@yahoogroups.com, "mrintp2000" <shzys@n...>
        wrote:
        > I'm not sure what you're asking Is that not a 97% correlation?
        Please
        > delineate your objections more clearly.


        You'll have to excuse me. I am a simple cave man, unfrozen in your
        modern world, of which I have no grasp.


        I was just wondering how you arrived at your correlation, and what
        it means.

        "Is that not a 97% correlation?", really doesn't tell me anything.


        >
        >
        > --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
        ...
        > > ultimately, if one "corrects" by discounting uncorrelated stats,
        > > ... it's just Point-differential that
        > > correlates to winning, on a team level.
        > >
        > > Breaking those down by position, of course, doesn't change the
        > > total. ..
      • mrintp2000
        I used Excel to calculate the correlation, it s under Tools and then Data Analysis (which you may need to download. As for the 97%, a correlation of .8 is
        Message 3 of 30 , Jul 22, 2004
          I used Excel to calculate the correlation, it's under Tools and then
          Data Analysis (which you may need to download. As for the 97%, a
          correlation of >.8 is considered to be very strong. So you might say
          this correlation is very....compelling.

          --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
          > --- In APBR_analysis@yahoogroups.com, "mrintp2000" <shzys@n...>
          > wrote:
          > > I'm not sure what you're asking Is that not a 97% correlation?
          > Please
          > > delineate your objections more clearly.
          >
          >
          > You'll have to excuse me. I am a simple cave man, unfrozen in your
          > modern world, of which I have no grasp.
          >
          >
          > I was just wondering how you arrived at your correlation, and what
          > it means.
          >
          > "Is that not a 97% correlation?", really doesn't tell me anything.
          >
          >
          > >
          > >
          > > --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
          > ...
          > > > ultimately, if one "corrects" by discounting uncorrelated stats,
          > > > ... it's just Point-differential that
          > > > correlates to winning, on a team level.
          > > >
          > > > Breaking those down by position, of course, doesn't change the
          > > > total. ..
        • wizardskev
          I m still waiting for someone to explain why a .97 correlation with PER should be viewed as The Holy Grail, when using a different linear stat method, I got
          Message 4 of 30 , Jul 22, 2004
            I'm still waiting for someone to explain why a .97 correlation with
            PER should be viewed as The Holy Grail, when using a different linear
            stat method, I got the same correlation. A 97% correlation is very
            strong, no doubt. But why focus on PER when another (simpler) method
            gets the same result?

            --- In APBR_analysis@yahoogroups.com, "mrintp2000" <shzys@n...> wrote:
            > I used Excel to calculate the correlation, it's under Tools and then
            > Data Analysis (which you may need to download. As for the 97%, a
            > correlation of >.8 is considered to be very strong. So you might say
            > this correlation is very....compelling.
            >
            > --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
            > > --- In APBR_analysis@yahoogroups.com, "mrintp2000" <shzys@n...>
            > > wrote:
            > > > I'm not sure what you're asking Is that not a 97% correlation?
            > > Please
            > > > delineate your objections more clearly.
            > >
            > >
            > > You'll have to excuse me. I am a simple cave man, unfrozen in your
            > > modern world, of which I have no grasp.
            > >
            > >
            > > I was just wondering how you arrived at your correlation, and what
            > > it means.
            > >
            > > "Is that not a 97% correlation?", really doesn't tell me anything.
            > >
            > >
            > > >
            > > >
            > > > --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
            > > ...
            > > > > ultimately, if one "corrects" by discounting uncorrelated stats,
            > > > > ... it's just Point-differential that
            > > > > correlates to winning, on a team level.
            > > > >
            > > > > Breaking those down by position, of course, doesn't change the
            > > > > total. ..
          • Mike G
            ... linear ... method ... As has been mentioned, the correlation between team point- differential is also about .97, so there isn t any improvement when you
            Message 5 of 30 , Jul 22, 2004
              --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
              wrote:
              > I'm still waiting for someone to explain why a .97 correlation with
              > PER should be viewed as The Holy Grail, when using a different
              linear
              > stat method, I got the same correlation. A 97% correlation is very
              > strong, no doubt. But why focus on PER when another (simpler)
              method
              > gets the same result?

              As has been mentioned, the correlation between team point-
              differential is also about .97, so there isn't any improvement when
              you factor in rebounds, turnovers, etc.

              My own quasi-linear correlation (not a "differential") is only
              about .68 correlation; it improves when I adjust winning % by
              Conference (+.025 for the West, -.025 for the East).

              When I use my Sagarin correction, I get a .985 correlation.

              However, I believe there's some merit to the free-standing nature of
              the PER-diff -- if you have that data. Since I like to evaluate
              seasons from the distant past, I'm determined to use "existing
              stats" for my formulae.


              And thanks (mrintp) for the tip on Excel. I didn't have to
              download; under Tools, there's Add-Ins; then just check Analysis
              ToolPak, and you're in business. (Excel 2000)
            • wizardskev
              I m not knocking PER differential at all. My point is that by taking the team differential from a very simple linear system -- pts + reb + ast + stl + blk -
              Message 6 of 30 , Jul 22, 2004
                I'm not knocking PER differential at all. My point is that by taking
                the team differential from a very simple linear system -- pts + reb +
                ast + stl + blk - missed fga - missed fta - to - pf -- I get the same
                97% correlation with winning percentage. Why go through all the
                complex calculations necessary for PER if we can get the same result
                using a simpler system?

                --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
                > --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                > wrote:
                > > I'm still waiting for someone to explain why a .97 correlation with
                > > PER should be viewed as The Holy Grail, when using a different
                > linear
                > > stat method, I got the same correlation. A 97% correlation is very
                > > strong, no doubt. But why focus on PER when another (simpler)
                > method
                > > gets the same result?
                >
                > As has been mentioned, the correlation between team point-
                > differential is also about .97, so there isn't any improvement when
                > you factor in rebounds, turnovers, etc.
                >
                > My own quasi-linear correlation (not a "differential") is only
                > about .68 correlation; it improves when I adjust winning % by
                > Conference (+.025 for the West, -.025 for the East).
                >
                > When I use my Sagarin correction, I get a .985 correlation.
                >
                > However, I believe there's some merit to the free-standing nature of
                > the PER-diff -- if you have that data. Since I like to evaluate
                > seasons from the distant past, I'm determined to use "existing
                > stats" for my formulae.
                >
                >
                > And thanks (mrintp) for the tip on Excel. I didn't have to
                > download; under Tools, there's Add-Ins; then just check Analysis
                > ToolPak, and you're in business. (Excel 2000)
              • mrintp2000
                Because the Team-PER-diff is a sum of the Player PER-Diff s. This correlation shows that there is no synergistic effect whatsoever, again-it s a simple sum.
                Message 7 of 30 , Jul 22, 2004
                  Because the Team-PER-diff is a sum of the Player PER-Diff's. This
                  correlation shows that there is no synergistic effect whatsoever,
                  again-it's a simple sum. There are many important implications of
                  this, here are a few
                  1. No position on the floor is more important than any other
                  2. You don't need a top ten player to have a title contending team
                  3. The tremendous value of a superstar is effectively quantified (KG
                  +16 for example is worth more than 3X as much as say Kenyon Martin +5)

                  Back to the PER. First, the PER is the only system I've seen that
                  stands up to subjective analysis. It had MJ rated #1 all those years,
                  and then Shaq #1 for 5 years straight. Maybe you guys disagree?

                  Second, the PER allows players to contribute in a multitude of ways
                  yet still be considered effective players. Third, the PER-diff
                  accurately accounts for defense. Well with maybe a 25% "fudge factor"
                  when the defensive assingment might be mismatched:(

                  Again, the real value here is we can take a very effective rating
                  system for individual players, and simply add that up to create an
                  equally effective rating for teams.

                  --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                  wrote:
                  > I'm not knocking PER differential at all. My point is that by taking
                  > the team differential from a very simple linear system -- pts + reb +
                  > ast + stl + blk - missed fga - missed fta - to - pf -- I get the same
                  > 97% correlation with winning percentage. Why go through all the
                  > complex calculations necessary for PER if we can get the same result
                  > using a simpler system?
                  >
                  > --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
                  > > --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                  > > wrote:
                  > > > I'm still waiting for someone to explain why a .97 correlation with
                  > > > PER should be viewed as The Holy Grail, when using a different
                  > > linear
                  > > > stat method, I got the same correlation. A 97% correlation is very
                  > > > strong, no doubt. But why focus on PER when another (simpler)
                  > > method
                  > > > gets the same result?
                  > >
                  > > As has been mentioned, the correlation between team point-
                  > > differential is also about .97, so there isn't any improvement when
                  > > you factor in rebounds, turnovers, etc.
                  > >
                  > > My own quasi-linear correlation (not a "differential") is only
                  > > about .68 correlation; it improves when I adjust winning % by
                  > > Conference (+.025 for the West, -.025 for the East).
                  > >
                  > > When I use my Sagarin correction, I get a .985 correlation.
                  > >
                  > > However, I believe there's some merit to the free-standing nature of
                  > > the PER-diff -- if you have that data. Since I like to evaluate
                  > > seasons from the distant past, I'm determined to use "existing
                  > > stats" for my formulae.
                  > >
                  > >
                  > > And thanks (mrintp) for the tip on Excel. I didn't have to
                  > > download; under Tools, there's Add-Ins; then just check Analysis
                  > > ToolPak, and you're in business. (Excel 2000)
                • mrintp2000
                  Excel has some pretty cool stat capabilities, I don t even know what most of the terms mean yet.
                  Message 8 of 30 , Jul 22, 2004
                    Excel has some pretty cool stat capabilities, I don't even know what
                    most of the terms mean yet.


                    --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
                    > --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                    > wrote:
                    > > I'm still waiting for someone to explain why a .97 correlation with
                    > > PER should be viewed as The Holy Grail, when using a different
                    > linear
                    > > stat method, I got the same correlation. A 97% correlation is very
                    > > strong, no doubt. But why focus on PER when another (simpler)
                    > method
                    > > gets the same result?
                    >
                    > As has been mentioned, the correlation between team point-
                    > differential is also about .97, so there isn't any improvement when
                    > you factor in rebounds, turnovers, etc.
                    >
                    > My own quasi-linear correlation (not a "differential") is only
                    > about .68 correlation; it improves when I adjust winning % by
                    > Conference (+.025 for the West, -.025 for the East).
                    >
                    > When I use my Sagarin correction, I get a .985 correlation.
                    >
                    > However, I believe there's some merit to the free-standing nature of
                    > the PER-diff -- if you have that data. Since I like to evaluate
                    > seasons from the distant past, I'm determined to use "existing
                    > stats" for my formulae.
                    >
                    >
                    > And thanks (mrintp) for the tip on Excel. I didn't have to
                    > download; under Tools, there's Add-Ins; then just check Analysis
                    > ToolPak, and you're in business. (Excel 2000)
                  • wizardskev
                    Each of those points in favor of PER are true of taking the differential from other linear weight measures such as the one I mentioned. Again, why use a
                    Message 9 of 30 , Jul 22, 2004
                      Each of those points in favor of PER are true of taking the
                      differential from other linear weight measures such as the one I
                      mentioned. Again, why use a complex system such as PER, when taking
                      the differential from a more simply calculated method yields similar
                      results?

                      --- In APBR_analysis@yahoogroups.com, "mrintp2000" <shzys@n...> wrote:
                      > Because the Team-PER-diff is a sum of the Player PER-Diff's. This
                      > correlation shows that there is no synergistic effect whatsoever,
                      > again-it's a simple sum. There are many important implications of
                      > this, here are a few
                      > 1. No position on the floor is more important than any other
                      > 2. You don't need a top ten player to have a title contending team
                      > 3. The tremendous value of a superstar is effectively quantified (KG
                      > +16 for example is worth more than 3X as much as say Kenyon Martin +5)
                      >
                      > Back to the PER. First, the PER is the only system I've seen that
                      > stands up to subjective analysis. It had MJ rated #1 all those years,
                      > and then Shaq #1 for 5 years straight. Maybe you guys disagree?
                      >
                      > Second, the PER allows players to contribute in a multitude of ways
                      > yet still be considered effective players. Third, the PER-diff
                      > accurately accounts for defense. Well with maybe a 25% "fudge factor"
                      > when the defensive assingment might be mismatched:(
                      >
                      > Again, the real value here is we can take a very effective rating
                      > system for individual players, and simply add that up to create an
                      > equally effective rating for teams.
                      >
                      > --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                      > wrote:
                      > > I'm not knocking PER differential at all. My point is that by taking
                      > > the team differential from a very simple linear system -- pts + reb +
                      > > ast + stl + blk - missed fga - missed fta - to - pf -- I get the same
                      > > 97% correlation with winning percentage. Why go through all the
                      > > complex calculations necessary for PER if we can get the same result
                      > > using a simpler system?
                      > >
                      > > --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
                      > > > --- In APBR_analysis@yahoogroups.com, "wizardskev"
                      <kevinbroom@r...>
                      > > > wrote:
                      > > > > I'm still waiting for someone to explain why a .97 correlation
                      with
                      > > > > PER should be viewed as The Holy Grail, when using a different
                      > > > linear
                      > > > > stat method, I got the same correlation. A 97% correlation is
                      very
                      > > > > strong, no doubt. But why focus on PER when another (simpler)
                      > > > method
                      > > > > gets the same result?
                      > > >
                      > > > As has been mentioned, the correlation between team point-
                      > > > differential is also about .97, so there isn't any improvement when
                      > > > you factor in rebounds, turnovers, etc.
                      > > >
                      > > > My own quasi-linear correlation (not a "differential") is only
                      > > > about .68 correlation; it improves when I adjust winning % by
                      > > > Conference (+.025 for the West, -.025 for the East).
                      > > >
                      > > > When I use my Sagarin correction, I get a .985 correlation.
                      > > >
                      > > > However, I believe there's some merit to the free-standing
                      nature of
                      > > > the PER-diff -- if you have that data. Since I like to evaluate
                      > > > seasons from the distant past, I'm determined to use "existing
                      > > > stats" for my formulae.
                      > > >
                      > > >
                      > > > And thanks (mrintp) for the tip on Excel. I didn't have to
                      > > > download; under Tools, there's Add-Ins; then just check Analysis
                      > > > ToolPak, and you're in business. (Excel 2000)
                    • Gabe Farkas
                      I don t necessarily agree with your point #1. Could you speculate that Ben Wallace s imposing defensive presence in the middle (a) discouraged opposing guards
                      Message 10 of 30 , Jul 22, 2004
                        I don't necessarily agree with your point #1. Could
                        you speculate that Ben Wallace's imposing defensive
                        presence in the middle (a) discouraged opposing guards
                        from taking the ball into the lane more, thus forcing
                        them to settle for outside shots, and (b) ed to more
                        blocks when these players did decide to drive the
                        ball?

                        Thus, those PGs, SGs and SFs had slightly lower PERs,
                        and consequently Chauncey's, Rip's and Tayshaun's
                        PER-diffs were slightly inflated?

                        I would be interested to see a correlative analysis
                        between the PER-diff of every position vs the average
                        of the other 4. In other words, look at the
                        statistical correlation of the PER-diff of the C vs
                        (PG+SG+SF+PF)/4, and then the same with PF vs
                        (PG+SG+SF+C) and so on and so forth for all 5
                        positions.


                        --- mrintp2000 <shzys@...> wrote:
                        > Because the Team-PER-diff is a sum of the Player
                        > PER-Diff's. This
                        > correlation shows that there is no synergistic
                        > effect whatsoever,
                        > again-it's a simple sum. There are many important
                        > implications of
                        > this, here are a few
                        > 1. No position on the floor is more important than
                        > any other
                        > 2. You don't need a top ten player to have a title
                        > contending team
                        > 3. The tremendous value of a superstar is
                        > effectively quantified (KG
                        > +16 for example is worth more than 3X as much as say
                        > Kenyon Martin +5)
                        >
                        > Back to the PER. First, the PER is the only system
                        > I've seen that
                        > stands up to subjective analysis. It had MJ rated #1
                        > all those years,
                        > and then Shaq #1 for 5 years straight. Maybe you
                        > guys disagree?
                        >
                        > Second, the PER allows players to contribute in a
                        > multitude of ways
                        > yet still be considered effective players. Third,
                        > the PER-diff
                        > accurately accounts for defense. Well with maybe a
                        > 25% "fudge factor"
                        > when the defensive assingment might be mismatched:(
                        >
                        > Again, the real value here is we can take a very
                        > effective rating
                        > system for individual players, and simply add that
                        > up to create an
                        > equally effective rating for teams.
                        >
                        > --- In APBR_analysis@yahoogroups.com, "wizardskev"
                        > <kevinbroom@r...>
                        > wrote:
                        > > I'm not knocking PER differential at all. My point
                        > is that by taking
                        > > the team differential from a very simple linear
                        > system -- pts + reb +
                        > > ast + stl + blk - missed fga - missed fta - to -
                        > pf -- I get the same
                        > > 97% correlation with winning percentage. Why go
                        > through all the
                        > > complex calculations necessary for PER if we can
                        > get the same result
                        > > using a simpler system?
                        > >
                        > > --- In APBR_analysis@yahoogroups.com, "Mike G"
                        > <msg_53@h...> wrote:
                        > > > --- In APBR_analysis@yahoogroups.com,
                        > "wizardskev" <kevinbroom@r...>
                        > > > wrote:
                        > > > > I'm still waiting for someone to explain why a
                        > .97 correlation with
                        > > > > PER should be viewed as The Holy Grail, when
                        > using a different
                        > > > linear
                        > > > > stat method, I got the same correlation. A
                        > 97% correlation is very
                        > > > > strong, no doubt. But why focus on PER when
                        > another (simpler)
                        > > > method
                        > > > > gets the same result?
                        > > >
                        > > > As has been mentioned, the correlation between
                        > team point-
                        > > > differential is also about .97, so there isn't
                        > any improvement when
                        > > > you factor in rebounds, turnovers, etc.
                        > > >
                        > > > My own quasi-linear correlation (not a
                        > "differential") is only
                        > > > about .68 correlation; it improves when I adjust
                        > winning % by
                        > > > Conference (+.025 for the West, -.025 for the
                        > East).
                        > > >
                        > > > When I use my Sagarin correction, I get a .985
                        > correlation.
                        > > >
                        > > > However, I believe there's some merit to the
                        > free-standing nature of
                        > > > the PER-diff -- if you have that data. Since I
                        > like to evaluate
                        > > > seasons from the distant past, I'm determined to
                        > use "existing
                        > > > stats" for my formulae.
                        > > >
                        > > >
                        > > > And thanks (mrintp) for the tip on Excel. I
                        > didn't have to
                        > > > download; under Tools, there's Add-Ins; then
                        > just check Analysis
                        > > > ToolPak, and you're in business. (Excel 2000)
                        >
                        >




                        __________________________________
                        Do you Yahoo!?
                        New and Improved Yahoo! Mail - 100MB free storage!
                        http://promotions.yahoo.com/new_mail
                      • carlos12155
                        The problem of course is that we can t predict PER differential (or any other differential, in fact). It s not very surprising that almost any linear weights
                        Message 11 of 30 , Jul 22, 2004
                          The problem of course is that we can't predict PER differential (or
                          any other differential, in fact). It's not very surprising that almost
                          any linear weights differentials give such strong correlations if
                          simple points diff. already produces a similar result. The key
                          question is ¿how team dependent is PER diff? ¿How stable is it from
                          year to year?.

                          Individual PER is reasonably stable and relatively team independent,
                          so we can make projections based on that. I've been trying to predict
                          team offensive efficiency based on linear ratings and I found a 96%
                          correlation between team offensive linear rating and offensive
                          efficiency (not high enough for me).

                          --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                          wrote:
                          > Each of those points in favor of PER are true of taking the
                          > differential from other linear weight measures such as the one I
                          > mentioned. Again, why use a complex system such as PER, when taking
                          > the differential from a more simply calculated method yields similar
                          > results?
                          >
                          > --- In APBR_analysis@yahoogroups.com, "mrintp2000" <shzys@n...> wrote:
                          > > Because the Team-PER-diff is a sum of the Player PER-Diff's. This
                          > > correlation shows that there is no synergistic effect whatsoever,
                          > > again-it's a simple sum. There are many important implications of
                          > > this, here are a few
                          > > 1. No position on the floor is more important than any other
                          > > 2. You don't need a top ten player to have a title contending team
                          > > 3. The tremendous value of a superstar is effectively quantified (KG
                          > > +16 for example is worth more than 3X as much as say Kenyon Martin +5)
                          > >
                          > > Back to the PER. First, the PER is the only system I've seen that
                          > > stands up to subjective analysis. It had MJ rated #1 all those years,
                          > > and then Shaq #1 for 5 years straight. Maybe you guys disagree?
                          > >
                          > > Second, the PER allows players to contribute in a multitude of ways
                          > > yet still be considered effective players. Third, the PER-diff
                          > > accurately accounts for defense. Well with maybe a 25% "fudge factor"
                          > > when the defensive assingment might be mismatched:(
                          > >
                          > > Again, the real value here is we can take a very effective rating
                          > > system for individual players, and simply add that up to create an
                          > > equally effective rating for teams.
                          > >
                          > > --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                          > > wrote:
                          > > > I'm not knocking PER differential at all. My point is that by taking
                          > > > the team differential from a very simple linear system -- pts +
                          reb +
                          > > > ast + stl + blk - missed fga - missed fta - to - pf -- I get the
                          same
                          > > > 97% correlation with winning percentage. Why go through all the
                          > > > complex calculations necessary for PER if we can get the same result
                          > > > using a simpler system?
                          > > >
                          > > > --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
                          > > > > --- In APBR_analysis@yahoogroups.com, "wizardskev"
                          > <kevinbroom@r...>
                          > > > > wrote:
                          > > > > > I'm still waiting for someone to explain why a .97 correlation
                          > with
                          > > > > > PER should be viewed as The Holy Grail, when using a different
                          > > > > linear
                          > > > > > stat method, I got the same correlation. A 97% correlation is
                          > very
                          > > > > > strong, no doubt. But why focus on PER when another (simpler)
                          > > > > method
                          > > > > > gets the same result?
                          > > > >
                          > > > > As has been mentioned, the correlation between team point-
                          > > > > differential is also about .97, so there isn't any improvement
                          when
                          > > > > you factor in rebounds, turnovers, etc.
                          > > > >
                          > > > > My own quasi-linear correlation (not a "differential") is only
                          > > > > about .68 correlation; it improves when I adjust winning % by
                          > > > > Conference (+.025 for the West, -.025 for the East).
                          > > > >
                          > > > > When I use my Sagarin correction, I get a .985 correlation.
                          > > > >
                          > > > > However, I believe there's some merit to the free-standing
                          > nature of
                          > > > > the PER-diff -- if you have that data. Since I like to evaluate
                          > > > > seasons from the distant past, I'm determined to use "existing
                          > > > > stats" for my formulae.
                          > > > >
                          > > > >
                          > > > > And thanks (mrintp) for the tip on Excel. I didn't have to
                          > > > > download; under Tools, there's Add-Ins; then just check Analysis
                          > > > > ToolPak, and you're in business. (Excel 2000)
                        • Joe Francis
                          mrintp2000 wrote: Team PER-differential(P-diff) Regular season winning % Golden State -2.5 0.451 Can yuo remind me how you computed the per diff for a team?
                          Message 12 of 30 , Jul 22, 2004
                            mrintp2000 wrote:
                            Team            PER-differential(P-diff)            Regular season winning %
                            Golden State              -2.5                             0.451
                            Can yuo remind me how you computed the per diff for a team?  When I computed the per diff for GS a while back I did it by multiplying player per by % of minutes played, and summing.  With that I got a positve per dif for GS (3+ as I recall). 

                            thanks,
                            -joe  (aka floppymoose)
                          • Joe Francis
                            mrintp2000 wrote: The team PER-diff is calculated by adding up the differentials at each position. Those differentials are created by the players who play at
                            Message 13 of 30 , Jul 22, 2004
                              mrintp2000 wrote:
                              The team PER-diff is calculated by adding up the differentials at each
                              position. Those differentials are created by the players who play at
                              those respective positions. It really is that simple!

                              So that
                              PG +5
                              SG 0
                              SF 0
                              PF +5
                              C  0

                              Adds up to a team PER-diff of +10. If each position was a +2, that
                              would also add up to +10. The beauty of all this is that it does not
                              matter how you get to +10, just that you get there.

                              Apologies, I missed this before.  Why is this method preferred to a method that rates by minutes spent on the floor?

                              thanks,
                              -joe
                            • mrintp2000
                              the PER is the only system I ve seen that stands up to subjective analysis. ... reb + ... same ... when
                              Message 14 of 30 , Jul 22, 2004
                                "the PER is the only system I've seen that stands up to subjective
                                analysis."

                                --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                                wrote:
                                > Each of those points in favor of PER are true of taking the
                                > differential from other linear weight measures such as the one I
                                > mentioned. Again, why use a complex system such as PER, when taking
                                > the differential from a more simply calculated method yields similar
                                > results?
                                >
                                > --- In APBR_analysis@yahoogroups.com, "mrintp2000" <shzys@n...> wrote:
                                > > Because the Team-PER-diff is a sum of the Player PER-Diff's. This
                                > > correlation shows that there is no synergistic effect whatsoever,
                                > > again-it's a simple sum. There are many important implications of
                                > > this, here are a few
                                > > 1. No position on the floor is more important than any other
                                > > 2. You don't need a top ten player to have a title contending team
                                > > 3. The tremendous value of a superstar is effectively quantified (KG
                                > > +16 for example is worth more than 3X as much as say Kenyon Martin +5)
                                > >
                                > > Back to the PER. First, the PER is the only system I've seen that
                                > > stands up to subjective analysis. It had MJ rated #1 all those years,
                                > > and then Shaq #1 for 5 years straight. Maybe you guys disagree?
                                > >
                                > > Second, the PER allows players to contribute in a multitude of ways
                                > > yet still be considered effective players. Third, the PER-diff
                                > > accurately accounts for defense. Well with maybe a 25% "fudge factor"
                                > > when the defensive assingment might be mismatched:(
                                > >
                                > > Again, the real value here is we can take a very effective rating
                                > > system for individual players, and simply add that up to create an
                                > > equally effective rating for teams.
                                > >
                                > > --- In APBR_analysis@yahoogroups.com, "wizardskev" <kevinbroom@r...>
                                > > wrote:
                                > > > I'm not knocking PER differential at all. My point is that by taking
                                > > > the team differential from a very simple linear system -- pts +
                                reb +
                                > > > ast + stl + blk - missed fga - missed fta - to - pf -- I get the
                                same
                                > > > 97% correlation with winning percentage. Why go through all the
                                > > > complex calculations necessary for PER if we can get the same result
                                > > > using a simpler system?
                                > > >
                                > > > --- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
                                > > > > --- In APBR_analysis@yahoogroups.com, "wizardskev"
                                > <kevinbroom@r...>
                                > > > > wrote:
                                > > > > > I'm still waiting for someone to explain why a .97 correlation
                                > with
                                > > > > > PER should be viewed as The Holy Grail, when using a different
                                > > > > linear
                                > > > > > stat method, I got the same correlation. A 97% correlation is
                                > very
                                > > > > > strong, no doubt. But why focus on PER when another (simpler)
                                > > > > method
                                > > > > > gets the same result?
                                > > > >
                                > > > > As has been mentioned, the correlation between team point-
                                > > > > differential is also about .97, so there isn't any improvement
                                when
                                > > > > you factor in rebounds, turnovers, etc.
                                > > > >
                                > > > > My own quasi-linear correlation (not a "differential") is only
                                > > > > about .68 correlation; it improves when I adjust winning % by
                                > > > > Conference (+.025 for the West, -.025 for the East).
                                > > > >
                                > > > > When I use my Sagarin correction, I get a .985 correlation.
                                > > > >
                                > > > > However, I believe there's some merit to the free-standing
                                > nature of
                                > > > > the PER-diff -- if you have that data. Since I like to evaluate
                                > > > > seasons from the distant past, I'm determined to use "existing
                                > > > > stats" for my formulae.
                                > > > >
                                > > > >
                                > > > > And thanks (mrintp) for the tip on Excel. I didn't have to
                                > > > > download; under Tools, there's Add-Ins; then just check Analysis
                                > > > > ToolPak, and you're in business. (Excel 2000)
                              • mrintp2000
                                I remember having this very discussion about Tayshaun Prince before the NBA finals. I contended that his opposing PER showed that he was a great defender. Most
                                Message 15 of 30 , Jul 22, 2004
                                  I remember having this very discussion about Tayshaun Prince before
                                  the NBA finals. I contended that his opposing PER showed that he was a
                                  great defender. Most people wanted to give the credit to Ben Wallace.
                                  My response was "why not just give the credit for Ben's defense to
                                  Tayshaun?" That wouldn't make sense either.

                                  I believe Roland did a study on the correlation between success at an
                                  individual position and winning percentage. He found no higher
                                  correlation with one position over another.


                                  --- In APBR_analysis@yahoogroups.com, Gabe Farkas <gabefark@y...> wrote:
                                  > I don't necessarily agree with your point #1. Could
                                  > you speculate that Ben Wallace's imposing defensive
                                  > presence in the middle (a) discouraged opposing guards
                                  > from taking the ball into the lane more, thus forcing
                                  > them to settle for outside shots, and (b) ed to more
                                  > blocks when these players did decide to drive the
                                  > ball?
                                  >
                                  > Thus, those PGs, SGs and SFs had slightly lower PERs,
                                  > and consequently Chauncey's, Rip's and Tayshaun's
                                  > PER-diffs were slightly inflated?
                                  >
                                  > I would be interested to see a correlative analysis
                                  > between the PER-diff of every position vs the average
                                  > of the other 4. In other words, look at the
                                  > statistical correlation of the PER-diff of the C vs
                                  > (PG+SG+SF+PF)/4, and then the same with PF vs
                                  > (PG+SG+SF+C) and so on and so forth for all 5
                                  > positions.
                                  >
                                  >
                                  > --- mrintp2000 <shzys@n...> wrote:
                                  > > Because the Team-PER-diff is a sum of the Player
                                  > > PER-Diff's. This
                                  > > correlation shows that there is no synergistic
                                  > > effect whatsoever,
                                  > > again-it's a simple sum. There are many important
                                  > > implications of
                                  > > this, here are a few
                                  > > 1. No position on the floor is more important than
                                  > > any other
                                  > > 2. You don't need a top ten player to have a title
                                  > > contending team
                                  > > 3. The tremendous value of a superstar is
                                  > > effectively quantified (KG
                                  > > +16 for example is worth more than 3X as much as say
                                  > > Kenyon Martin +5)
                                  > >
                                  > > Back to the PER. First, the PER is the only system
                                  > > I've seen that
                                  > > stands up to subjective analysis. It had MJ rated #1
                                  > > all those years,
                                  > > and then Shaq #1 for 5 years straight. Maybe you
                                  > > guys disagree?
                                  > >
                                  > > Second, the PER allows players to contribute in a
                                  > > multitude of ways
                                  > > yet still be considered effective players. Third,
                                  > > the PER-diff
                                  > > accurately accounts for defense. Well with maybe a
                                  > > 25% "fudge factor"
                                  > > when the defensive assingment might be mismatched:(
                                  > >
                                  > > Again, the real value here is we can take a very
                                  > > effective rating
                                  > > system for individual players, and simply add that
                                  > > up to create an
                                  > > equally effective rating for teams.
                                  > >
                                  > > --- In APBR_analysis@yahoogroups.com, "wizardskev"
                                  > > <kevinbroom@r...>
                                  > > wrote:
                                  > > > I'm not knocking PER differential at all. My point
                                  > > is that by taking
                                  > > > the team differential from a very simple linear
                                  > > system -- pts + reb +
                                  > > > ast + stl + blk - missed fga - missed fta - to -
                                  > > pf -- I get the same
                                  > > > 97% correlation with winning percentage. Why go
                                  > > through all the
                                  > > > complex calculations necessary for PER if we can
                                  > > get the same result
                                  > > > using a simpler system?
                                  > > >
                                  > > > --- In APBR_analysis@yahoogroups.com, "Mike G"
                                  > > <msg_53@h...> wrote:
                                  > > > > --- In APBR_analysis@yahoogroups.com,
                                  > > "wizardskev" <kevinbroom@r...>
                                  > > > > wrote:
                                  > > > > > I'm still waiting for someone to explain why a
                                  > > .97 correlation with
                                  > > > > > PER should be viewed as The Holy Grail, when
                                  > > using a different
                                  > > > > linear
                                  > > > > > stat method, I got the same correlation. A
                                  > > 97% correlation is very
                                  > > > > > strong, no doubt. But why focus on PER when
                                  > > another (simpler)
                                  > > > > method
                                  > > > > > gets the same result?
                                  > > > >
                                  > > > > As has been mentioned, the correlation between
                                  > > team point-
                                  > > > > differential is also about .97, so there isn't
                                  > > any improvement when
                                  > > > > you factor in rebounds, turnovers, etc.
                                  > > > >
                                  > > > > My own quasi-linear correlation (not a
                                  > > "differential") is only
                                  > > > > about .68 correlation; it improves when I adjust
                                  > > winning % by
                                  > > > > Conference (+.025 for the West, -.025 for the
                                  > > East).
                                  > > > >
                                  > > > > When I use my Sagarin correction, I get a .985
                                  > > correlation.
                                  > > > >
                                  > > > > However, I believe there's some merit to the
                                  > > free-standing nature of
                                  > > > > the PER-diff -- if you have that data. Since I
                                  > > like to evaluate
                                  > > > > seasons from the distant past, I'm determined to
                                  > > use "existing
                                  > > > > stats" for my formulae.
                                  > > > >
                                  > > > >
                                  > > > > And thanks (mrintp) for the tip on Excel. I
                                  > > didn't have to
                                  > > > > download; under Tools, there's Add-Ins; then
                                  > > just check Analysis
                                  > > > > ToolPak, and you're in business. (Excel 2000)
                                  > >
                                  > >
                                  >
                                  >
                                  >
                                  >
                                  > __________________________________
                                  > Do you Yahoo!?
                                  > New and Improved Yahoo! Mail - 100MB free storage!
                                  > http://promotions.yahoo.com/new_mail
                                • mrintp2000
                                  Hey Floppy, how are things? Are the sheeple distraught over the Fish signing/disaster of an albatross? Here s the link for the W s
                                  Message 16 of 30 , Jul 22, 2004
                                    Hey Floppy, how are things? Are the sheeple distraught over the Fish
                                    signing/disaster of an albatross?

                                    Here's the link for the W's
                                    http://www.82games.com/0304GSW5.HTM

                                    shawns

                                    --- In APBR_analysis@yahoogroups.com, Joe Francis <jfrancis@f...> wrote:
                                    >
                                  • mrintp2000
                                    I m not sure I understand your question.
                                    Message 17 of 30 , Jul 22, 2004
                                      I'm not sure I understand your question.

                                      --- In APBR_analysis@yahoogroups.com, Joe Francis <jfrancis@f...> wrote:
                                      >
                                    • Gabe Farkas
                                      2 thoughts: 1) Ben was known for his defense before Tayshaun got there, and i d wager that his PER would indicate just that. 2) i wasn t making any point about
                                      Message 18 of 30 , Jul 22, 2004
                                        2 thoughts:

                                        1) Ben was known for his defense before Tayshaun got
                                        there, and i'd wager that his PER would indicate just
                                        that.

                                        2) i wasn't making any point about team success
                                        correlating to individual positional achievement.
                                        rather, i was suggesting that success at one position
                                        might have something to do with the success of one's
                                        teammates.


                                        --- mrintp2000 <shzys@...> wrote:
                                        > I remember having this very discussion about
                                        > Tayshaun Prince before
                                        > the NBA finals. I contended that his opposing PER
                                        > showed that he was a
                                        > great defender. Most people wanted to give the
                                        > credit to Ben Wallace.
                                        > My response was "why not just give the credit for
                                        > Ben's defense to
                                        > Tayshaun?" That wouldn't make sense either.
                                        >
                                        > I believe Roland did a study on the correlation
                                        > between success at an
                                        > individual position and winning percentage. He found
                                        > no higher
                                        > correlation with one position over another.
                                        >
                                        >
                                        > --- In APBR_analysis@yahoogroups.com, Gabe Farkas
                                        > <gabefark@y...> wrote:
                                        > > I don't necessarily agree with your point #1.
                                        > Could
                                        > > you speculate that Ben Wallace's imposing
                                        > defensive
                                        > > presence in the middle (a) discouraged opposing
                                        > guards
                                        > > from taking the ball into the lane more, thus
                                        > forcing
                                        > > them to settle for outside shots, and (b) ed to
                                        > more
                                        > > blocks when these players did decide to drive the
                                        > > ball?
                                        > >
                                        > > Thus, those PGs, SGs and SFs had slightly lower
                                        > PERs,
                                        > > and consequently Chauncey's, Rip's and Tayshaun's
                                        > > PER-diffs were slightly inflated?
                                        > >
                                        > > I would be interested to see a correlative
                                        > analysis
                                        > > between the PER-diff of every position vs the
                                        > average
                                        > > of the other 4. In other words, look at the
                                        > > statistical correlation of the PER-diff of the C
                                        > vs
                                        > > (PG+SG+SF+PF)/4, and then the same with PF vs
                                        > > (PG+SG+SF+C) and so on and so forth for all 5
                                        > > positions.
                                        > >
                                        > >
                                        > > --- mrintp2000 <shzys@n...> wrote:
                                        > > > Because the Team-PER-diff is a sum of the Player
                                        > > > PER-Diff's. This
                                        > > > correlation shows that there is no synergistic
                                        > > > effect whatsoever,
                                        > > > again-it's a simple sum. There are many
                                        > important
                                        > > > implications of
                                        > > > this, here are a few
                                        > > > 1. No position on the floor is more important
                                        > than
                                        > > > any other
                                        > > > 2. You don't need a top ten player to have a
                                        > title
                                        > > > contending team
                                        > > > 3. The tremendous value of a superstar is
                                        > > > effectively quantified (KG
                                        > > > +16 for example is worth more than 3X as much as
                                        > say
                                        > > > Kenyon Martin +5)
                                        > > >
                                        > > > Back to the PER. First, the PER is the only
                                        > system
                                        > > > I've seen that
                                        > > > stands up to subjective analysis. It had MJ
                                        > rated #1
                                        > > > all those years,
                                        > > > and then Shaq #1 for 5 years straight. Maybe
                                        > you
                                        > > > guys disagree?
                                        > > >
                                        > > > Second, the PER allows players to contribute in
                                        > a
                                        > > > multitude of ways
                                        > > > yet still be considered effective players.
                                        > Third,
                                        > > > the PER-diff
                                        > > > accurately accounts for defense. Well with maybe
                                        > a
                                        > > > 25% "fudge factor"
                                        > > > when the defensive assingment might be
                                        > mismatched:(
                                        > > >
                                        > > > Again, the real value here is we can take a very
                                        > > > effective rating
                                        > > > system for individual players, and simply add
                                        > that
                                        > > > up to create an
                                        > > > equally effective rating for teams.
                                        > > >
                                        > > > --- In APBR_analysis@yahoogroups.com,
                                        > "wizardskev"
                                        > > > <kevinbroom@r...>
                                        > > > wrote:
                                        > > > > I'm not knocking PER differential at all. My
                                        > point
                                        > > > is that by taking
                                        > > > > the team differential from a very simple
                                        > linear
                                        > > > system -- pts + reb +
                                        > > > > ast + stl + blk - missed fga - missed fta - to
                                        > -
                                        > > > pf -- I get the same
                                        > > > > 97% correlation with winning percentage. Why
                                        > go
                                        > > > through all the
                                        > > > > complex calculations necessary for PER if we
                                        > can
                                        > > > get the same result
                                        > > > > using a simpler system?
                                        > > > >
                                        > > > > --- In APBR_analysis@yahoogroups.com, "Mike G"
                                        > > > <msg_53@h...> wrote:
                                        > > > > > --- In APBR_analysis@yahoogroups.com,
                                        > > > "wizardskev" <kevinbroom@r...>
                                        > > > > > wrote:
                                        > > > > > > I'm still waiting for someone to explain
                                        > why a
                                        > > > .97 correlation with
                                        > > > > > > PER should be viewed as The Holy Grail,
                                        > when
                                        > > > using a different
                                        > > > > > linear
                                        > > > > > > stat method, I got the same correlation.
                                        > A
                                        > > > 97% correlation is very
                                        > > > > > > strong, no doubt. But why focus on PER
                                        > when
                                        > > > another (simpler)
                                        > > > > > method
                                        > > > > > > gets the same result?
                                        > > > > >
                                        > > > > > As has been mentioned, the correlation
                                        > between
                                        > > > team point-
                                        > > > > > differential is also about .97, so there
                                        > isn't
                                        > > > any improvement when
                                        > > > > > you factor in rebounds, turnovers, etc.
                                        > > > > >
                                        > > > > > My own quasi-linear correlation (not a
                                        > > > "differential") is only
                                        > > > > > about .68 correlation; it improves when I
                                        > adjust
                                        > > > winning % by
                                        > > > > > Conference (+.025 for the West, -.025 for
                                        > the
                                        > > > East).
                                        > > > > >
                                        > > > > > When I use my Sagarin correction, I get a
                                        > .985
                                        > > > correlation.
                                        > > > > >
                                        > > > > > However, I believe there's some merit to the
                                        > > > free-standing nature of
                                        > > > > > the PER-diff -- if you have that data.
                                        > Since I
                                        > > > like to evaluate
                                        > > > > > seasons from the distant past, I'm
                                        > determined to
                                        > > > use "existing
                                        > > > > > stats" for my formulae.
                                        > > > > >
                                        > > > > >
                                        > > > > > And thanks (mrintp) for the tip on Excel. I
                                        > > > didn't have to
                                        > > > > > download; under Tools, there's Add-Ins; then
                                        > > > just check Analysis
                                        > > > > > ToolPak, and you're in business. (Excel
                                        > 2000)
                                        > > >
                                        > > >
                                        > >
                                        > >
                                        > >
                                        > >
                                        > > __________________________________
                                        > > Do you Yahoo!?
                                        > > New and Improved Yahoo! Mail - 100MB free storage!
                                        > > http://promotions.yahoo.com/new_mail
                                        >
                                        >


                                        __________________________________________________
                                        Do You Yahoo!?
                                        Tired of spam? Yahoo! Mail has the best spam protection around
                                        http://mail.yahoo.com
                                      Your message has been successfully submitted and would be delivered to recipients shortly.