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

Re: (PER differential)

Expand Messages
  • tajallie@hotmail.com
    ... I don t think anyone would argue the team PER-Diff will have a high correlation to winning%. The arguement is that a players PER and PER- Diff are
    Message 1 of 23 , May 4, 2004
    • 0 Attachment
      >
      > I believe the effect is much less than what you may think. I submit
      > the high correlation of PER-diff to winning%, as well as the high
      > correlation of PER-diff to a positive +/- as evidence.

      I don't think anyone would argue the team PER-Diff will have a high
      correlation to winning%. The arguement is that a players PER and PER-
      Diff are transferable to any situation. Personally, I think it is
      self evident from watching and playing basketball, that the context
      of who you play with matters . . . and yes I believe Magic, Bird,
      Walton, MJ, Russell all made players around them much better. If you
      watched Parish in Golden State and the first year he played in Boston
      you see a pretty big difference (and I believe the stat would bear it
      out). Do I think it was bcs Parrish got better or bcs he played with
      Larry Bird instead of Purvis Short? I'll take Bird for $100.

      Didn't someone run an anlysis of PER pre and post trade within the
      season earlier this year.
    • bchaikin@aol.com
      I like your work and think it adds insight. However, if I was to guess one element that should benefit from team chemistry it would be EFG% . . .I would
      Message 2 of 23 , May 4, 2004
      • 0 Attachment

        I like your work and think it adds insight. However, if I was to guess one element that should benefit from "team chemistry" it would be EFG% . . .I would guess stats on the Kings and maybe Utah would bear that out, but secondary, I wouldn't be shocked to see ST, BS, turnovers  and other stats all be dependant on other team players. Just the other day I witnessed Boykins get credited for two bad passes (on 82games.com) bcs Nene failed to roll to the basket. Would Karl have rolled enabling Earl to complete the pass (or even Camby for that matter), absolutely.

        when i think of team chemistry, i think of how well players play together, and to me anyway that means who is getting what touches - are players satisfied with their roles on the team, with how often they are seeing the ball on offense, have they had to any sacrifice parts of their "normal" game (touches, shot attempted, etc) because of the efforts of others. my point is that if miller was the starting point guard on two different teams in two consecutive years, played about the same minutes, yet had the same touches/min, and shot, passed, got fouled, and turned the ball over per ball possession just as often per ball possession both years, to me that shows that odds are his role on both teams was pretty much the same...

        if his role was pretty much the same i'm guessing his shot selection was about the same too (and i could be dead wrong here), so the question would then be why the large difference in FG% from clipperland to the nuggets. either he simply shot much worse on a similar shot selection over the course of the season, or he shot worse because he was taking harder shots (or less easy layups, short shots). the fact that he got fouled about the same per ball possession both seasons tells me he was more than likely driving to the basket just about as often (or else you have to assume that one season he drove much less but got fouled just as often on jumpers, which i would tend to doubt)....

        wonder if there is any info from 82games.com that could help us here, like jump, close, dunk, and tips shots (percentage-wise) from the two seasons for miller...

        bob chaikin
        bchaikin@...

















      • mrintp2000
        ... here, like jump, close, dunk, and tips shots (percentage-wise) from the two seasons for miller... Andre Miller this season Jump 62% .376 Close 36%
        Message 3 of 23 , May 4, 2004
        • 0 Attachment
          --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
          >
          >"wonder if there is any info from 82games.com that could help us
          here, like jump, close, dunk, and tips shots (percentage-wise) from
          the two seasons for miller..."

          Andre Miller this season

          Jump
          62% .376
          Close
          36% .607
          Dunk
          1% .800
          Tips
          1% .714
          Inside
          38% .612

          Last season

          Jump
          74% .369
          Close
          26% .557
          Dunk
          0% .667
          Tips
          0% .667
          Inside
          26% .560

          So his jumper stayed about the same, the difference was he took the
          ball to the hoop more this year. This could be attributed to more open
          court play with Denver, or maybe Miller just played harder? The fact
          that he dunked and got a few tips may indicate that he was healthier
          this season? The rest of his stats look only slightly better with
          Denver. Based on his PER-diff and +/-, I think Miller was the MVP of
          that team this year. His strengths are rebounding and getting to the
          FT line.
        • mrintp2000
          ... I m glad:) The arguement is that a players PER and PER- ... No doubt and we will never be able to completely remove context, unless the game becomes
          Message 4 of 23 , May 4, 2004
          • 0 Attachment
            --- In APBR_analysis@yahoogroups.com, tajallie@h... wrote:
            > >

            > "I don't think anyone would argue the team PER-Diff will have a high
            > correlation to winning%."

            I'm glad:)


            "The arguement is that a players PER and PER-
            > Diff are transferable to any situation."

            No doubt and we will never be able to completely remove context,
            unless the game becomes one-on-one. The best we can do is look at how
            guys perform in head-to-head matchups over the course of a season. The
            alternative, using team stats like +/-, is useful in a team context.

            "Personally, I think it is
            > self evident from watching and playing basketball, that the context
            > of who you play with matters . . . and yes I believe Magic, Bird,
            > Walton, MJ, Russell all made players around them much better. If you
            > watched Parish in Golden State and the first year he played in Boston
            > you see a pretty big difference (and I believe the stat would bear it
            > out). Do I think it was bcs Parrish got better or bcs he played with
            > Larry Bird instead of Purvis Short? I'll take Bird for $100."

            I don't think the stats would support your position on this, but I
            would like to see the research that's been done.
            >
            > "Didn't someone run an anlysis of PER pre and post trade within the
            > season earlier this year."

            I would really like to know how stable PER and PER-diff is from
            year-to-year?
          • Mike G
            ... They pay all that money and don t know what they ve got? ... play ... Ideally, your 5 best players play 5 different positions. I m not sure how you d
            Message 5 of 23 , May 5, 2004
            • 0 Attachment
              --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
              > the sonics coaches didn't realize what they had sitting on
              > the bench (daniels).

              They pay all that money and don't know what they've got?


              > there are far far more instances of teams succeeding when they
              play
              > their best players...

              Ideally, your 5 best players play 5 different positions. I'm not
              sure how you'd count up the teams that managed this, but I suspect
              it would require some luck on top of careful planning.

              >[..run a "Sim" season with the Sonics going bigger minutes with
              > their most productive/efficient players; and reduce the minutes of
              those
              > unproductive big men -- ].
              >
              > send me the allocated minutes you'd like to see and i'll post the
              results...

              min player
              36 Allen
              32 Lewis
              32 Barry
              28 Daniels
              20 Murray
              20 Radmanovic
              16 Booth
              16 James
              12 Ridnour
              12 Sesay
              8 Potapenko
              4 Evans

              I've taken the 3rd-power of their per-minute production, to get
              these ratios. Actually, Allen should play 48 minutes (and a few
              others, fewer), but he only went 37, and I don't know if you have a
              fatigue factor...
              And then I normalized to 240 player-minutes.

              4 of the top 5 are guards. And Frahm wasn't one of the top 12
              minutes-guys, so he's out.

              Hope you appreciate the 4-minute intervals.
            • bchaikin@aol.com
              ... They pay all that money and don t know what they ve got? ... play ... Ideally, your 5 best players play 5 different positions.  I m not sure how you d
              Message 6 of 23 , May 5, 2004
              • 0 Attachment

                > the sonics coaches didn't realize what they had sitting on
                > the bench (daniels).

                They pay all that money and don't know what they've got?


                > there are far far more instances of teams succeeding when they
                play
                > their best players...

                Ideally, your 5 best players play 5 different positions.  I'm not
                sure how you'd count up the teams that managed this, but I suspect
                it would require some luck on top of careful planning.

                >[..run a "Sim" season with the Sonics going bigger minutes with
                > their most productive/efficient players; and reduce the minutes of
                those
                > unproductive big men -- ].
                >
                > send me the allocated minutes you'd like to see and i'll post the
                results...

                min     player
                36    Allen
                32    Lewis
                32    Barry
                28    Daniels
                20    Murray
                20    Radmanovic
                16    Booth
                16    James
                12    Ridnour
                12    Sesay
                8    Potapenko
                4    Evans

                I've taken the 3rd-power of their per-minute production, to get these ratios.  Actually, Allen should play 48 minutes (and a few others, fewer), but he only went 37, and I don't know if you have a fatigue factor...And then I normalized to 240 player-minutes.

                4 of the top 5 are guards.  And Frahm wasn't one of the top 12 minutes-guys, so he's out.

                Hope you appreciate the 4-minute intervals.

                the sonics went 37-45 in 03-04, and their point differential per game was that of a team that should have gone 39-43. simulation using a default substitution pattern of all of the players' totals minutes each divided by 82 games (rounded to the closest multiple of 4) results in an average W-L record per 82 game season of 38-44 to 37.5-44.5...
                                               default
                                   actual      sim         sim         sim
                                    37-45    38-44       40-42     48.5-33.5
                                   min/82 37.5-44.5  40.5-41.5    49-33
                c.booth           15         16           16            36
                r.evans            16         16            4            36
                r.allen              26         28          36            36
                b.barry            22         24           32            36
                a.daniels         18         16           28            36
                j.james            12         12           16            12
                v.radmanovic    28         28           20            12
                r.lewis             36         36           32            36
                r.murray          25         24           20              0
                l.ridnour           14        16           12              0
                potapenko       17         16            8              0
                sesay               7         12           12             0

                using your proposed substitution pattern, the sonics improve by 2.5 games per 82 game season to 40-40.5 wins per 82 game season. but for example the last substitution pattern listed if simulated results in an average W-L record of 48.5-33.5 to 49-33, about 11.5-12 games better than their actual real life W-L record...

                bob chaikin
                bchaikin@...




















              • Mike G
                ... game was ... default ... divided by 82 games ... record per 82 ... Thanks for the work, Bob. Let me know if I m not interpreting this properly: The 2nd
                Message 7 of 23 , May 6, 2004
                • 0 Attachment
                  --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
                  >
                  > the sonics went 37-45 in 03-04, and their point differential per
                  game was
                  > that of a team that should have gone 39-43. simulation using a
                  default
                  > substitution pattern of all of the players' totals minutes each
                  divided by 82 games
                  > (rounded to the closest multiple of 4) results in an average W-L
                  record per 82
                  > game season of 38-44 to 37.5-44.5...
                  > default
                  > actual sim sim sim
                  > 37-45 38-44 40-42 48.5-33.5
                  > min/82 37.5-44.5 40.5-41.5 49-33
                  > c.booth 15 16 16 36
                  > r.evans 16 16 4 36
                  > r.allen 26 28 36 36
                  > b.barry 22 24 32 36
                  > a.daniels 18 16 28 36
                  > j.james 12 12 16 12
                  > v.radmanovic 28 28 20 12
                  > r.lewis 36 36 32 36
                  > r.murray 25 24 20 0
                  > l.ridnour 14 16 12 0
                  > potapenko 17 16 8 0
                  > sesay 7 12 12 0

                  Thanks for the work, Bob.

                  Let me know if I'm not interpreting this properly: The 2nd column
                  represents actual player minutes divided by 82 games (and rounded to
                  the nearest multiple of 4 min); hence, Allen and Barry only play 20-
                  odd minutes, on average.

                  Then Sim 3 takes the MPG I suggested, but allows everyone to play
                  all 82 games? And still the Sonics only improve by a couple of
                  games?

                  Then, in Sim 4, you've given 2 of the least productive players
                  (Booth and Evans) major minutes, and the team gains another 8 wins?

                  Since the effect of granting Allen and Barry the full 82 games of
                  contribution (apparently running 3 guards a good part of the time)
                  is dwarfed by the effect of having (any) 2 big men in the lineup, it
                  looks as if your simulation does indeed recognize the necessity of
                  rebounding and interior defense, even when the players supplying
                  those things don't do much else.

                  It seems if the Sonics had Barry and Allen all year, AND went with
                  their most "productive" players more minutes, the combined effect is
                  almost insignificant (1-2 wins).

                  But having those guys healthy and sticking to a couple of random
                  widebodies makes them a contender?

                  Or is there another interpretation?


                  >
                  > using your proposed substitution pattern, the sonics improve by
                  2.5 games per
                  > 82 game season to 40-40.5 wins per 82 game season. but for example
                  the last
                  > substitution pattern listed if simulated results in an average W-L
                  record of
                  > 48.5-33.5 to 49-33, about 11.5-12 games better than their actual
                  real life W-L
                  > record...
                  >
                  > bob chaikin
                  > bchaikin@b...
                • bchaikin@aol.com
                  ... 82 games ... 82 ... Thanks for the work, Bob. Let me know if I m not interpreting this properly:  The 2nd column represents actual player minutes divided
                  Message 8 of 23 , May 6, 2004
                  • 0 Attachment

                    > the sonics went 37-45 in 03-04, and their point differential per game was
                    > that of a team that should have gone 39-43. simulation using a default
                    > substitution pattern of all of the players' totals minutes each divided by 82 games
                    > (rounded to the closest multiple of 4) results in an average W-L record per 82
                    > game season of 38-44 to 37.5-44.5...
                    >                                default
                    >                  actual      sim         sim         sim
                    >                   37-45    38-44       40-42     48.5-33.5
                    >                  min/82  37.5-44.5  40.5-41.5    49-33
                    > c.booth           15         16           16            36
                    > r.evans           16         16            4            36
                    > r.allen           26         28           36            36
                    > b.barry           22         24           32            36
                    > a.daniels         18         16           28            36
                    > j.james           12         12           16            12
                    > v.radmanovic      28         28           20            12
                    > r.lewis           36         36           32            36
                    > r.murray          25         24           20             0
                    > l.ridnour         14         16           12             0
                    > potapenko         17         16            8             0
                    > sesay              7         12           12             0

                    Thanks for the work, Bob.

                    Let me know if I'm not interpreting this properly:  The 2nd column represents actual player minutes divided by 82 games (and rounded to the nearest multiple of 4 min); hence, Allen and Barry only play 20-odd minutes, on average. 

                    yes - to approximate what actually happened in real life in 03-04, the sim tries to play everyone as close as possible to the same amount of total minutes over the span of 82 games that they actually played in real life. not as precise as possible (what if one guy played 1500 min but only 41 games?) but a close approximation...

                    Then Sim 3 takes the MPG I suggested, but allows everyone to play all 82 games?  And still the Sonics only improve by a couple of games?

                    2-3 games, and that could be enough to make the playoffs. yes - i believe they shot a little better, had a slightly better def FG%, but rebounded significantly worse. also remember you had murray and ridnour playing a combined 32 min/g, and potapenko and sesay another 20 min/g...

                    Then, in Sim 4, you've given 2 of the least productive players (Booth and Evans) major minutes, and the team gains another 8 wins?

                    evidently they are providing something to the effort, that when added to the play of players like lewis, allen, barry, and daniels, complements them well enough to win more games...

                    but remember - although in both sub patterns allen plays 36 min/g, in the 3rd one barry played 32 min/g yet 36 min/g in the last pattern, and daniels just 28 min/g in the 3rd yet 36 min/g in the last one...

                    also part of the reason is how i have the players rated defensively. i rate players on a scale of -5.0% to +5.0%, meaning if a defender is rated at -5.0% (excellent) he will decrease a 50% shooter to 45% on average, and a +5.0% defender (worst) would increase his FG% to 55% (example ron artest is a -5.0% and peja stojakovic is a +5.0%). for the sonics here is how i have them rated, based on 82games.com data and some other small factors...

                    -1.0% - booth
                    +4.5% - radmonovic
                    +4.0% - lewis
                    -2.0% - allen
                    0.0% - barry
                    +1.0% - james
                    +3.0% -potapenko
                    +5.0% - evans
                    +2.0% - murray
                    -0.5% - daniels
                    +2.5% - ridnour
                    +3.5% - sesay

                    what 82games.com also has data for but the sim does not model (yet) is how much a defender prevents an offensive player from shooting, or allows him to shoot more or less than, the average player at a specific position. i.e. ron artest holds the players he guards to 15% less FGA than the average SF and someone like larry hughes allows the SGs he defends 20% more FGA than the average SG. could be they are reducing/increasing the touches of the players they guard (i.e. deny their man the ball - artest - or letting him get open more often than others do - hughes - don't know that yet), or they simply play good (or bad) defense on FGAs...

                    Since the effect of granting Allen and Barry the full 82 games of contribution (apparently running 3 guards a good part of the time) is dwarfed by the effect of having (any) 2 big men in the lineup, it looks as if your simulation does indeed recognize the necessity of rebounding and interior defense, even when the players supplying those things don't do much else....

                    not that sure about the interior defense as i have reggie evans rated poorly on defense...

                    It seems if the Sonics had Barry and Allen all year, AND went with their most "productive" players more minutes, the combined effect is almost insignificant (1-2 wins).

                    But having those guys healthy and sticking to a couple of random widebodies makes them a contender? Or is there another interpretation?...

                    not sure than booth and evans are merely "random wisebodies", both have enviable attributes in that booth was a premier shot-blocker last season and evans an excellent rebounder, plus their low touches/min means more touches/min for the efficient scorers allen, barry, and daniels (plus minutes previously given to murray, ridnour, potapenko, and sesay are going elsewhere)...

                    kinda like my previous statements that antoine walker is not only a poor shooter (inefficient scorer) but that he also has high touches/min such that he takes touches away from better more efficient scorers like nowitzki, nash, finley, and jamison, plus his turnovers per ball possession is the highest of anyone on the mavs that played 1000+ minutes. dallas would have been better off giving walker's minutes to jamison, bradley, and others...

                    i didn't sim a ton of different sub patterns for the sonics, just tried the last one above with booth and evans playing alongside the 3 efficient scorers (booth, evans, lewis, allen, barry, and daniels all getting 36 min/g). some of their per game numbers for that 8200 game simulation were:

                    booth 11 pts/g, 8 reb/g, 4 fouls/g, 3 bs/g, 0.5 touches/min (same touches/min as in real life)
                    evans 7 pts/g, 10 reb/g, 4 fouls/g, 0.5 touches/min (0.4 touches/min in real life)
                    lewis 19 pts/g, 6 reb/g, 1.0 touches/min (0.9 touches/min in real life)
                    allen  23 pts/g, 4.5 ast/g, 4.6 reb/g, 1.4 touches/min (about the same touches/min in real life)
                    barry 14 pts/g, 7 ast/g, 4 reb/g, 1.6 touches/min (1.5 touches/min in real life)
                    daniels 15 pts/g, 8 ast/g, 1.7 touches/min (1.6 touches/min in real life)

                    the question is - is it unreasonable to play both booth and evans 36 min/g? i don't think so (lots of team in the playoffs now are going with just a 7 or 8 man rotation). each did commit 300+ fouls in each simulated 82 game season replay playing those minutes, yet the team still won more games...

                    >
                    > using your proposed substitution pattern, the sonics improve by 2.5 games per
                    > 82 game season to 40-40.5 wins per 82 game season. but for example the last
                    > substitution pattern listed if simulated results in an average W-L record of
                    > 48.5-33.5 to 49-33, about 11.5-12 games better than their actual real life W-L
                    > record...

                    bob chaikin
                    bchaikin@...

                  • Kevin Pelton
                    Interesting stuff, Bob, and finally some proof for my belief that Evans needed to play more than he did, because his skills so well complement those of the
                    Message 9 of 23 , May 6, 2004
                    • 0 Attachment
                      Interesting stuff, Bob, and finally some proof for my belief that
                      Evans needed to play more than he did, because his skills so well
                      complement those of the Sonics' stars.

                      I am curious why you rated him as the worst defender on the team --
                      more as an exercise in the theory than as a specific debate on Evans.
                      He's generally considered a solid defender and the Sonics were 2.2
                      points per 100 possessions better on defense with him in the game,
                      per 82games (though that could be explained away by his defensive
                      rebounding).

                      What are you seeing?
                    • bchaikin@aol.com
                      Interesting stuff, Bob, and finally some proof for my belief that Evans needed to play more than he did, because his skills so well complement those of the
                      Message 10 of 23 , May 6, 2004
                      • 0 Attachment
                        Interesting stuff, Bob, and finally some proof for my belief that Evans needed to play more than he did, because his skills so well complement those of the Sonics' stars.

                        I am curious why you rated him as the worst defender on the team -- more as an exercise in the theory than as a specific debate on Evans. He's generally considered a solid defender and the Sonics were 2.2 points per 100 possessions better on defense with him in the game, per 82games (though that could be explained away by his defensive
                        rebounding).

                        What are you seeing?

                        data from 82games.com...

                        remember my model does not (yet) take into account how many FGA above or below the average PF (or C, SF, SG, or PG) a player allows. so even tho evans gave up +5.0%, it could very well be that he gave up less or even significantly less FGM/FGA per min than the avg PF, meaning that even tho evn gave up a high FG% he may not have given up a high number of FGM - i just don't know yet...

                        bob chaikin
                        bchaikin@...













                      • roland_beech
                        Bob s simulator had the Sonics doing better with Evans and Booth getting big minutes... interestingly of the 5man units Seattle did actually use:
                        Message 11 of 23 , May 6, 2004
                        • 0 Attachment
                          Bob's simulator had the Sonics doing better with Evans and Booth
                          getting big minutes...

                          interestingly of the 5man units Seattle did actually use:

                          Barry-Allen-Lewis-Evans-Booth were +29 in 28 minutes together
                          (obviously a tiny sample)

                          and

                          Barry-Allen-Lewis-Evans-Potapenko were +21 in 176 minutes

                          all told, the Evans-Booth player pair was on court together for 234
                          minutes and had a -1 plus/minus (not adjusted by possession
                          difference which probably hurts them since they would more often
                          than not I imagine get taken out when Seattle was about to have the
                          ball)

                          Evans counterpart defense doesn't look good (53% eFG allowed to the
                          opposing PF), but he pulls down 5 more rebounds per 48 minutes so
                          makes up for it there. Evans also commits fewer turnovers per 48
                          than the opposing PF while he's on the court (no surprise since he
                          barely touches the ball on offense as Bob's possession data shows)

                          When we begin charting individual defense more closely there will be
                          numbers looking at how often a defender gets help, etc which could
                          show that a player with "suspect" defensive numbers is actually
                          better than it looks since he is guarding guys one on one without
                          help most of the time, leaving his man to help out on guys driving
                          in from the perimeter who then pass to his now open guy, etc
                        Your message has been successfully submitted and would be delivered to recipients shortly.