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

Season-per-game Similarity tests

Expand Messages
  • Dean Oliver
    Rules for calculating similarity, based on comparison of player-season values of per-game averages: Subtract (from 1000) 1 point for each difference of 1 Game
    Message 1 of 3 , Sep 18, 2001
    • 0 Attachment
      Rules for calculating similarity, based on comparison of
      player-season
      values of per-game averages:

      Subtract (from 1000) 1 point for each difference of 1 Game played and
      differences in per game averages as below:

      Min 0.5
      fgm 0.2
      fga 0.45
      fg% 0.0015
      fg3m 0.04
      fg3a 0.1
      fg3% 0.1
      ftm 0.08
      fta 0.1
      ft% 0.005
      oreb 0.2
      dreb 0.35
      treb 0.5
      ast 0.06
      stl 0.025
      tov 0.04
      blk 0.025
      pf 0.12
      pts 0.38

      I've done a few tests, but the one I wanted to do was on ROY Mike
      Miller:

      Player Team Season SimScore
      lenard,voshon mia 1998 930
      person,wesley pho 1996 926
      person,chuck san 1995 918
      person,chuck san 1996 917
      ferry,danny cle 1996 916
      person,chuck min 1994 912
      ellis,dale sea 1999a 911
      scott,dennis pho 1998 910
      ferry,danny cle 1997 909
      ellis,dale den 1996 908
      majerle,dan cle 1996 905
      person,wesley pho 1997 903
      person,wesley cle 1999a 903
      scott,dennis orl 1991 902
      dickerson,micha hou 1999a 902
      lenard,voshon den 2001 901

      No real stars here, but, as James likes to point out, one of the most
      important things is age. Miller was only 20 last year and all these
      guys above were older. I'd really need to control for age. Dennis
      Scott's '91 season may be about right age-wise (but I don't have my
      encyclopedia in front of me).

      G Min fgm fga fg3m fg3a
      Miller 82 29.2 4.5 10.4 1.8 4.5
      Scott 82 28.5 6.1 14.4 1.5 4.1

      ftm fta oreb treb ast
      1.1 1.6 0.8 4.0 1.7
      1.9 2.5 0.8 2.9 1.6

      stl tov blk pf dq pts
      0.61 1.2 0.23 2.5 0.0 12.0
      0.76 1.5 0.30 2.5 0.0 15.7

      Pretty similar seasons, though Miller didn't do quite as much with
      his
      minutes (might also be a decline in league pace).

      Dean Oliver
      Journal of Basketball Studies
    • harlanzo@yahoo.com
      This is just another note in passing but shouldn t age also be an issue in comparison? It is obviously not relevant if you just want to know a player who
      Message 2 of 3 , Sep 18, 2001
      • 0 Attachment
        This is just another note in passing but shouldn't age also be an
        issue in comparison? It is obviously not relevant if you just want
        to know a player who could fill a similar role to a player on your
        team. However, if the point of your comparison is to figure out the
        upside of your player I think accounting for the age difference is
        the players maybe relevant.


        --- In APBR_analysis@y..., "Dean Oliver" <deano@t...> wrote:
        >
        > Rules for calculating similarity, based on comparison of
        > player-season
        > values of per-game averages:
        >
        > Subtract (from 1000) 1 point for each difference of 1 Game played
        and
        > differences in per game averages as below:
        >
        > Min 0.5
        > fgm 0.2
        > fga 0.45
        > fg% 0.0015
        > fg3m 0.04
        > fg3a 0.1
        > fg3% 0.1
        > ftm 0.08
        > fta 0.1
        > ft% 0.005
        > oreb 0.2
        > dreb 0.35
        > treb 0.5
        > ast 0.06
        > stl 0.025
        > tov 0.04
        > blk 0.025
        > pf 0.12
        > pts 0.38
        >
        > I've done a few tests, but the one I wanted to do was on ROY Mike
        > Miller:
        >
        > Player Team Season SimScore
        > lenard,voshon mia 1998 930
        > person,wesley pho 1996 926
        > person,chuck san 1995 918
        > person,chuck san 1996 917
        > ferry,danny cle 1996 916
        > person,chuck min 1994 912
        > ellis,dale sea 1999a 911
        > scott,dennis pho 1998 910
        > ferry,danny cle 1997 909
        > ellis,dale den 1996 908
        > majerle,dan cle 1996 905
        > person,wesley pho 1997 903
        > person,wesley cle 1999a 903
        > scott,dennis orl 1991 902
        > dickerson,micha hou 1999a 902
        > lenard,voshon den 2001 901
        >
        > No real stars here, but, as James likes to point out, one of the
        most
        > important things is age. Miller was only 20 last year and all
        these
        > guys above were older. I'd really need to control for age. Dennis
        > Scott's '91 season may be about right age-wise (but I don't have my
        > encyclopedia in front of me).
        >
        > G Min fgm fga fg3m fg3a
        > Miller 82 29.2 4.5 10.4 1.8 4.5
        > Scott 82 28.5 6.1 14.4 1.5 4.1
        >
        > ftm fta oreb treb ast
        > 1.1 1.6 0.8 4.0 1.7
        > 1.9 2.5 0.8 2.9 1.6
        >
        > stl tov blk pf dq pts
        > 0.61 1.2 0.23 2.5 0.0 12.0
        > 0.76 1.5 0.30 2.5 0.0 15.7
        >
        > Pretty similar seasons, though Miller didn't do quite as much with
        > his
        > minutes (might also be a decline in league pace).
        >
        > Dean Oliver
        > Journal of Basketball Studies
      • Dean Oliver
        ... the ... Exactly right. I don t have a very complete db of ages/birthdates. Once I do, I would certainly include it in trying to make predictive similarity
        Message 3 of 3 , Sep 18, 2001
        • 0 Attachment
          --- In APBR_analysis@y..., harlanzo@y... wrote:
          > This is just another note in passing but shouldn't age also be an
          > issue in comparison? It is obviously not relevant if you just want
          > to know a player who could fill a similar role to a player on your
          > team. However, if the point of your comparison is to figure out
          the
          > upside of your player I think accounting for the age difference is
          > the players maybe relevant.

          Exactly right. I don't have a very complete db of ages/birthdates.
          Once I do, I would certainly include it in trying to make predictive
          similarity scores. Below, for instance, Dennis Scott was 22 in his
          91 season, 2 years older than Miller -- a pretty important
          difference, I'm thinking. He may still end up the best comparison,
          though, because of the dearth of 20 year olds in the NBA who did
          anything. (Michael Dickerson, by the way, was 23 in 1999.)


          > > scott,dennis orl 1991 902
          > > dickerson,micha hou 1999a 902
          > > lenard,voshon den 2001 901
          > >
          > > No real stars here, but, as James likes to point out, one of the
          > most
          > > important things is age. Miller was only 20 last year and all
          > these
          > > guys above were older. I'd really need to control for age.
          Dennis
          > > Scott's '91 season may be about right age-wise (but I don't have
          my
          > > encyclopedia in front of me).
          > >
          > > G Min fgm fga fg3m fg3a
          > > Miller 82 29.2 4.5 10.4 1.8 4.5
          > > Scott 82 28.5 6.1 14.4 1.5 4.1
          > >
          > > ftm fta oreb treb ast
          > > 1.1 1.6 0.8 4.0 1.7
          > > 1.9 2.5 0.8 2.9 1.6
          > >
          > > stl tov blk pf dq pts
          > > 0.61 1.2 0.23 2.5 0.0 12.0
          > > 0.76 1.5 0.30 2.5 0.0 15.7
          > >
          > > Pretty similar seasons, though Miller didn't do quite as much
          with
          > > his
          > > minutes (might also be a decline in league pace).

          The pace declined from 100 poss/G in '91 to 94 /game last year. The
          league rating went from 105.9 to 101.3. Miller had a tougher context
          to put up similar numbers. That, plus the age difference, suggests
          that Miller should be a little better than Dennis Scott.

          Dean Oliver
          Journal of Basketball Studies
        Your message has been successfully submitted and would be delivered to recipients shortly.