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

Re: NBA win-shares

Expand Messages
  • harlanzo
    win shares is definitely an interesting concept but I wonder if it is a bit forced. to me, the most useful stat is points produced (versus of course points
    Message 1 of 5 , Jun 23, 2002
    • 0 Attachment
      win shares is definitely an interesting concept but I wonder if it is
      a bit forced. to me, the most useful stat is points produced (versus
      of course points allowed on the defensive end). I guess you could
      extroapolate a win loss record from such data but to run a tendex
      rating and then guess wins does not seem to reveal too much.

      --- In APBR_analysis@y..., "mikel_ind" <msg_53@h...> wrote:
      > --- In APBR_analysis@y..., "HoopStudies" <deano@r...> wrote:
      > >
      > > Someone took a shot at it using a TENDEX system:
      > >
      > > http://www.mindspring.com/~lzamuda/winsexplained.htm
      >
      > I actually like this. It's got some iffy assumptions, but it seems
      > to merge the concepts of individual and team success.
      >
      > Looking at his alltime rankings, a lot of the differences with my
      own
      > list relate to a few items: He doesn't include playoffs or ABA
      > stats; and he has a cumulative-totals system that rewards longevity
      > over the brief-but-brilliant career.
      >
      > His version of T#nd#x is pretty crude, as most are. Game pace is
      > determined by opponents' points, but rebound totals are assumed to
      be
      > equivalently related.
      >
      > The convention of standardizing to 1.000 point per possession is
      > maintained (I don't agree), and then he compensates for earlier
      eras
      > very crudely by giving whole decades their own multiplier, like .90
      > for the '60s and .80 for the '50s.
      >
      >
      > I haven't found a clear pattern yet as to whether he overrates high
      > shooting pct or anything else, relative to my system. Here is a
      list
      > of players he ranks higher than I do:
      >
      > Stockton, Kareem, Karl Malone, Unseld, Havlicek, Parish, Wilt,
      Oscar,
      > Moses, Mark Jackson, Cheeks, Magic, Barkley, Russell, Paul Silas,
      > Rodman, Drexler, Payton, Terry Porter, Horace Grant
      >
      >
      > Incidentally, my list is re-ranked to exclude playoffs.
      >
      >
      > Here are players I rank higher (not including guys from the ABA):
      >
      > Duncan, McAdoo, Gervin, Grant Hill, Aguirre, Mourning, Willis Reed,
      > Chambers, Mark Price, Marques Johnson, Bailey Howell, Bernard King,
      > Pettit, Nance, Mutombo, Cliff Robinson, Bellamy
      >
      > And these players who didn't make his top 100, but make mine:
      >
      > Neil Johnston, Webber, McGinnis, Garnett, Mikan, Lovellette,
      Haywood,
      > Richmond, Arizin, Foust, Coleman, Daugherty, Maurice Lucas, Beaty,
      > Hagan, Smits, Iverson
      >
      >
      > I may tinker with this a bit more. Please don't hold me to any of
      > it. Other than his bias to longevity, and my bias toward
      brilliance,
      > does anyone see a pattern? Do I overrate the scorers?
    • HoopStudies
      ... I guess I looked at the numbers for the wins per year, rather than career wins. I don t think there really is a good philosophical way of evaluating
      Message 2 of 5 , Jun 24, 2002
      • 0 Attachment
        --- In APBR_analysis@y..., "mikel_ind" <msg_53@h...> wrote:
        > --- In APBR_analysis@y..., "HoopStudies" <deano@r...> wrote:
        > >
        > > Someone took a shot at it using a TENDEX system:
        > >
        > > http://www.mindspring.com/~lzamuda/winsexplained.htm
        >
        > I actually like this. It's got some iffy assumptions, but it seems
        > to merge the concepts of individual and team success.

        I guess I looked at the numbers for the wins per year, rather than
        career wins. I don't think there really is a good philosophical way
        of evaluating longevity vs peak value. So exactly how you rank
        players is up to the user with this list.

        What I found interesting is that the range of wins/82 games is pretty
        similar to what I have estimated myself. I think Smith's stuff has
        slightly higher max wins than I estimate, which I have no problems
        with. I don't think I've ever seen 20 wins per season on an
        average. Jordan peaked out about that high, with maybe 1 loss per
        season, too.

        I would still protest the lack of losses per season or total losses.
        But that's a weakness with the win-shares approach. Rodman may win 8
        per season (not too far from what I estimated in general), but did he
        contribute 1 loss or 6 losses? It makes a difference in assessing
        his real contribution. At the highest level, it isn't such a big
        deal because players can really not be responsible for more than
        about 20 wins in a season, so wins are very proportional to win%.
        But other guys win% are influenced also by how much playing time they
        get, how much of a role they play in the offense and defense.

        I can't tell whether this technique accounts for defense reasonably
        (the defensive stop stuff, which looks vaguely like what I did 10
        years ago). I think probably not. I think that Joe Dumars' low rate
        suggests that. Hard to say how good his D was, even now, kinda like
        Bruce Bowen, because of so many mixed indicators.

        In terms of bias, I can't tell. I always like to see these things
        applied to teams. He didn't so you can't see whether wins add up.
        You can't see whether there is a bias to fast teams or to offensive
        teams or to teams that have more assists. Those are pretty typical
        biases that I can't read from what he's done.

        DeanO
      • monepeterson
        ... I ve chatted with Sean (that s the feller s name) about his system. The wins do add up for teams as a whole, but only because the system s final team
        Message 3 of 5 , Jul 12 10:06 AM
        • 0 Attachment
          Chiming in late on an old topic:

          > In terms of bias, I can't tell. I always like to see these things
          > applied to teams. He didn't so you can't see whether wins add up.
          > You can't see whether there is a bias to fast teams or to offensive
          > teams or to teams that have more assists. Those are pretty typical
          > biases that I can't read from what he's done.

          I've chatted with Sean (that's the feller's name) about his system.
          The wins do add up for teams as a whole, but only because the
          system's final team adjustment (essentially adjusted tendex points
          per win) ensures that it does.

          The intriguing thing about his system is establishing 60% of league
          level production (tendex/minute) as a "zero" level, something that
          seems, from a glance, to work pretty well. For this type of thing
          anyway.

          Moné
        Your message has been successfully submitted and would be delivered to recipients shortly.