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

Re: the Ruben Patterson effect

Expand Messages
  • Dean Oliver
    ... 02-03 Bob s stuff typically results in similar projections as mine (I also get about a 10 win difference putting Kobe in for Johnson, for example). I go
    Message 1 of 57 , Jul 31, 2004
    • 0 Attachment
      --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
      > with derek anderson 31-51, allen iverson just 31-51 (35-47 using his
      02-03

      Bob's stuff typically results in similar projections as mine (I also
      get about a 10 win difference putting Kobe in for Johnson, for
      example). I go about it through net points or individual win-loss
      records, but the ultimate result Bob puts out of team win-loss record
      is nice.

      I highlight the passage above from his email to highlight the
      difficulty with any of these context-sensitive approaches: Players
      change. AI's 2004 season was a bit worse than his 2003 season, which
      was worse than his 2001 season. Why? Was that predictable? Is it
      age? Is it coaching? Is it injuries (the guy doesn't seem to show
      much pain)?

      Bob's sim shows sensitivity to those kinds of changes. My stuff
      definitely shows sensitivity, but I'd like to think I understand those
      changes better. I knew AI would do worse once Larry Brown left, but
      that doesn't mean I understand it better.

      This is relevant for Q, too. His first 2 seasons in the league were
      much better than his last two. I've been thinking it's
      injury-related, but some have suggested that it's related to playing
      with Corey Maggette. There are reasons to doubt the Maggette
      rationale, but if it is true, he should be better in Phoenix. If the
      injuries have dogged him the last two years, which is true, and, more
      speculatively, he is injury-free in 2005, then perhaps he returns to
      form. I need to look into it more, but my sense is that Q will get a
      bit closer to where he was in 2002, which is a player who is about 4
      wins better than he was last year.

      Finally, Bob mentioned that McGrady's 2003 season was about as good as
      some of MJs best. I can see why he says that, but one big thing would
      keep me from agreeing -- consistency. In MJ's top seasons, he was
      dominant every night. My "game-by-game win-loss record" for MJ was
      typically 60+ wins and about 20 losses, implying that he had a good
      night about 3/4ths of the time. McGrady, on the other hand, had some
      monster nights that made his season total look good, but also had some
      mediocre nights, for a similar total of 47-28 that season.

      >
      > so the question might be why not employ 2 or 3 bowen type players on
      the
      > spurs to increase duncan's touches even more? because there is a
      point (i'm not
      > quite sure exactly where that point is yet) where this logic breaks
      down, and if
      > duncan's teammates cannot be a factor on offense he will get double and
      > triple teamed consistently.

      This relates to the skill curves I put in Chapter 19 of the book.
      Duncan can maintain his efficiency while using a lot of possessions
      (upwards of 30%), but if you have 3 Bowens using 11% of the team's
      possessions, that's 3*9%=27% of the team's possessions that have to be
      used by the other 2 guys. Duncan can't handle more than 14% or so of
      those before getting inefficient. And the Spurs don't have anyone
      else who can even pick up about 9% of the rest.

      >
      > simulation also shows that the league's very best players can -
      individually
      > - make their teams as much as 18-20 games better per 82 game season
      than the
      > worst players (starters) at their same position. play tim duncan 40
      min/g on
      > the spurs and they finish 18-20 games better per 82 game season than
      if duncan
      > was replaced by the likes of antoine walker, walter mccarty, or
      clifford
      > robinson.

      Interestingly, my range is about 14-17, not 18-20, for these players.
      I'm not sure if you can do the comparison with reality to say which
      is "right". Regardless, that's a lot of wins between some of the best
      and worst starters.

      DeanO

      Dean Oliver
      Author, Basketball on Paper
      http://www.basketballonpaper.com
      When basketball teams start playing Moneyball, this is the book
      they'll use!
    • Daniel Flemming
      ... Hash: SHA1 ... No, this is wrong. It s wrong because it favours bench players. Guys like Bowen and Artest (and especially Bowen, because he s older)
      Message 57 of 57 , Aug 5 7:45 AM
      • 0 Attachment
        -----BEGIN PGP SIGNED MESSAGE-----
        Hash: SHA1

        On Wed, 4 Aug 2004 smckibbi@... wrote:

        > Great idea. Perhaps the criteria for deciding which swingman is guarding who
        > would be the team's overall def points per 100 possessions when the swingman
        > is on the court. The one with the lower pts/100 poss is guarding the better
        > offensive player.

        No, this is wrong. It's wrong because it favours bench players. Guys like
        Bowen and Artest (and especially Bowen, because he's older) usually don't
        guard the opposing bench's best player; they take a rest when the other
        side's coach brings his bench players onto the floor.

        You get a guy like Peterson on the Raptors, he's usually on the floor when
        the other side's bench is, and usually not when their starters are. Now,
        he's a good defender, but his stats are going to get enhanced under this
        formula.

        Here's a Modest Proposal: Track the overall offensive production per
        minute or possession of the opposing team's players that a particular
        player defends. Compare that to their offensive production per minute or
        possession while he's on the floor. A negative result means a good
        defender, a positive result means a bad one, zero means average.
        - --
        GnuPG public key available from
        http://homepage.mac.com/danflemming/gnupg_key.html
        -----BEGIN PGP SIGNATURE-----
        Version: GnuPG v1.2.5 (GNU/Linux)
        Comment: Made with pgp4pine 1.76

        iD8DBQFBEkgWER493M5r4PIRAsXkAKCPGIxtXyXhpN6VA+qrO35/hOxNFwCdH890
        eyD+SA/LG+nvJOZv5cfzskc=
        =Fajf
        -----END PGP SIGNATURE-----
      Your message has been successfully submitted and would be delivered to recipients shortly.