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

1626Dilution, balance, and Bob-bashing

Expand Messages
  • Dean Oliver <deano@rawbw.com>
    Dec 31, 2002
      I guess 'tis the season to beat up on BobC. A lot of it happened
      while I was gone, that's for sure.

      I could spend a lot of time going through and showing how
      statistically different the various eras of the NBA are. I think it
      is very clear that pace has changed most significantly, with a huge
      decrease in turnovers. That's a big change from _any_ past period to
      the present, not just '60's to today. I, for one, would not say that
      the '80's are like today either. I would say that the mid-'90's
      aren't even much like today. Hell, last year is beginning to look
      pretty different from this year with the increase in zone D being
      prominent. Statistically, you can find pace, 3pt shooting, and
      turnover rates as good indicators of how the game has changed. Non-
      statistically, the league has trended significantly toward selecting
      athletic tall players who can dribble and defend, rather than tall
      players who can just play the low post. All these 7 footers who play
      on the perimeter didn't exist even in the early '90's. Does that
      mean that Wilt couldn't have played on the perimeter? We don't know -
      - that's the point. Can you simulate something? Sure. Can you
      calibrate it or validate it? No. There is no reality check I can
      think of to do so. Calibration in my mind means having a real
      experiment -- actually playing the 1996 Bulls vs the 1972 Lakers --
      to check a simulation against. Obviously impossible. What we can do
      better is come up with reasonable statistical metrics for what any
      cross-generational simulator should have to live up to. Should they
      meet the average pace of the two generations? The average pace of
      the two teams? I guess I'd find Bob-bashing more productive if we
      can all come up with rules for what cross-generational simulators
      would do, not in specifics (Lakers should beat Bulls), but in
      generalities. For example, any games between the two should
      obviously result in stats where the two teams have the same number of
      possessions. That's a basketball rule. Or just stating the
      assumptions of a cross-generational simulation would help. We assume
      that the Mavericks of 2003 are not allowed to play a zone when
      competing against the Lakers of 1987. Or, more likely, we assume
      that the Mavericks defensive ability is independent of how they play
      defense (which is, of course, wrong but statistically practical).

      Talent dilution and competitive balance

      These are two of the hardest topics around and hot topics in sports
      economics right now. We aren't going to solve them here. Standard
      deviations of various numbers across teams are a valid approach that
      gets at one way of looking at competitive balance, but is definitely
      not the whole story. (I've looked at how long it takes for teams to
      go from good to bad, bad to average, etc., as well as standard
      deviations of win-loss records, gini coefficients, how many similar
      playoff teams from one season to the next, blah blah blah.) What
      teams have the greatest chance of winning and how great those chances
      are -- that's important. I personally think that knowing the Vegas
      line on who would win the title -- at the start of the season and at
      the start of the post-season -- then who won the title is getting
      better at the underlying concern of competitive balance questions,
      which is that predetermined outcomes of sports are bad. Purely
      random outcomes may also be bad, but things that are almost
      guarantees -- the Lakers winning the title in 2001 -- is not healthy
      over a sustained period. That's why gambling is bad – it sets up
      situations where guys intentionally losing is possible and results
      become more predetermined. This is what some people blame the
      downfall of professional wrestling on – the WWE (formerly WWF)
      admitted that they script matches, even though most people believed
      it for a long time. Now that everyone knows it, wrestling is less
      interesting. For some of us, it was always less interesting.

      Talent dilution fits into the question of competitive balance, too,
      but I like to approach it from another angle, having tried all sorts
      of statistical things to measure it. (The one statistical measure of
      talent that works decently – not perfectly – is turnovers per
      possession, I think. I say that because it gets lower leaguewise
      from grade school to middle school to high school to college to pros
      and, being somewhat sexist here, it gets lower from female to male.
      You could use this to say that previous eras of the NBA were also
      less talented.) The fundamental problem is defining "talent". (I
      know I've said this before to this group.) Performance reflects
      talent PLUS team system, role on a team, rules, coaching, rest
      between games (there is more rest now than in the `60's, I think),
      home vs road effects and a host of (hopefully) smaller factors. We
      can measure performance and, in this discussion, have implicitly
      assumed that it only reflects talent (hence, we are measuring
      talent). But we assume wrong and, as Coach K says, "Only fools
      assume," or as the Bad News Bears said, "If you assume, it makes an
      ass out of u and me." I learned how many of life's lessons from the
      Bad News Bears? (One of those movies said that kissing makes babies
      and, boy, did that screw my teen dating life up.)

      But I get sidetracked. What statistics can be used to indicate
      talent? Using league stats has some problems, mainly due to the
      impact of rules, which change skill sets and change the balance
      between O and D, though perhaps don't change overall talent. Looking
      at how much subs played vs starters may have some role in the
      discussion. But a big hurdle is if offensive and defensive "talent"
      have been increasing or decreasing at equal rates, the performance
      stats won't show anything. Does anyone have any idea on how to deal
      with that?

      Someone pointed out that talent almost assuredly gets diluted in
      expansion years. This is true and is a minimum test of any statistic
      we use for assessing average level of talent. Many statistics should
      show this, however. I believe MikeG had some stat a long time ago
      showing how much net gain in minutes played per player there was,
      which was a weird stat (one I'm not sure about), but certainly shows
      expansion years and possibly other things. I think
      turnovers/possession would show expansion years, too. Would FG% do
      that? It's a purely offensive stat. Does this mean that "talent" is
      primarily an offensive thing? What other statistics can we use?

    • Show all 35 messages in this topic