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

Re: [APBR_analysis] My Version 2 of WINVAL

Expand Messages
  • igor eduardo küpfer
    From: dan_t_rosenbaum To: Sent: Sunday, February 29, 2004 1:45 AM Subject: [APBR_analysis] My Version 2
    Message 1 of 5 , Mar 1 12:39 AM
    • 0 Attachment
      From: "dan_t_rosenbaum" <rosenbaum@...>
      To: <APBR_analysis@yahoogroups.com>
      Sent: Sunday, February 29, 2004 1:45 AM
      Subject: [APBR_analysis] My Version 2 of WINVAL


      > This version is possessions-based (and thus accounts for pace) and
      > also includes offensive and defensive ratings.
      >
      > In order to make it easier to read, I just included it and my
      > comments on it on a web-page that I created. Here is the link.
      >
      > http://www.uncg.edu/bae/people/rosenbaum/NBA/winval1.htm
      >
      > (This thing has almost turned into a mini-paper.)
      >

      "Mini"? :-)

      Great stuff, Dan. It will take me a few days to digest the whole thing. A
      couple of comments in passing:

      * <quote>
      Null Hypothesis: the cost of two point FGA is the same as the cost of three
      point FGA

      The null hypothesis is soundly rejected with three point FGA having a lower
      cost, even after accounting for points produced and free throw attempts
      generated.
      </q>

      Would it be asking too much to try this one again, this time including
      rebounds in the accounting? Common Wisdom has it that 3 pointers generate
      more offensive rebounds, which could add to the value of the long attempts.
      (I'm aware that DeanO disputes CW in his book, and I accept his conclusion
      that 3 misses are no more likely to be offensively rebounded than 2 misses.
      Independent confirmation would be very nice, however.)

      * <quote>
      Null Hypothesis: steals and turnovers have the same value, but opposite
      signs

      It is a fairly close call, but the null hypothesis is rejected. It appears
      that steals help teams more than turnovers hurt teams. This could be because
      guys who get steals disrupt offenses in other ways that are not picked up in
      traditional statistics.
      </q>

      Is there some double counting going on here? A steal is always a turnover,
      but a turnover is not always a steal. Is it possible to isolate the types of
      turnovers (steals, bad pass, OFF foul, etc.) in this analysis?

      * <quote>
      Null Hypothesis: offensive and defensive rebounds have the same value

      There is no evidence to reject this hypothesis. Thus, even though these two
      regressions have higher coefficients for defensive rebounds (versus
      offensive rebounds), there really is no evidence to conclude statistically
      that the value of defensive rebounds is higher than that of offensive
      rebounds. That said, it is very, very difficult to draw the conclusion from
      these data that offensive rebounds should be valued higher than defensive
      rebounds, as is often argued.
      </q>

      Still a topic of great interest for me. My opinion, backed tentatively by
      the data I've looked at, is that defensive rebounds have less marginal value
      than offensive rebounds. That is, an individual player's defensive rebound
      adds less to his team's defensive rebounding percentage than that player's
      offensive rebound adds to his team's offensive rebounding percentage. I'm
      not sure if you address this: it could be that DRebs and ORebs add the same
      amount to a team's production, while at the same time saying each _player's_
      O and D rebounds have a different value.




      --
      cheers,
      ed
    • Mike G
      ... attempts... ... Part of the value of a 2-pt attempt is that fouls are created. Part of this value is in the high-% FT that ensue; and part is that fouls
      Message 2 of 5 , Mar 1 4:30 AM
      • 0 Attachment
        --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer
        <edkupfer@r...> wrote:
        > ... Common Wisdom has it that 3 pointers generate
        > more offensive rebounds, which could add to the value of the long
        attempts...
        >

        Part of the value of a "2-pt attempt" is that fouls are created.
        Part of this value is in the high-% FT that ensue; and part is that
        fouls can weaken the opponent for the remainder of the game.

        However, a study that measures blocks of time when certain players
        are in the game will not pick up an effect that removes a player
        from the lineup. By definition, that effect is shifted to a
        later 'time block'.


        > Is there some double counting going on here? A steal is always a
        turnover,
        > but a turnover is not always a steal. Is it possible to isolate
        the types of
        > turnovers (steals, bad pass, OFF foul, etc.) in this analysis?

        My impression is that Dan gets it right to conjecture that each
        Recorded steal is associated with other disruptions by the defense.
        Conversely, a turnover may not be 'all that bad' if it's equivalent
        to a desperation shot; i.e., a desperation clock-fighting pass.

        > .. My opinion, backed tentatively by
        > the data I've looked at, is that defensive rebounds have less
        marginal value
        > than offensive rebounds. That is, an individual player's defensive
        rebound
        > adds less to his team's defensive rebounding percentage than that
        player's
        > offensive rebound adds to his team's offensive rebounding
        percentage...

        Again, each defensive rebound that gets recorded may represent
        several 'assisted rebounds', in which there are several defensive
        players boxing out and otherwise making it easier for one player to
        get the rebound. 3 players may work together to gather 20 rebounds;
        how they officially divide them up for credit, is immaterial to the
        success of the team.

        So each DR recorded may represent several "partial" DR, if such
        things were recorded.
      • John Hollinger
        My research in the 2002 Prospectus showed just the opposite -- that a 3-pointer had LESS chance of being rebounded than a two-point shot. Which makes sense,
        Message 3 of 5 , Mar 1 12:00 PM
        • 0 Attachment
          My research in the 2002 Prospectus showed just the opposite -- that a
          3-pointer had LESS chance of being rebounded than a two-point shot.
          Which makes sense, because coaches are fond of saying the most
          dangerous rebounder is the shooter, and on a 3-pointer the shooter is
          too far away to get most rebounds.



          --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer
          <edkupfer@r...> wrote:
          > From: "dan_t_rosenbaum" <rosenbaum@u...>
          > To: <APBR_analysis@yahoogroups.com>
          > Sent: Sunday, February 29, 2004 1:45 AM
          > Subject: [APBR_analysis] My Version 2 of WINVAL
          >
          >
          > > This version is possessions-based (and thus accounts for pace) and
          > > also includes offensive and defensive ratings.
          > >
          > > In order to make it easier to read, I just included it and my
          > > comments on it on a web-page that I created. Here is the link.
          > >
          > > http://www.uncg.edu/bae/people/rosenbaum/NBA/winval1.htm
          > >
          > > (This thing has almost turned into a mini-paper.)
          > >
          >
          > "Mini"? :-)
          >
          > Great stuff, Dan. It will take me a few days to digest the whole
          thing. A
          > couple of comments in passing:
          >
          > * <quote>
          > Null Hypothesis: the cost of two point FGA is the same as the cost
          of three
          > point FGA
          >
          > The null hypothesis is soundly rejected with three point FGA having
          a lower
          > cost, even after accounting for points produced and free throw
          attempts
          > generated.
          > </q>
          >
          > Would it be asking too much to try this one again, this time
          including
          > rebounds in the accounting? Common Wisdom has it that 3 pointers
          generate
          > more offensive rebounds, which could add to the value of the long
          attempts.
          > (I'm aware that DeanO disputes CW in his book, and I accept his
          conclusion
          > that 3 misses are no more likely to be offensively rebounded than 2
          misses.
          > Independent confirmation would be very nice, however.)
          >
          > * <quote>
          > Null Hypothesis: steals and turnovers have the same value, but
          opposite
          > signs
          >
          > It is a fairly close call, but the null hypothesis is rejected. It
          appears
          > that steals help teams more than turnovers hurt teams. This could
          be because
          > guys who get steals disrupt offenses in other ways that are not
          picked up in
          > traditional statistics.
          > </q>
          >
          > Is there some double counting going on here? A steal is always a
          turnover,
          > but a turnover is not always a steal. Is it possible to isolate the
          types of
          > turnovers (steals, bad pass, OFF foul, etc.) in this analysis?
          >
          > * <quote>
          > Null Hypothesis: offensive and defensive rebounds have the same
          value
          >
          > There is no evidence to reject this hypothesis. Thus, even though
          these two
          > regressions have higher coefficients for defensive rebounds (versus
          > offensive rebounds), there really is no evidence to conclude
          statistically
          > that the value of defensive rebounds is higher than that of
          offensive
          > rebounds. That said, it is very, very difficult to draw the
          conclusion from
          > these data that offensive rebounds should be valued higher than
          defensive
          > rebounds, as is often argued.
          > </q>
          >
          > Still a topic of great interest for me. My opinion, backed
          tentatively by
          > the data I've looked at, is that defensive rebounds have less
          marginal value
          > than offensive rebounds. That is, an individual player's defensive
          rebound
          > adds less to his team's defensive rebounding percentage than that
          player's
          > offensive rebound adds to his team's offensive rebounding
          percentage. I'm
          > not sure if you address this: it could be that DRebs and ORebs add
          the same
          > amount to a team's production, while at the same time saying each
          _player's_
          > O and D rebounds have a different value.
          >
          >
          >
          >
          > --
          > cheers,
          > ed
        • Mike G
          ... that a ... shot. ... is ... This platitude surely predates the 3-point shot. And the danger in a player getting a rebound may lie more in what he ll do
          Message 4 of 5 , Mar 2 7:27 AM
          • 0 Attachment
            --- In APBR_analysis@yahoogroups.com, "John Hollinger"
            <alleyoop2@y...> wrote:
            > My research in the 2002 Prospectus showed just the opposite --
            that a
            > 3-pointer had LESS chance of being rebounded than a two-point
            shot.
            > Which makes sense, because coaches are fond of saying the most
            > dangerous rebounder is the shooter, and on a 3-pointer the shooter
            is
            > too far away to get most rebounds.

            This platitude surely predates the 3-point shot. And the "danger"
            in a player getting a rebound may lie more in what he'll do with it,
            than in the chances of his getting it.

            So given that a missed 3 is no more likely (or less likely) to be
            rebounded offensively, what accounts for the difference detected by
            Dr. Dan ?

            What about the missed shot that goes out of bounds? Probably more
            likely on a 3 than on a 2. Does the defense get a rebound? No, but
            effectively it's the same. Do they then get a fast break
            opportunity? No.

            So teams that shoot a lot of 3's may have relatively fewer fast
            breaks run against them. Yet they may have relatively more
            offensive rebounds. For both reasons, the outside-shooting team
            gains a cushion against a bigger/stronger opponent.

            This brings up the effect of "difference in team style". For a team
            without inside scoring, it does make sense to shoot more long
            shots. Forcing the ball inside makes the offense less efficient.

            Meanwhile, it's possible for an inside-efficient team to be even
            more efficient when shooting the 3. A mediocre 3-pt shooter can be
            a lot better when he's wide open all day, due to doubling inside.

            Smart coaches and teams take what is given them, most of the time.
            For the past several years (since before the shortened-arc era of
            the mid '90s), 3-pt shooting is (league-wide) more efficient than 2-
            pt and FT combined.




            >
            >
            >
            > --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer
            > <edkupfer@r...> wrote:
            > > From: "dan_t_rosenbaum" <rosenbaum@u...>
            > > * <quote>
            > > Null Hypothesis: the cost of two point FGA is the same as the
            cost
            > of three
            > > point FGA
            > >
            > > The null hypothesis is soundly rejected with three point FGA
            having
            > a lower
            > > cost, even after accounting for points produced and free throw
            > attempts
            > > generated.
            > > </q>
            > >
            > > Would it be asking too much to try this one again, this time
            > including
            > > rebounds in the accounting? Common Wisdom has it that 3 pointers
            > generate
            > > more offensive rebounds, which could add to the value of the
            long
            > attempts.
            > > (I'm aware that DeanO disputes CW in his book, and I accept his
            > conclusion
            > > that 3 misses are no more likely to be offensively rebounded
            than 2
            > misses.
            > > Independent confirmation would be very nice, however.)
          Your message has been successfully submitted and would be delivered to recipients shortly.