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

WNBA turnaround

Expand Messages
  • Dean Oliver
    I did a quick analysis of the Charlotte Sting, who went 1-10 to start the season and 17-4 to end the season, to finish at 18-14. I d like to do a historical
    Message 1 of 11 , Aug 15, 2001
    View Source
    • 0 Attachment
      I did a quick analysis of the Charlotte Sting, who went 1-10 to start
      the season and 17-4 to end the season, to finish at 18-14. I'd like
      to do a historical search for streaks of this kind in any league, but
      knowing that such a search takes time and doing math doesn't, I
      figured I'd present the odds of such a thing occurring.

      Using basic binomial probability theory, the odds of a true 0.563 team
      going 1-10 or worse over 11 games is 0.17% (not 17%, 0.17%). Really
      low. The chances of a true 0.563 team going 17-4 or better in 21
      games is 1.67%. The odds of having both things happening is 0.00285%
      (the two numbers multiplied together), ridiculously low. I don't
      think I did anything wrong. If anything, a true Bayesian analysis
      would arrive at a lower number (it doesn't assume that the team is
      truly 0.563).

      So, now can anyone help me find the history of teams that might have
      had such streaks in one season? This is truly amazing. (The odds
      that the LA Sparks would go 18-0 is 9.04%, fairly high compared to
      what Charlotte did.) I can look later at the game/personnel reasons
      why Charlotte did this (someone is doing some of this analysis for
      me).

      Dean Oliver
      Journal of Basketball Studies
    • harlanzo@yahoo.com
      I can think of a couple of less dramatic but still streaky teams. the 95-96 cavs made the playoffs after a 0-7 (or 8) start. AFter a really mediocre start,
      Message 2 of 11 , Aug 17, 2001
      View Source
      • 0 Attachment
        I can think of a couple of less dramatic but still streaky teams.
        the 95-96 cavs made the playoffs after a 0-7 (or 8) start. AFter a
        really mediocre start, the Houston Rockets of 90-91 had a ridiculous
        hot streak mid-season after Hakeem fractured his eye socket (courtesy
        of Bill Cartwright's elbow).



        --- In APBR_analysis@y..., "Dean Oliver" <deano@t...> wrote:
        >
        > I did a quick analysis of the Charlotte Sting, who went 1-10 to
        start
        > the season and 17-4 to end the season, to finish at 18-14. I'd
        like
        > to do a historical search for streaks of this kind in any league,
        but
        > knowing that such a search takes time and doing math doesn't, I
        > figured I'd present the odds of such a thing occurring.
        >
        > Using basic binomial probability theory, the odds of a true 0.563
        team
        > going 1-10 or worse over 11 games is 0.17% (not 17%, 0.17%).
        Really
        > low. The chances of a true 0.563 team going 17-4 or better in 21
        > games is 1.67%. The odds of having both things happening is
        0.00285%
        > (the two numbers multiplied together), ridiculously low. I don't
        > think I did anything wrong. If anything, a true Bayesian analysis
        > would arrive at a lower number (it doesn't assume that the team is
        > truly 0.563).
        >
        > So, now can anyone help me find the history of teams that might
        have
        > had such streaks in one season? This is truly amazing. (The odds
        > that the LA Sparks would go 18-0 is 9.04%, fairly high compared to
        > what Charlotte did.) I can look later at the game/personnel
        reasons
        > why Charlotte did this (someone is doing some of this analysis for
        > me).
        >
        > Dean Oliver
        > Journal of Basketball Studies
      • Michael K. Tamada
        ... Though you conclusion is correct, I don t think the formula you used is correct. E.g. if a team goes 1-10, with a true probability of winning of .563,
        Message 3 of 11 , Aug 17, 2001
        View Source
        • 0 Attachment
          On Thu, 16 Aug 2001, Dean Oliver wrote:

          >
          > I did a quick analysis of the Charlotte Sting, who went 1-10 to start
          > the season and 17-4 to end the season, to finish at 18-14. I'd like
          > to do a historical search for streaks of this kind in any league, but
          > knowing that such a search takes time and doing math doesn't, I
          > figured I'd present the odds of such a thing occurring.
          >
          > Using basic binomial probability theory, the odds of a true 0.563 team
          > going 1-10 or worse over 11 games is 0.17% (not 17%, 0.17%). Really
          > low. The chances of a true 0.563 team going 17-4 or better in 21
          > games is 1.67%. The odds of having both things happening is 0.00285%
          > (the two numbers multiplied together), ridiculously low. I don't
          > think I did anything wrong. If anything, a true Bayesian analysis
          > would arrive at a lower number (it doesn't assume that the team is
          > truly 0.563).
          >
          > So, now can anyone help me find the history of teams that might have
          > had such streaks in one season? This is truly amazing. (The odds

          Though you conclusion is correct, I don't think the formula you used is
          correct. E.g. if a team goes 1-10, with a true probability of winning of
          .563, then although that event may have a 0.17% probability of occuring,
          the next event, going 17-4, has a *100%* probability of occuring. Because
          once a team goes 1-10, if it's going to end up 18-14, it HAS to go 17-4
          the rest of the way.

          I.e. we're looking at a team with a fixed 18-14 record (not a random
          sample of 32 games), and have to use "sampling without replacement"
          methods rather than "sampling with replacement", as the binomial equation
          assumes.


          I think (I'm not sure) that the correct formula is along these lines:
          there are 32 games, and 18 victories. If we look at the first 11 games,
          and then the last 21 games, let's call those the "first portion" and "last
          portion" of the season. How many different combinations are there of
          winning 0 or 1 games in the first portion (and therefore 17 or 18 in the
          last portion)?

          A bit of notation: xCy (pronounced "x choose y") is the combinatoric
          function [more commonly written like this: ( x ) ].
          y

          The formula for xCy is x!/(y!(x-y)!), where ! means factorial.

          Let "w" stand for the number of wins in the first portion of the season.
          Then, looking at the first portion, there are 11Cw ways of winning w games
          in the first portion. The Sting will therefore have won 18-w games in the
          last portion of the season, with 21C(18-w) ways of doing this. Total
          number of combinations for w wins is:

          11Cw * 21C(18-w)

          For w = 0 and 1, this is

          11C0 * 21C18 plus
          11C1 * 21C17


          which is

          1 * (21*20*19/3*2*1) plus
          11 * (21*20*19*18/4*3*2*1)

          or

          1,330 plus 65,835 =
          67,165 ways for the Sting to win 0 or 1 games in their first 11


          How many combinations of 18 wins and 14 defeats are there? The answer is
          32C18, which is 471,435,600 (there's danger of memory overflows with
          factorials this big, so I confirmed this calculation by summing all the
          11Cw * 21C(18-w) values for w = 0 to 11).

          So the probability of the Sting winning 0 or 1 games in their first 11 is
          65,385 / 471,435,600 = .00001425, or .001425%.


          This is exactly half of the figure that you came up with, which is
          probably not a coincidence, though I can't figure out what the reason for
          the relationship would be. (Possibly, either you or I made an arithmetic
          mistake somewhere.)



          --MKT
        • Charles Steinhardt
          I once looked at the question of whether baseball teams were streaky from a mathematical point of view. For anybody familiar with generating functions I d
          Message 4 of 11 , Aug 17, 2001
          View Source
          • 0 Attachment
            I once looked at the question of whether baseball teams were "streaky"
            from a mathematical point of view. For anybody familiar with generating
            functions I'd be happy to explain the basic approach, though I don't have
            time now (work keeps me *very* busy at the moment).

            However, the correct mathematical approach (which can be exact or
            approximate, as this approach will give you an series to sum for the
            probability) is to, in this case, assume that there are randomly 18 wins
            and 14 losses put together and to look at all 32!/(18!14!) combinations
            for how many have a longest streak of exactly N wins or losses. This can
            be solved directly if in a messy way, and I'll see if I still have my
            program for summing the series that I wrote a few years ago.

            If I don't, then I might be able to rewrite it if I have time.

            The answer for baseball, if you're curious, was that teams have many more
            streaks than expected of length 3-4, many fewer of length 5 or 6, and
            many more of length 8+.

            This can be explained to a large extent by noting that every 5 games, a
            team will have its best and its worst pitchers both pitching, and a team's
            performance is most sensitive (by far!) to its starting pitcher.

            In basketball, I might guess that there would be fewer streaks than
            expected of any length greater than about 3 because the homestands are so
            short and because basketball is so sensitive to the home team. For
            example, in baseball when I did my study, the home team was .511 in the
            previous decade. In basketball I don't know that number, but from seeing
            the best teams in the league frequently just break .500 on the road
            (particularly the Celtics back when they were very good), I might guess
            the home winning percentage to be around .625 or so, good enough to skew
            this sort of result.

            Then again, the other major difference is that in baseball the best team
            in the league will lose to the worst team about 1/3 of the time. In
            basketball, this would be more like 1/20 of the time I believe, or
            certainly more than their records alone would indicate. So while in
            baseball I could generally make the assumption that beyond the length of
            one 3-4 game series (the reason for more of those streaks IMO) the
            schedule was about constant, in basketball this is not true.

            Any thoughts?




            >
            >
            > On Thu, 16 Aug 2001, Dean Oliver wrote:
            >
            > >
            > > I did a quick analysis of the Charlotte Sting, who went 1-10 to start
            > > the season and 17-4 to end the season, to finish at 18-14. I'd like
            > > to do a historical search for streaks of this kind in any league, but
            > > knowing that such a search takes time and doing math doesn't, I
            > > figured I'd present the odds of such a thing occurring.
            > >
            > > Using basic binomial probability theory, the odds of a true 0.563 team
            > > going 1-10 or worse over 11 games is 0.17% (not 17%, 0.17%). Really
            > > low. The chances of a true 0.563 team going 17-4 or better in 21
            > > games is 1.67%. The odds of having both things happening is 0.00285%
            > > (the two numbers multiplied together), ridiculously low. I don't
            > > think I did anything wrong. If anything, a true Bayesian analysis
            > > would arrive at a lower number (it doesn't assume that the team is
            > > truly 0.563).
            > >
            > > So, now can anyone help me find the history of teams that might have
            > > had such streaks in one season? This is truly amazing. (The odds
            >
            > Though you conclusion is correct, I don't think the formula you used is
            > correct. E.g. if a team goes 1-10, with a true probability of winning of
            > .563, then although that event may have a 0.17% probability of occuring,
            > the next event, going 17-4, has a *100%* probability of occuring. Because
            > once a team goes 1-10, if it's going to end up 18-14, it HAS to go 17-4
            > the rest of the way.
            >
            > I.e. we're looking at a team with a fixed 18-14 record (not a random
            > sample of 32 games), and have to use "sampling without replacement"
            > methods rather than "sampling with replacement", as the binomial equation
            > assumes.
            >
            >
            > I think (I'm not sure) that the correct formula is along these lines:
            > there are 32 games, and 18 victories. If we look at the first 11 games,
            > and then the last 21 games, let's call those the "first portion" and "last
            > portion" of the season. How many different combinations are there of
            > winning 0 or 1 games in the first portion (and therefore 17 or 18 in the
            > last portion)?
            >
            > A bit of notation: xCy (pronounced "x choose y") is the combinatoric
            > function [more commonly written like this: ( x ) ].
            > y
            >
            > The formula for xCy is x!/(y!(x-y)!), where ! means factorial.
            >
            > Let "w" stand for the number of wins in the first portion of the season.
            > Then, looking at the first portion, there are 11Cw ways of winning w games
            > in the first portion. The Sting will therefore have won 18-w games in the
            > last portion of the season, with 21C(18-w) ways of doing this. Total
            > number of combinations for w wins is:
            >
            > 11Cw * 21C(18-w)
            >
            > For w = 0 and 1, this is
            >
            > 11C0 * 21C18 plus
            > 11C1 * 21C17
            >
            >
            > which is
            >
            > 1 * (21*20*19/3*2*1) plus
            > 11 * (21*20*19*18/4*3*2*1)
            >
            > or
            >
            > 1,330 plus 65,835 =
            > 67,165 ways for the Sting to win 0 or 1 games in their first 11
            >
            >
            > How many combinations of 18 wins and 14 defeats are there? The answer is
            > 32C18, which is 471,435,600 (there's danger of memory overflows with
            > factorials this big, so I confirmed this calculation by summing all the
            > 11Cw * 21C(18-w) values for w = 0 to 11).
            >
            > So the probability of the Sting winning 0 or 1 games in their first 11 is
            > 65,385 / 471,435,600 = .00001425, or .001425%.
            >
            >
            > This is exactly half of the figure that you came up with, which is
            > probably not a coincidence, though I can't figure out what the reason for
            > the relationship would be. (Possibly, either you or I made an arithmetic
            > mistake somewhere.)
            >
            >
            >
            > --MKT
            >
            >
            > To unsubscribe from this group, send an email to:
            > APBR_analysis-unsubscribe@yahoogroups.com
            >
            >
            >
            > Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
            >
            >
            >
          • Dean Oliver
            ... is ... winning of ... occuring, ... Because ... 17-4 ... The multiplication of the two numbers doesn t seem quite right. My calculations would have been
            Message 5 of 11 , Aug 17, 2001
            View Source
            • 0 Attachment
              --- In APBR_analysis@y..., "Michael K. Tamada" <tamada@o...> wrote:
              >
              > Though you conclusion is correct, I don't think the formula you used
              is
              > correct. E.g. if a team goes 1-10, with a true probability of
              winning of
              > .563, then although that event may have a 0.17% probability of
              occuring,
              > the next event, going 17-4, has a *100%* probability of occuring.
              Because
              > once a team goes 1-10, if it's going to end up 18-14, it HAS to go
              17-4
              > the rest of the way.
              >

              The multiplication of the two numbers doesn't seem quite right. My
              calculations would have been the same (except for the multiplication)
              if it was a team that went 180-140. The assumption that it is a 0.563
              team is where traditional statistics fail in comparison to Bayesian
              stats. The bigger question that Bayesian can help with is whether
              this team truly is a 0.563 team. Their victory over Cleveland last
              night suggests even further that they're better than a 0.563 team.

              I just use the BINOMDIST() function in excel to calculate the
              probabilities of exceedences. This function assumes a fixed
              probability of winning (0.563 in this case). Then it calculates the
              chances of a certain number of victories in a certain number of games.

              What you're looking at are the number of ways to get to 18-14 and the
              chances that they do it the way they did. That is probably a better
              way to calculate the odds of both such streaks occurring in the
              season.

              > I.e. we're looking at a team with a fixed 18-14 record (not a random
              > sample of 32 games), and have to use "sampling without replacement"
              > methods rather than "sampling with replacement", as the binomial
              equation
              > assumes.
              >

              > So the probability of the Sting winning 0 or 1 games in their first
              11 is
              > 65,385 / 471,435,600 = .00001425, or .001425%.
              >
              >
              > This is exactly half of the figure that you came up with, which is
              > probably not a coincidence, though I can't figure out what the
              reason for
              > the relationship would be. (Possibly, either you or I made an
              arithmetic
              > mistake somewhere.)

              Hmm. You calculate lower odds than I. I would have expected higher,
              if anything. Either way, we seem to be saying that the odds of such a
              season (assuming that 0.563 is their true p_win) are ridiculously low.
              This is going to be the case even if you look at home/road
              distribution or at the quality of competition (which was something of
              a factor for Charlotte).

              (Ignore my comment in my previous email about 0.17% vs. 0.14% -- I was
              trying to work on memory and mine is horrible.)

              The apparent big factor in turning Charlotte around was improved
              defense. This seemed to have been sparked by placing Tammy
              Sutton-Brown in the starting lineup over Machanguana (who they got
              for defense, ironically). The team was 16-3 with her starting, I
              believe (I can check tonight). Allison Feaster, despite her 3-point
              shooting, might also have been most influential on the defensive
              end, but I don't have a lot of evidence for that. On the offensive
              side, somehow they got Dawn Staley to cut down on her turnovers, but
              that was not as big as the defensive turnaround.

              Dean Oliver
              Journal of Basketball Studies
            • Dean Oliver
              ... streaky ... generating ... have ... Horrible how work gets in the way of actual quality science. ... more ... and ... games, a ... team s ... Pitching
              Message 6 of 11 , Aug 17, 2001
              View Source
              • 0 Attachment
                --- In APBR_analysis@y..., Charles Steinhardt <charles@p...> wrote:
                >
                > I once looked at the question of whether baseball teams were
                "streaky"
                > from a mathematical point of view. For anybody familiar with
                generating
                > functions I'd be happy to explain the basic approach, though I don't
                have
                > time now (work keeps me *very* busy at the moment).
                >

                Horrible how work gets in the way of actual quality science.

                >
                > The answer for baseball, if you're curious, was that teams have many
                more
                > streaks than expected of length 3-4, many fewer of length 5 or 6,
                and
                > many more of length 8+.
                >
                > This can be explained to a large extent by noting that every 5
                games, a
                > team will have its best and its worst pitchers both pitching, and a
                team's
                > performance is most sensitive (by far!) to its starting pitcher.
                >

                Pitching and the 4-5 man rotation really is important in analyzing
                baseball streaks. I can't think of any other sports like this.

                > In basketball, I might guess that there would be fewer streaks than
                > expected of any length greater than about 3 because the homestands
                are so
                > short and because basketball is so sensitive to the home team. For
                > example, in baseball when I did my study, the home team was .511 in
                the
                > previous decade. In basketball I don't know that number, but from
                seeing
                > the best teams in the league frequently just break .500 on the road
                > (particularly the Celtics back when they were very good), I might
                guess
                > the home winning percentage to be around .625 or so, good enough to
                skew
                > this sort of result.
                >

                Basketball is between 58% and about 65% typically. It's a strong
                force. I'd love to understand why it's so much more important. We
                can qualitatively say that the fans are much closer, but we should
                also be able to say that baseball teams can tailor their teams to the
                quirkiness of their ballpark. I guess the emotional power is
                stronger.

                > Then again, the other major difference is that in baseball the best
                team
                > in the league will lose to the worst team about 1/3 of the time. In
                > basketball, this would be more like 1/20 of the time I believe, or
                > certainly more than their records alone would indicate. So while in
                > baseball I could generally make the assumption that beyond the
                length of
                > one 3-4 game series (the reason for more of those streaks IMO) the
                > schedule was about constant, in basketball this is not true.

                First, a 65-17 team vs. a 17-65 team should win about 94% of the time,
                so good guess.

                Second, in baseball, a 104-58 team vs. a 58-104 team should win about
                76% of the time, so about 3/4, not 2/3.

                Basketball streakiness should be pretty much on par with what stats
                predict, I'd think. NBA teams don't have too many long homestands
                (the 7 game streak John brought up notwithstanding) to bias those
                streaks. They don't play too many home-home matchups, so their season
                ends up a pretty good random sampling through time. If the East or
                West is particularly weak, that might have an effect.

                Dean Oliver
                Journal of Basketball Studies
              • Charles Steinhardt
                ... Well, the other work I m doing is (hopefully) quality science too... :) ... I might suggest a few things, though admittedly without facts to back them up.
                Message 7 of 11 , Aug 17, 2001
                View Source
                • 0 Attachment
                  On Fri, 17 Aug 2001, Dean Oliver wrote:

                  > --- In APBR_analysis@y..., Charles Steinhardt <charles@p...> wrote:
                  > >
                  > > I once looked at the question of whether baseball teams were
                  > "streaky"
                  > > from a mathematical point of view. For anybody familiar with
                  > generating
                  > > functions I'd be happy to explain the basic approach, though I don't
                  > have
                  > > time now (work keeps me *very* busy at the moment).
                  > >
                  >
                  > Horrible how work gets in the way of actual quality science.
                  >

                  Well, the other work I'm doing is (hopefully) quality science too... :)

                  > Pitching and the 4-5 man rotation really is important in analyzing
                  > baseball streaks. I can't think of any other sports like this.
                  >
                  > > In basketball, I might guess that there would be fewer streaks than
                  > > expected of any length greater than about 3 because the homestands
                  > are so
                  > > short and because basketball is so sensitive to the home team. For
                  > > example, in baseball when I did my study, the home team was .511 in
                  > the
                  > > previous decade. In basketball I don't know that number, but from
                  > seeing
                  > > the best teams in the league frequently just break .500 on the road
                  > > (particularly the Celtics back when they were very good), I might
                  > guess
                  > > the home winning percentage to be around .625 or so, good enough to
                  > skew
                  > > this sort of result.
                  > >
                  >
                  > Basketball is between 58% and about 65% typically. It's a strong
                  > force. I'd love to understand why it's so much more important. We
                  > can qualitatively say that the fans are much closer, but we should
                  > also be able to say that baseball teams can tailor their teams to the
                  > quirkiness of their ballpark. I guess the emotional power is
                  > stronger.
                  >

                  I might suggest a few things, though admittedly without facts to back them
                  up. Certainly the proximity of the fans (and buildings designed for
                  maximum noise - look at the decibel level in Utah during the playoffs
                  sometime!) helps. Some other ideas:

                  1) The rotation is an equalizer in baseball. More to the point, the best
                  teams in basketball have a much better winning percentage than in
                  baseball.

                  2) There is no equivalent of the free throw in baseball, and in fact all
                  new stadia are forced to put some sort of blue/black screen in
                  straightaway center so that the batter has a good line of sight. Fans can
                  have a very direct impact in basketball that I'd guess is worth as much as
                  5 points per game (some in increasing the home FT%, some in decreasing
                  that of opponents. Maybe somebody has statistics on this?

                  3) Basketball is by its very nature a faster-paced game than baseball and
                  thus more prone to momentum-based runs. In baseball, it's always easy for
                  the struggling team to take the equivalent of a timeout. Can you imagine
                  what basketball would be like if there were a 20" timeout after every
                  posession? Whatever else you'd expect, I'd think that game would be less
                  prone to long runs. Limited timeouts increase the value of momentum, and
                  a home crowd is helpful in that respect.

                  4) Basketball is a younger sport than baseball, and as a result there are
                  many fewer traditional fans of one team that end up in another market. In
                  addition, the markets are a little bit better spaced than in baseball and
                  include a few more cities that only have one sports team. As a result,
                  95% of the fans at a game will be home fans. I was recently at a
                  Yankees-Phillies game in Philadelphia at which there were 55000 people,
                  about 40000 of them Yankees fans. In that atmosphere any home crowd
                  advantage must go away.

                  5) Basketball generally sells out, whereas baseball only does in select
                  markets. 10000 people at an Expos game isn't going to provide much of an
                  advantage, and if anything all those empty seats could be demoralizing.

                  6) Baseball can tailor its field to one or two hitters, but rarely an
                  entire lineup of 8 or 9 people as well as pitchers. That is to say, the
                  park can favor pitching or hitting, and can favor right- or left-handed
                  hitters, but any pro team has all of those. So this effect should be
                  minor, if any, and is mostly in relation to how fielders react to
                  different plays. In addition, with the unbalanced schedule most teams
                  that visit play there 8 or 9 games a year, and the rest 3 or 4. So there
                  is plenty of time to adjust to a park for a veteran or even by the end of
                  a series for a rookie. Incidentally, basketball does have its share of
                  home-court advantages if you know where to look: I remember that before
                  the Celtics moved out of the Garden, there used to be well-placed/hidden
                  dead spots in the parquet for example. But certainly it's not as
                  prevalent - basketball doesn't exactly have ground rules to worry about.
                  Then again, there were a bunch of complaints a few years ago about playoff
                  rims (I forget where) being bad for jump shooters and favoring the home
                  team with a strong inside game. And, in basketball one can tailor the
                  team's style of play to almost always take inside shots much better than
                  in baseball one can get a group of players to always hit to left field,
                  say.






                  > > Then again, the other major difference is that in baseball the best
                  > team
                  > > in the league will lose to the worst team about 1/3 of the time. In
                  > > basketball, this would be more like 1/20 of the time I believe, or
                  > > certainly more than their records alone would indicate. So while in
                  > > baseball I could generally make the assumption that beyond the
                  > length of
                  > > one 3-4 game series (the reason for more of those streaks IMO) the
                  > > schedule was about constant, in basketball this is not true.
                  >
                  > First, a 65-17 team vs. a 17-65 team should win about 94% of the time,
                  > so good guess.
                  >
                  > Second, in baseball, a 104-58 team vs. a 58-104 team should win about
                  > 76% of the time, so about 3/4, not 2/3.
                  >

                  Where are you getting these numbers? Either way, the difference is again
                  in the rotation. That 104-58 team usually has starters with the following
                  records (let's say records when the team has them start, including W/L by
                  the bullpen)

                  A: 27-6
                  B: 24-9
                  C: 19-13
                  D: 18-14 (usually a collection of people in this slot at this point)
                  E: 16-16

                  Some have 3 top starters and two worse ones.

                  Meanwhile, the worst team will have pretty much the reverse, with those
                  last couple of slots usually a collection of people promoted and demoted
                  from starting or from AAA ball. That worst team also will very often
                  juggle the rotation to try to win one from the best team, as they're not
                  in a pennant race anyway but know that they'll have more people watching
                  the game against the good team and want to impress potential fans (and
                  play the role of spoiler). As a result, at least one of the three games
                  in the series will give the worst team an even chance or better of
                  winning, and quite possibly another. As a result, one rarely sees the
                  best team in the league sweep the worst team, or at least it is more rare
                  than you might predict.

                  In basketball, this is not the case. The teams are pretty much the same,
                  pending injuries/suspensions.


                  > Basketball streakiness should be pretty much on par with what stats
                  > predict, I'd think. NBA teams don't have too many long homestands
                  > (the 7 game streak John brought up notwithstanding) to bias those
                  > streaks. They don't play too many home-home matchups, so their season
                  > ends up a pretty good random sampling through time. If the East or
                  > West is particularly weak, that might have an effect.
                  >

                  I might expect otherwise for a few reasons:

                  1) Prevalence of young players. Meaning that veteran teams should do a
                  little better in the first half of the season and young teams in the
                  second half rather than being constant (does somebody have stats on this
                  one?)

                  2) Coaches make adjustments including an overhaul of the offense/defense
                  midseason if they struggle. Should cause the team to play differently,
                  one way or the other.

                  3) Relative importance of injuries. Baseball is not so sensitive to the
                  injury even of a superstar. For example, the Red Sox this year are 66-53
                  without the best pitcher in the game, an all-star catcher, an all-star
                  shortstop, and a bunch of other players that are important to their team
                  all for at least a month (and in the case of the SS and C, 3 months).
                  With no injuries, they should not be better than about 72-47, and even
                  that would be a very impressive record for this group of players.
                  However, what would the Magic have done with Grant Hill last year? The
                  Sixers without Iverson? While baseball doesn't have a lower injury rate
                  than basketball, an injury to one of five starters is more critical, and
                  thus more likely to affect the team's performance. Particularly an
                  important starter.

                  I'd be interested to see some statistics on this, particularly if somebody
                  actually tried to calculate the effect of different injuries/potential
                  injuries.

                  -Charles

                  > Dean Oliver
                  > Journal of Basketball Studies
                  >
                  >
                  > To unsubscribe from this group, send an email to:
                  > APBR_analysis-unsubscribe@yahoogroups.com
                  >
                  >
                  >
                  > Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
                  >
                  >
                  >
                • Dean Oliver
                  ... Lucky bum. ... baseball and ... easy for ... imagine ... every ... less ... momentum, and ... I have my doubts about this given the general info suggesting
                  Message 8 of 11 , Aug 17, 2001
                  View Source
                  • 0 Attachment
                    --- In APBR_analysis@y..., Charles Steinhardt <charles@p...> wrote:

                    >
                    > Well, the other work I'm doing is (hopefully) quality science too...
                    :)

                    Lucky bum.


                    > 3) Basketball is by its very nature a faster-paced game than
                    baseball and
                    > thus more prone to momentum-based runs. In baseball, it's always
                    easy for
                    > the struggling team to take the equivalent of a timeout. Can you
                    imagine
                    > what basketball would be like if there were a 20" timeout after
                    every
                    > posession? Whatever else you'd expect, I'd think that game would be
                    less
                    > prone to long runs. Limited timeouts increase the value of
                    momentum, and
                    > a home crowd is helpful in that respect.
                    >

                    I have my doubts about this given the general info suggesting that
                    within game streaks don't exist. (I'm not convinced about the
                    validity of that research either.)

                    > 4) Basketball is a younger sport than baseball, and as a result
                    there are
                    > many fewer traditional fans of one team that end up in another
                    market. In
                    > addition, the markets are a little bit better spaced than in
                    baseball and
                    > include a few more cities that only have one sports team. As a
                    result,
                    > 95% of the fans at a game will be home fans. I was recently at a
                    > Yankees-Phillies game in Philadelphia at which there were 55000
                    people,
                    > about 40000 of them Yankees fans. In that atmosphere any home crowd
                    > advantage must go away.
                    >

                    You haven't been to Golden St, have you?

                    > 5) Basketball generally sells out, whereas baseball only does in
                    select
                    > markets. 10000 people at an Expos game isn't going to provide much
                    of an
                    > advantage, and if anything all those empty seats could be
                    demoralizing.
                    >

                    Uh, Golden St. again.

                    > 6) Baseball can tailor its field to one or two hitters, but rarely
                    an
                    > entire lineup of 8 or 9 people as well as pitchers. That is to say,
                    the
                    > park can favor pitching or hitting, and can favor right- or
                    left-handed
                    > hitters, but any pro team has all of those. So this effect should
                    be
                    > minor, if any, and is mostly in relation to how fielders react to
                    > different plays. In addition, with the unbalanced schedule most
                    teams
                    > that visit play there 8 or 9 games a year, and the rest 3 or 4. So
                    there
                    > is plenty of time to adjust to a park for a veteran or even by the
                    end of
                    > a series for a rookie. Incidentally, basketball does have its share
                    of
                    > home-court advantages if you know where to look: I remember that
                    before
                    > the Celtics moved out of the Garden, there used to be
                    well-placed/hidden
                    > dead spots in the parquet for example. But certainly it's not as
                    > prevalent - basketball doesn't exactly have ground rules to worry
                    about.
                    > Then again, there were a bunch of complaints a few years ago about
                    playoff
                    > rims (I forget where) being bad for jump shooters and favoring the
                    home
                    > team with a strong inside game. And, in basketball one can tailor
                    the
                    > team's style of play to almost always take inside shots much better
                    than
                    > in baseball one can get a group of players to always hit to left
                    field,
                    > say.
                    >
                    >

                    Well, other than the old Boston Garden, I don't remember hearing of
                    any physical reason for an arena to favor one team over another.


                    > > First, a 65-17 team vs. a 17-65 team should win about 94% of the
                    time,
                    > > so good guess.
                    > >
                    > > Second, in baseball, a 104-58 team vs. a 58-104 team should win
                    about
                    > > 76% of the time, so about 3/4, not 2/3.
                    > >
                    >
                    > Where are you getting these numbers? Either way, the difference is

                    Bill James:

                    Win% Team A vs. Team B



                    >
                    > > Basketball streakiness should be pretty much on par with what
                    stats
                    > > predict, I'd think. NBA teams don't have too many long homestands
                    > > (the 7 game streak John brought up notwithstanding) to bias those
                    > > streaks. They don't play too many home-home matchups, so their
                    season
                    > > ends up a pretty good random sampling through time. If the East
                    or
                    > > West is particularly weak, that might have an effect.
                    > >
                    >
                    > I might expect otherwise for a few reasons:
                    >
                    > 1) Prevalence of young players. Meaning that veteran teams should
                    do a
                    > little better in the first half of the season and young teams in the
                    > second half rather than being constant (does somebody have stats on
                    this
                    > one?)

                    Testable. However, the general hypothesis has actually been the other
                    way around. Young players supposedly hit a wall and do worse in the
                    2nd half. I tend to think you're right, but have had a hard time
                    testing it, not having a very up-to-date player directory with
                    birthdays.

                    >
                    > 2) Coaches make adjustments including an overhaul of the
                    offense/defense
                    > midseason if they struggle. Should cause the team to play
                    differently,
                    > one way or the other.
                    >

                    No difference between baseball and basketball on this one.

                    > 3) Relative importance of injuries. Baseball is not so sensitive to
                    the
                    > injury even of a superstar. For example, the Red Sox this year are
                    66-53
                    > without the best pitcher in the game, an all-star catcher, an
                    all-star
                    > shortstop, and a bunch of other players that are important to their
                    team
                    > all for at least a month (and in the case of the SS and C, 3
                    months).
                    > With no injuries, they should not be better than about 72-47, and
                    even
                    > that would be a very impressive record for this group of players.
                    > However, what would the Magic have done with Grant Hill last year?
                    The
                    > Sixers without Iverson? While baseball doesn't have a lower injury
                    rate
                    > than basketball, an injury to one of five starters is more critical,
                    and
                    > thus more likely to affect the team's performance. Particularly an
                    > important starter.
                    >
                    > I'd be interested to see some statistics on this, particularly if
                    somebody
                    > actually tried to calculate the effect of different
                    injuries/potential
                    > injuries.

                    The problem is always how you replace a superstar. If you replace
                    them with a bad player, the team gets much worse. The Bulls without
                    Jordan actually did pretty well the first year, then suffered the
                    next. The Sixers did play a few without Iverson this year, so we can
                    check. The Raptors without Vince. I am forming this unjustified
                    opinion in my head that teams that play 1-3 games without their
                    superstar generally do about the same. Teams that play more than
                    about 10 games without their superstar really start to hurt. I need
                    to form the hypothesis a little better, but I think I've seen it.

                    Dean Oliver
                    Journal of Basketball Studies
                  • Dean Oliver
                    Forgot to finish the formula on Win% calculations... ... is ... Win%A_B = [Win%A*(1-Win%B)]/[Win%A*(1-Win%B)+(1-Win%A)*Win%B]
                    Message 9 of 11 , Aug 17, 2001
                    View Source
                    • 0 Attachment
                      Forgot to finish the formula on Win% calculations...

                      >
                      > > > First, a 65-17 team vs. a 17-65 team should win about 94% of the
                      > time,
                      > > > so good guess.
                      > > >
                      > > > Second, in baseball, a 104-58 team vs. a 58-104 team should win
                      > about
                      > > > 76% of the time, so about 3/4, not 2/3.
                      > > >
                      > >
                      > > Where are you getting these numbers? Either way, the difference
                      is
                      >
                      > Bill James:
                      >
                      > Win% Team A vs. Team B
                      >

                      Win%A_B = [Win%A*(1-Win%B)]/[Win%A*(1-Win%B)+(1-Win%A)*Win%B]

                      http://www.rawbw.com/~deano/methdesc.html#matchup

                      has the detailed info.
                    • Dean LaVergne
                      ... From: Charles Steinhardt [mailto:charles@princeton.edu] 2) There is no equivalent of the free throw in baseball, and in fact all new stadia are forced to
                      Message 10 of 11 , Aug 17, 2001
                      View Source
                      • 0 Attachment
                         
                        -----Original Message-----
                        From: Charles Steinhardt [mailto:charles@...]



                        2) There is no equivalent of the free throw in baseball, and in fact all
                        new stadia are forced to put some sort of blue/black screen in
                        straightaway center so that the batter has a good line of sight.  Fans can
                        have a very direct impact in basketball that I'd guess is worth as much as
                        5 points per game (some in increasing the home FT%, some in decreasing
                        that of opponents.  Maybe somebody has statistics on this? 


                        [Dean LaVergne] It doesn't seem to hold out.  For the last ten years:
                         
                        Free Throw Percentage:
                         
                        Season   Away    Home    Diff
                        1992    75.62%  76.12%   0.51%
                        1993    75.19%  75.65%   0.46%
                        1994    73.60%  73.25%  -0.35%
                        1995    73.49%  73.84%   0.34%
                        1996    73.91%  74.03%   0.12%
                        1997    73.98%  73.67%  -0.31%
                        1998    73.54%  73.82%   0.28%
                        1999    72.29%  73.32%   1.02%
                        2000    74.55%  75.44%   0.89%
                        2001    74.48%  74.99%   0.51%
                                   
                                   
                        However, free throws attempted seem a little more significant:
                                    
                                   
                        Season   Away    Home    Diff    % Diff
                        1992   28,522  30,553   2,031    7.12%
                        1993   29,715  31,659   1,944    6.54%
                        1994   28,688  30,131   1,443    5.03%
                        1995   29,248  30,690   1,442    4.93%
                        1996   30,587  32,178   1,591    5.20%
                        1997   29,522  30,272     750    2.54%
                        1998   30,704  31,799   1,095    3.57%
                        1999   18,399  19,000     601    3.27%
                        2000   29,410  30,352     942    3.20%
                        2001   28,570  29,308     738    2.58%
                                            
                        Dean L
                         
                      • Charles Steinhardt
                        Very interesting (and obviously not what I d expect looking from the point of view of a fan). Anybody know why the FT disparity has dropped (and how strongly
                        Message 11 of 11 , Aug 17, 2001
                        View Source
                        • 0 Attachment
                          Very interesting (and obviously not what I'd expect looking from the point
                          of view of a fan).

                          Anybody know why the FT disparity has dropped (and how strongly that
                          disparity correlates with winning %)? Have there been new instructions to
                          officials?

                          On Fri, 17 Aug 2001, Dean LaVergne wrote:

                          >
                          > -----Original Message-----
                          > From: Charles Steinhardt [mailto:charles@...]
                          >
                          >
                          >
                          >
                          > 2) There is no equivalent of the free throw in baseball, and in fact all
                          > new stadia are forced to put some sort of blue/black screen in
                          > straightaway center so that the batter has a good line of sight. Fans can
                          > have a very direct impact in basketball that I'd guess is worth as much as
                          > 5 points per game (some in increasing the home FT%, some in decreasing
                          > that of opponents. Maybe somebody has statistics on this?
                          >
                          >
                          > [Dean LaVergne] It doesn't seem to hold out. For the last ten years:
                          >
                          > Free Throw Percentage:
                          >
                          > Season Away Home Diff
                          > 1992 75.62% 76.12% 0.51%
                          > 1993 75.19% 75.65% 0.46%
                          > 1994 73.60% 73.25% -0.35%
                          > 1995 73.49% 73.84% 0.34%
                          > 1996 73.91% 74.03% 0.12%
                          > 1997 73.98% 73.67% -0.31%
                          > 1998 73.54% 73.82% 0.28%
                          > 1999 72.29% 73.32% 1.02%
                          > 2000 74.55% 75.44% 0.89%
                          > 2001 74.48% 74.99% 0.51%
                          >
                          >
                          > However, free throws attempted seem a little more significant:
                          >
                          >
                          > Season Away Home Diff % Diff
                          > 1992 28,522 30,553 2,031 7.12%
                          > 1993 29,715 31,659 1,944 6.54%
                          > 1994 28,688 30,131 1,443 5.03%
                          > 1995 29,248 30,690 1,442 4.93%
                          > 1996 30,587 32,178 1,591 5.20%
                          > 1997 29,522 30,272 750 2.54%
                          > 1998 30,704 31,799 1,095 3.57%
                          > 1999 18,399 19,000 601 3.27%
                          > 2000 29,410 30,352 942 3.20%
                          > 2001 28,570 29,308 738 2.58%
                          >
                          >
                          > Dean L
                          >
                          >
                        Your message has been successfully submitted and would be delivered to recipients shortly.