## comparing eras again

Expand Messages
• A while back, we had a long discussion on comparing eras. In that spirit, I saw this article on baseball and eras that was interesting:
Message 1 of 6 , Mar 22, 2004
A while back, we had a long discussion on comparing eras. In that
spirit, I saw this article on baseball and eras that was interesting:

http://baseballprimer.com/articles/cdial_2004-03-18_0.shtml

It doesn't resolve the issues and it doesn't translate perfectly to
hoops. I also seem to remember someone here (maybe mike g) also
looking at standard deviation. But this author makes some
interesting arguments and asks some interesting questions (without
answering them) that help attack the problem for other angles:

-If Babe Ruth in his baseball prime was transported into our modern
league, would he be worth more wins than Barry Bonds over a career?
Over one season?

-If Babe Ruth was born in the same year as Barry Bonds, and grew up
in the same environment, would he be worth more wins than Barry
Bonds over a career? Over one season?

-Who was worth more wins in his contextual era?

Switch and Ruth and Bonds with Wilt and Shaq and you've got the same
issues.
• Here is the procedure I would do if I wanted to get to the bottom of this question. 1. Take a given statistical index (such as PER or Tendex) and normalize it
Message 2 of 6 , Mar 22, 2004
Here is the procedure I would do if I wanted to get to the bottom of
this question.

1. Take a given statistical index (such as PER or Tendex) and
normalize it to have the same mean in every season.

2. By season, estimate the fraction of total PER (or Tendex)
accounted for by each birth year cohort. This will result in about
20 observations for each season.

3. Run this regression using as many seasons as possible - preferably
going back to the 40s, 50s, and 60s. (The index may need to be

PER_FR = b1*BIRTH_YEAR_DUMMIES + b2*AGE_DUMMIES + e,

where PER_FR is the fraction of total PER (or Tendex) accounted for
by a given birth year cohort in a given season,
BIRTH_YEAR_DUMMIES are a series of dummy variables for birth year
AGE_DUMMIES are a series of dummy variables for age

If the b1 coefficients tend to more positive for later birth year
cohorts, then succeeding generations of players have been more
productive than their predecessors. If the b1 coefficients are
negative, succeeding generations of players have been less productive
than their predecessors.

I do not have the data to run this regression, but I would be very
interested in hearing what happens if anyone has the data, time, and
inclination to run this regression.

A handful of birth year cohorts may get a little bump due to the
timing of expansion, but overall this method should be do a good job
of handling the complications brought on by expansion.

This model assumes many things, including that the relationship
between age and NBA production has not changed over time. Slight
variations of the basic model above could deal with that issue.

--- In APBR_analysis@yahoogroups.com, "harlanzo" <harlanzo@y...>
wrote:
> A while back, we had a long discussion on comparing eras. In that
> spirit, I saw this article on baseball and eras that was
interesting:
>
> http://baseballprimer.com/articles/cdial_2004-03-18_0.shtml
>
> It doesn't resolve the issues and it doesn't translate perfectly to
> hoops. I also seem to remember someone here (maybe mike g) also
> looking at standard deviation. But this author makes some
> interesting arguments and asks some interesting questions (without
> answering them) that help attack the problem for other angles:
>
> -If Babe Ruth in his baseball prime was transported into our modern
> league, would he be worth more wins than Barry Bonds over a career?
> Over one season?
>
> -If Babe Ruth was born in the same year as Barry Bonds, and grew up
> in the same environment, would he be worth more wins than Barry
> Bonds over a career? Over one season?
>
> -Who was worth more wins in his contextual era?
>
>
> Switch and Ruth and Bonds with Wilt and Shaq and you've got the
same
> issues.
• As a major protagonist/antagonist of part of the comparing eras debate of a year ago, I would summarize the contending issues as follows (and this relates to
Message 3 of 6 , Mar 22, 2004
As a major protagonist/antagonist of part of the "comparing eras" debate of a year ago, I
would summarize the contending issues as follows (and this relates to dan's remarks
below).

There were many near-ideological debates on "what is fair to compare?" Is it fair, for
example, to imagine a game between players in Nikes versus those in Converse? I think
on this simplest of criterion, my impression is that most people implicitly were in
agreement that one should not make comparisons based on such physical technology (not
that existing stats would even reveal an advantage - unless there were a lot of slow
adapters to new footwear). After that, there was little by way of agreement in terms of
"holding all else equal" and in fact, there were many basic disagreements on fact itself.
Thus, the view that players today are on average variously bigger, faster, and stronger was
not universal - some contributors disagreeing that this statement in fact represented the
truth. Dean, if I recall correctly, posted some interesting data on the fact that there was a
some period of time. (If that analysis is in BoP, forgive me, I haven't gotten to that chapter
yet.) But no data was offered about quickness or strength. And then there was a
discussion about the effect of race as a potential variable to be used in comparing the
relative strength of various eras.

Now, as regards Dan's regression, height would be something that age proxies would pick
up, I think (and racial composition too, but I don't have deep thoughts on this issue, and it
is a topic that requires them). But there is another important factor that these proxies
would pick up that I believe might result in misleading conclusions. Specifically, my
concern is that they would pick up the factor that I was claiming was hugely important -
namely, improvements in the "dis-embodied technology of offensive play", or, in other
words, better offensive coaching. Specifically, my point was that this factor must be the
overriding factor in the improvement in offensive productivity from the beginning of the
league until the "modern era" of the mid-80s. So, I am not sure that it is proper to
attribute this to player cohorts.

Where the cohort analysis did pop up a year ago was to specifically address the effect of
league expansion on potential talent dilution. And I just can't recall what the bottom line
was on that, except, perhaps that it hadn't yet been reached but that initial inquiries were
very interesting.

*************

--- In APBR_analysis@yahoogroups.com, "dan_t_rosenbaum" <rosenbaum@u...> wrote:
> Here is the procedure I would do if I wanted to get to the bottom of
> this question.
>
> 1. Take a given statistical index (such as PER or Tendex) and
> normalize it to have the same mean in every season.
>
> 2. By season, estimate the fraction of total PER (or Tendex)
> accounted for by each birth year cohort. This will result in about
> 20 observations for each season.
>
> 3. Run this regression using as many seasons as possible - preferably
> going back to the 40s, 50s, and 60s. (The index may need to be
>
> PER_FR = b1*BIRTH_YEAR_DUMMIES + b2*AGE_DUMMIES + e,
>
> where PER_FR is the fraction of total PER (or Tendex) accounted for
> by a given birth year cohort in a given season,
> BIRTH_YEAR_DUMMIES are a series of dummy variables for birth year
> AGE_DUMMIES are a series of dummy variables for age
>
> If the b1 coefficients tend to more positive for later birth year
> cohorts, then succeeding generations of players have been more
> productive than their predecessors. If the b1 coefficients are
> negative, succeeding generations of players have been less productive
> than their predecessors.
>
> I do not have the data to run this regression, but I would be very
> interested in hearing what happens if anyone has the data, time, and
> inclination to run this regression.
>
> A handful of birth year cohorts may get a little bump due to the
> timing of expansion, but overall this method should be do a good job
> of handling the complications brought on by expansion.
>
> This model assumes many things, including that the relationship
> between age and NBA production has not changed over time. Slight
> variations of the basic model above could deal with that issue.
>
> --- In APBR_analysis@yahoogroups.com, "harlanzo" <harlanzo@y...>
> wrote:
> > A while back, we had a long discussion on comparing eras. In that
> > spirit, I saw this article on baseball and eras that was
> interesting:
> >
> > http://baseballprimer.com/articles/cdial_2004-03-18_0.shtml
> >
> > It doesn't resolve the issues and it doesn't translate perfectly to
> > hoops. I also seem to remember someone here (maybe mike g) also
> > looking at standard deviation. But this author makes some
> > interesting arguments and asks some interesting questions (without
> > answering them) that help attack the problem for other angles:
> >
> > -If Babe Ruth in his baseball prime was transported into our modern
> > league, would he be worth more wins than Barry Bonds over a career?
> > Over one season?
> >
> > -If Babe Ruth was born in the same year as Barry Bonds, and grew up
> > in the same environment, would he be worth more wins than Barry
> > Bonds over a career? Over one season?
> >
> > -Who was worth more wins in his contextual era?
> >
> >
> > Switch and Ruth and Bonds with Wilt and Shaq and you've got the
> same
> > issues.
• This methodology that I am proposing would not really be comparing players from different eras. What it would be doing is comparing the guys born in 1951 to
Message 4 of 6 , Mar 22, 2004
This methodology that I am proposing would not really be comparing
players from different eras. What it would be doing is comparing the
guys born in 1951 to those born in 1955 in the years in which they
overlapped, accounting for the differences in ages. And those guys
born in 1955 would be compared to the guys born in 1960 and the guys
born in 1960 would be compared to the guys in 1966 and so on.
Through all of these comparisons, we can get a sense of which cohorts
produced the most. Only by transitivity do we get comparisons across
different eras.

Thus, gradual changes in height, race, offensive technology, etc., I
don't think will be a big deal here. (Remember the statistical index
is normalized to have the same mean in each year.) Early entry and
improved conditioning and medical technologies may give a slight bias
to later cohorts, but there are ways to deal with that in a second
round of regressions.

--- In APBR_analysis@yahoogroups.com, "schtevie2003" <schtevie@h...>
wrote:
> Now, as regards Dan's regression, height would be something that
age proxies would pick
> up, I think (and racial composition too, but I don't have deep
thoughts on this issue, and it
> is a topic that requires them). But there is another important
factor that these proxies
> would pick up that I believe might result in misleading
conclusions. Specifically, my
> concern is that they would pick up the factor that I was claiming
was hugely important -
> namely, improvements in the "dis-embodied technology of offensive
play", or, in other
> words, better offensive coaching. Specifically, my point was that
this factor must be the
> overriding factor in the improvement in offensive productivity from
the beginning of the
> league until the "modern era" of the mid-80s. So, I am not sure
that it is proper to
> attribute this to player cohorts.
>
> Where the cohort analysis did pop up a year ago was to specifically
> league expansion on potential talent dilution. And I just can't
recall what the bottom line
> was on that, except, perhaps that it hadn't yet been reached but
that initial inquiries were
> very interesting.
• See this post (and up and down the thread), from the archives: http://groups.yahoo.com/group/APBR_analysis/message/1578 Rather than get into production rates
Message 5 of 6 , Mar 23, 2004
See this post (and up and down the thread), from the archives:

http://groups.yahoo.com/group/APBR_analysis/message/1578

Rather than get into 'production rates' (and all we could debate
therein), I just checked how player minutes rise or fall from one
season to the next.

In general, a player's minutes diminish as the years go by. Does
this mean the league has generally gotten more competitive?

--- In APBR_analysis@yahoogroups.com, "dan_t_rosenbaum"
<rosenbaum@u...> wrote:
> This methodology that I am proposing would not really be comparing
> players from different eras. What it would be doing is comparing
the
> guys born in 1951 to those born in 1955 in the years in which they
> overlapped, accounting for the differences in ages. And those
guys
> born in 1955 would be compared to the guys born in 1960 and the
guys
> born in 1960 would be compared to the guys in 1966 and so on.
> Through all of these comparisons, we can get a sense of which
cohorts
> produced the most. Only by transitivity do we get comparisons
across
> different eras.
>
> Thus, gradual changes in height, race, offensive technology, etc.,
I
> don't think will be a big deal here. (Remember the statistical
index
> is normalized to have the same mean in each year.) Early entry
and
> improved conditioning and medical technologies may give a slight
bias
> to later cohorts, but there are ways to deal with that in a second
> round of regressions.
>
>
> --- In APBR_analysis@yahoogroups.com, "schtevie2003"
<schtevie@h...>
> wrote:
> > Now, as regards Dan's regression, height would be something that
> age proxies would pick
> > up, I think (and racial composition too, but I don't have deep
> thoughts on this issue, and it
> > is a topic that requires them). But there is another important
> factor that these proxies
> > would pick up that I believe might result in misleading
> conclusions. Specifically, my
> > concern is that they would pick up the factor that I was
claiming
> was hugely important -
> > namely, improvements in the "dis-embodied technology of
offensive
> play", or, in other
> > words, better offensive coaching. Specifically, my point was
that
> this factor must be the
> > overriding factor in the improvement in offensive productivity
from
> the beginning of the
> > league until the "modern era" of the mid-80s. So, I am not sure
> that it is proper to
> > attribute this to player cohorts.
> >
> > Where the cohort analysis did pop up a year ago was to
specifically
> > league expansion on potential talent dilution. And I just can't
> recall what the bottom line
> > was on that, except, perhaps that it hadn't yet been reached but
> that initial inquiries were
> > very interesting.
• But you are ignoring the huge increases and decreases players get in minutes in their first and last seasons. Presumably, what your results show is that the
Message 6 of 6 , Mar 23, 2004
But you are ignoring the huge increases and decreases players get in
minutes in their first and last seasons. Presumably, what your
results show is that the increase in the first season (from zero
minutes) tends to be much larger than the decrease in the last
season (to zero minutes). I am not quite sure how that is helpful
in comparing players across eras.

However, your results got me thinking that rather than some
statistical measure, it might make more sense to simply calculate
cohorts' shares of minutes played season by season.

--- In APBR_analysis@yahoogroups.com, "Mike G" <msg_53@h...> wrote:
> See this post (and up and down the thread), from the archives:
>
> http://groups.yahoo.com/group/APBR_analysis/message/1578
>
> Rather than get into 'production rates' (and all we could debate
> therein), I just checked how player minutes rise or fall from one
> season to the next.
>
> In general, a player's minutes diminish as the years go by. Does
> this mean the league has generally gotten more competitive?
>
>
>
> --- In APBR_analysis@yahoogroups.com, "dan_t_rosenbaum"
> <rosenbaum@u...> wrote:
> > This methodology that I am proposing would not really be
comparing
> > players from different eras. What it would be doing is
comparing
> the
> > guys born in 1951 to those born in 1955 in the years in which
they
> > overlapped, accounting for the differences in ages. And those
> guys
> > born in 1955 would be compared to the guys born in 1960 and the
> guys
> > born in 1960 would be compared to the guys in 1966 and so on.
> > Through all of these comparisons, we can get a sense of which
> cohorts
> > produced the most. Only by transitivity do we get comparisons
> across
> > different eras.
> >
> > Thus, gradual changes in height, race, offensive technology,
etc.,
> I
> > don't think will be a big deal here. (Remember the statistical
> index
> > is normalized to have the same mean in each year.) Early entry
> and
> > improved conditioning and medical technologies may give a slight
> bias
> > to later cohorts, but there are ways to deal with that in a
second
> > round of regressions.
> >
> >
> > --- In APBR_analysis@yahoogroups.com, "schtevie2003"
> <schtevie@h...>
> > wrote:
> > > Now, as regards Dan's regression, height would be something
that
> > age proxies would pick
> > > up, I think (and racial composition too, but I don't have deep
> > thoughts on this issue, and it
> > > is a topic that requires them). But there is another
important
> > factor that these proxies
> > > would pick up that I believe might result in misleading
> > conclusions. Specifically, my
> > > concern is that they would pick up the factor that I was
> claiming
> > was hugely important -
> > > namely, improvements in the "dis-embodied technology of
> offensive
> > play", or, in other
> > > words, better offensive coaching. Specifically, my point was
> that
> > this factor must be the
> > > overriding factor in the improvement in offensive productivity
> from
> > the beginning of the
> > > league until the "modern era" of the mid-80s. So, I am not
sure
> > that it is proper to
> > > attribute this to player cohorts.
> > >
> > > Where the cohort analysis did pop up a year ago was to
> specifically
> > address the effect of
> > > league expansion on potential talent dilution. And I just
can't
> > recall what the bottom line
> > > was on that, except, perhaps that it hadn't yet been reached
but
> > that initial inquiries were
> > > very interesting.
Your message has been successfully submitted and would be delivered to recipients shortly.