## 1715Re: [APBR_analysis] Re: Cross Generational Simulating/Comparisons

Expand Messages
• Jan 27, 2003
• 0 Attachment

Okay, let's take things point by point and be done with it?

if you tire of the exchange of ideas concerning this discussion, once you have already made your position clear, but do not wish to be bothered to be asked to defend your postion with stats in a stats analysis group, then maybe this wasn't the forum for you to "...add a few pennies worth of thought on the issue of inter-temporal comparsions of NBA performance..."...

Me:

the superiority of the average team in the early 80 to its 60s counterpart was on the order of 10 to 12 points per game (at a modern game pace).  This is equivalent to how much better the best Bulls team was to the league average.

Bob:

not sure what these numbers mean or where they are coming from....how did you arrive at this or derive these numbers?...

so let me get this straight, you are saying you have statistically come up with a way to show that if the average NBA team of the early 1980s (about 109 pts/g) played an average team from the 1960s (115 pts/g) or early 1960s (almost 116 pts/g), the avg 80s team would beat the average 60s team, on average, by 10-12 points, normalizing for modern times. the 95-96 bulls outscored their opponents that year by just over 12 pts/g. so you are saying the difference in ability to win of an average early 1980s NBA team versus/compared to an average 1960s or early 1960s is the same as the 95-96 bulls to the suns or hornets of 95-96 (.500 teams that year)...

is this correct?...

Me again:  The calculation I referred to was made by computing a measure of offensive (or equivalently defensive) productivity that I called "points per common possession" versus "common possessions per game".  In the former case, the computation is average points scored in the numerator and average common
possessions in the denominator.  Common possessions are, by definition and construction, the same for both teams in a particular game (hence attractive for purposes of comparison and analysis) and are basically calculated by subtracting the number of offensive rebounds from the conventionally defined number of possessions a team enjoys in a game.

fine...so fill in the blanks below so we can compare...or show us your numbers that show how you arrived at your result...

pts/      sec/      pts/
year        FG%      g      poss     poss
5960      .410     115.3
6061      .415     118.1
6162      .426     118.8
6263      .441     115.3
6364      .433     111.0
6465      .426     110.6
6566      .433     115.5
6667      .441     117.4
6768      .446     116.6
6869      .441     112.3
6970      .460     116.7
7071      .449     112.4
7172      .455     110.2
7273      .456     107.6
7374      .459     105.7
7475      .457     102.3
7576      .458     104.3
7677      .465     106.5

pts/
48min
7778      .469     107.7    13.4      1.000
7879      .485     109.8    13.5      1.029
7980      .481     108.5    13.8      1.041
8081      .486     107.5    14.0      1.043
8182      .491     107.9    14.1      1.058
8283      .485     107.9    13.8      1.037
8384      .492     109.2    14.0      1.063
8485      .491     110.2    13.9      1.066
8586      .487     109.5    13.9      1.059
8687      .480     109.2    14.1      1.067
8788      .480     107.6    14.3      1.065
8889      .477     108.5    14.1      1.062
8990      .476     106.3    14.4      1.065
9091      .474     105.5    14.5      1.063
9192      .472     104.5    14.7      1.067
9293      .473     104.5    14.7      1.066
9394      .466     101.0    15.0      1.049
9495      .466     100.6    15.3      1.069
9596      .462      98.8     15.5      1.063
9697      .455      96.2     15.8      1.053
9798      .450      94.8     15.7      1.036
9899      .437      90.9     16.0      1.009
9900      .449      96.9     15.3      1.030
0001      .443      94.0     15.6      1.020
0102      .445      94.8     15.7      1.034

DeanO has pointed out that "...this trend looks about right, even if my calculations have slight differences...". so if we are defining offensive productivity as pts/poss, i see similar pts/poss numbers with varying FG% and pts/g and game pace for 77-78 thru today. if your number show a different trend for prior to 77-78, lets see them...

For missing data series internal to the sums (that is for turnovers and offensive rebounds in the time periods in question) as I noted before, I extrapolated values

fine - you made some assumptins/estimates for missing parameters...

and conducted a bit of sensitivity analysis to see if reasonable deviations from
assumed values would overturn results.

They could not. (As noted, the offensive rebounding percentage was remarkably stable over succeeding time period.)

Finally, the "12 points per game" was, if I recall correctly, taking he observed increase in productivity and prorating it at the modern end-point game pace. (Which is to say it would have been higher using the earlier endpoint where there were more
possessions per game, but no more meaningful.)

so lets see how you came up with this using the your numbers....

Then to repeat, "what the numbers mean" is that teams were having fewer possessions (defined to encompass all they would do with the ball before it would revert to the opposing team) per game yet they were scoring more points.

lets see some numbers to back this up...

These are the off-setting trends that allow no other explanation (yet offered) than the offenses were - on net - getting better, much better.

this is how you defend your position, by stating that the results allow "...no other explanation..."?

Me:

Just looking at the average offensive productivity over the time in question, what one sees is that it improved dramatically (the 12 points per game above).

Bob:

would love to "look" at it but you're not showing anything...

Me again: As noted, I did not have these data on file or in hand, as they were done long ago.  Perhaps I will spend the time to recalculate them.  In the meantime, if you have these data at your fingertips, feel free to make the calculation.

feel free to explain it...

There is no ambiguity as to what they say.  The trends are almost strictly monotonic as I recall.  Pick any starting year in the mid to early 60s (or late 50s for that matter) and any end point in the late 70s to early 80s.  The numbers speak clearly.

show me the beef.....

Me:

Now, those who believe we stand on the shoulders of giants might be inclined to say that defenses got worse as opposed to offenses having mproved.  Is this view tenable?  The answer is "no" because the trend is accompanied by a dramatic slowing of the game.

the answer is not "no", the numbers from 77-78 thru today show that the slowing of the game does not correlate well with pts/poss, on the contrary from 77-78 thru today the values fluctuate from 1.000 to the early 90s there is an increase to almost 1.070. 77-78 and 98-99 show very close pts/poss numbers of 1.000 and 1.009 yet the game pace is significantly different. based on that why should anyone believe your pre-77-78 numbers without seeing them?....

This slowing, all else equal, implies that offensive productivity should have decreased. .  Why?  It either means fast-break baskets were being reduced or that the half-court offenses were having to work harder for shots.  But offenses improved, dramatically...

sorry, don't see it....

Bob:

here are some number for you to chew on:

pts/      sec/      pts/
year        FG%   48min   poss     poss
7778      .469     107.7    13.4      1.000
7879      .485  109.8    13.5      1.029
7980      .481     108.5    13.8      1.041
8081      .486     107.5    14.0      1.043
8182      .491  107.9    14.1      1.058
8283      .485     107.9    13.8      1.037
8384      .492     109.2    14.0      1.063
8485      .491     110.2    13.9      1.066
8586      .487     109.5    13.9      1.059
8687      .480     109.2    14.1      1.067
8788      .480     107.6    14.3      1.065
8889      .477     108.5    14.1      1.062
8990      .476     106.3    14.4      1.065
9091      .474     105.5    14.5      1.063
9192      .472     104.5    14.7      1.067
9293      .473     104.5    14.7      1.066
9394      .466     101.0    15.0      1.049
9495      .466     100.6    15.3      1.069
9596      .462      98.8     15.5      1.063
9697      .455      96.2     15.8      1.053
9798      .450      94.8     15.7    1.036
9899      .437      90.9     16.0      1.009
9900      .449      96.9     15.3      1.030
0001      .443      94.0     15.6    1.020
0102      .445      94.8     15.7      1.034

sec/poss is the avg time per team possession, pts/poss is the avg points scored per team possession. you can see a general trend in a slower game thru time secs/poss) but no such similar trend in offensive productivity...

as you can see total points scored and game pace have next to nothing to do with offensive productivity per team possession. 1978-79 and 1999-00 have the same pts scored per team possession (1.029 and 1.030) yet vastly different game paces, vastly different FG%s, etc. if you should doubt these pts/poss numbers, DeanO or any other of the stats gurus in this discussion group can verify (or for that fact deny) them....

since i can't accurately calculate pts/poss prior to 77-78, i can't present similar evidence. but the above clearly shows that over a 24 year time span of the NBA that offensive production per team possession is not directly correlated to game pace (slowing down the game or speeding it up)......

Me again: I have chewed on your numbers,and I am happty to tell you what flavor impressions they leave.  First, I should say that the food is not what I ordered.  You pick as a starting point of your series, the end point of the trend I asserted.

the stats are available from 77-78 on, you haven't presented your numbers nor estimates for data missing prior to that. are we to simply take your word on it?...

No matter.  Let's talk about what you present.  First of all, I am not sure how you
calculate possessions, but never mind.

i don't know how you or DeanO do either, but apparently his and my numbers are similar...

As I recall my rendering of the data in the timespan you present (and again, these are somewhat distant recollections) once you factor out the increasing weight of the three-point shot in augmenting average offensive productivity (as I noted previously, in passing, it took a remarkably long time for teams to figure out how to optimize the three point shot in the context of team offenses) what the 80s represents is a period of essential "stagnation" (neither offenses or defenses gaining at the expense of the other) followed by a decline beginning in the very late 80s or early 90s.

this i do not see...i see similar pts/poss for early 1980s to mid 1990s...and similar numbers for late 70s and late 90s early 2000s.....if you have different numbers for offensive productivity lets see them to back up your statements...

Now, nothing you present overturns this impression of mine.

surprising...

And as to these changes in the 90s, taking the data you present above (let's say '91-'92 to '01-'02) what they say is that there was a 6.8% increase in seconds per possession and a 3.1% decrease in points per ossession.

from 89-90 to 96-97 i see simlar pts/poss numbers but a game consistently slowing down...no correlation there.....i see a lowering of pts/poss in the next 5 years but it fluctuates, no discernable pattern...

At 91.72 possessions per game in '01-02 this means teams were scoring 3 points per game less than the decade prior.

in 01-02 the 1.034 pts/poss is similar to the 82-83 pts/poss of 1.037, yet they scored
more points per game in 82-83 (108 pts/g vs 95 pts/g) simply becuase the game pace was faster...

If this represents a net improvement in average defense over the decade, Is there anyone who doesn't think that that is significant?

looks to me like a fluctuation of pts/poss over a 24 season period, despite a general slowing of game pace...

in 78-79 i see .485 FG%, 110 pts/g, and 13.5 sec/poss with 1.029 pts/poss. in 99-00 i see .449 FG%, 97 pts/g, 15.3 sec/poss, but the same pts/poss of 1.030....are we not looking at the same data?...

Again, for context, the average NBA champion is only about 6 to 7 points (again pace dependent, blah, blah, blah) than the average team in the league.  So 3 points is close to half the way from mediocrity to championship status.

not sure what you point is again...

Me:

And the only explanation I can deduce is that offenses were disciplining themselves (through improved coaching I surmise) and expunging bad shot selection.  Now, as
> I also said before, I expect that defenses were also improving during this time so that the 12 point estimate is in fact a low estimate of the actual improvement in these two decades.

Bob:

have no idea where this is coming from...

Me again: About what have you no idea? Is it the deduction that the offsetting trends of game pace and offensive productivity in the 60s and 70s

show us some data...

necessarily mean (again, absent an alternative explanation) that there is a net offensive improvement in the average NBA team over time?  If so, I don't know how to explain it any more clearly.  Perhaps someone else can help.

perhaps...

Or is it the point that the "12 point" estimate is a low-ball?  This simply means that all the offsetting trends can identify is the net offensive improvement.  Now one can believe then that the relentless assault of offenses on defenses were met with no
countermoves on the latter's part.  In this case, the estimate is not low, but this of course seems rather hard to believe, hence the low-ball claim.

Me:

And to continue.  If one accepts the general and continuous improvement of the game over the two decades mentioned, then it becomes arbitrary to imagine that progress stopped then.  The reasonable prior belief is that the slow decrease in offensive productivity after this point is the result of defensive improvements, rather than some technological
retrogression.

you've made a number of generalizations...show us when - what years - the trend increased, with yearly numbers, when it stopped, when it decreased, etc....may be obvious to you, but you are not being definitive. the above 77-78 to 01-02 numbers show otherwise...

unless of course you are not defining offensive productivity with pts/poss but something else...

Bob:

what slow decrease in offensive productivity? i see similar offensive productivity in 77-78 to 80-81 as i see in 97-98 to 01-02, with the higher productivity per poss of the 1980s being higher and fairly constant...

technological retrogression?? like what, bring back the jump ball after every score??...

e again: Again, I refer you to the fact that the endpoints you chose for your analysis were not the period I was speaking of.

actual data for ast, st, bs doesn't exist for prior to 73-74 (except ABA) and TO prior to 77-78. you say you've made estimates for seasons prior to this but don't show them, and present your findings as dogma...

But additionally, to repeat somewhat, your selection is rather precious, as there is a lot going on over the 15 year period you show.  Productivity goes up and down again.

24 year period - while there is a general increase in time per poss, or decrease in game pace - no definitive pattern for pts/poss or offensive productivity versus game pace for this 24 year period is evident...and you want us to believe - that without you showing the data - that there is a trend for seasons prior to this...

And as for the term "technological retrogression", I am sorry that it confused you.  I was simply pointing out (regarding the 90s decrease in offensive productivity)

you mean 97-98 to 01-02 don't you?...

that the alternative to believing that defenses had on net improved relative to offenses, one must believe that somehow offenses got worse, somehow forgetting to do as well in a given year what they could do a few years past (that is to say that their "technology" - the means by which they produce points - had retrogressed - as opposed to progressed).  Clear?

if you say so...

Me:

And finally, I made what I think is a very persuasive argument for the superiority of the trained athlete in more recent times.  In the form of a question, if there are not some return to his efforts, why is he killing himself in the gym?

Bob:

a nice philosophical arguement but not pertinent for stats analysis..

in 73-74 nate archibald was making \$400,000 a year. the average salary (not mean but average) at the time was - i'm guessing - 1/8 to 1/4 of that, somewhere in between \$50,000 to \$100,000. the average american income for a working adult at that time was probably about \$5,000 - \$15,000 a year. for that kind of scratch are you saying athletes didn't keep themselves in shape, in condition, to play pro hoops? that that kind of dough wasn't enough incentive to also kill themselves in the gym in the early 1970s like athletes kill themselves today? i don't know the average salaries in the early 1960s, but i know chamberlain and russell were arounf \$100,000, and i'll bet the league avg salary was well above the average working adult salary at the time - and this doesn't take into account the salary war started by the ABA in the late 60s early 70s.....i'm also guessing the average player trying to make the NBA in the early 60s (let alone today), if he didn't make it, went on to become a doctor or lawyer with a high salary...

just becuase the athlete of the 1960s didn't have access to the weight training they had today and all the equipment doesn't mean they didn't do whatever they could to be inshape to play at the pro level...

Me again:  Are you kidding me?  Statistics are a useful and indispensible tool for divining truth, but they are not the sole means; no matter if they are your fetish of this paragraph in this argument.  As for numbers, the point was made, and made
unanswered, that in every athletic endeavor where competition is additionally against a fixed measure (say 100 yds run, or pounds lifted, or spins in air completed) there has been a relentless yet decreasing improvement over time.

so absolute records are broken against static measures. basketball is a relative measure, player versus player, like baseball or any of the other team sports. the arguement is pointless. are the players better athletes - better bodies, better nutrition - yes im sure, but better basketball players? if no one breaks wayne gretzky's records in the next 100 years, does that mean hockey had peaked in athletic performnace in the early to mid 1980s??

You seem to plead specially for basketball's exception when it is observed that its athletes are bigger and stronger.  And as for the question I pose regarding why a rational athlete would devote excruciating hours in a gym to no benefit, it is also pertinent for the analysis.

the assumption that pro basketball athletes train harder and longer now than 10, 20 or 30 years ago is ridiculous. players back in the 60s and 70s and late 50s ran just as much in practice, played 35+ or 40+ min/g just like they do today, at an even faster paced game...they all saw the benefit of good training, to make an excellent living. every generation has its hotplate williams and oliver millers, as they have their dolph schayes and john havliceks and john stocktons and karl malones...

Me:

There is the "proof".  I look forward to any alternate explanation of these facts and inferences.

Bob:

see above...

Me again: See above indeed.

Me:

All this aside, I should say that I have nothing but the highest regard for the notion of using simulations for analyzing the game of basketball.  To my mind, they are the best method for understanding the effect of marginal changes in game variables.

However, they are not useful - only highly deceptive - in determining the competitiveness of NBA teams from different eras.

no more deceptive than someone stating teams from one era are 10-12 points better than another era without showing the numbers...

Bob:

words from the wise i guess....

Me again:  see above...

i'd love to believe your statements, but you don't show how you arrive at the numbers...

bob chaikin
bchaikin@...

• Show all 29 messages in this topic