Re: talent dilution
- --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
> Obviously, the closer in time two teams are, the easier thecompare
> simulation. But you still want to check those simulations. Can
> anyone simulate the strength of the 1995 Rockets? Most simulators
> wouldn't have had them winning the title. So how do you even
> them with teams in the same year.a
> not quite sure how to interpret this statement - would you "expect"
> simulator to have the rockets win the title more often that seasonthan, say,
> any other team, simply because they indeed did win the title thatyear?....
>This is actually my point. The Rockets did win the title that year,
were not expected to, had a meak record, and a poor point
difference. But they won. They beat everyone else when it
mattered. You cannot take that title away from them. How would you
trust any simulator that says that the Rockets lost in the first
round of the playoff when they actually won it? Yes, I understand
the probabilistic arguments. But, does being an underdog
fundamentally take away from the achievement of winning a title?
They actually had the hardest route to a title of just about anyone,
playing on the road more, playing higher seeds. They did _something_
to win a title. The numbers say it was all luck, but they did it.
Those same numbers would almost never simulate their title. Those
same numbers would have the Rockets losing to every title team since
the inception of the NBA, probably. Does that mean that they are the
worst NBA Champion ever?
I honestly don't know. But I honestly don't know how I'd trust a
cross-generational simulator because there are questions in
implementation that are almost theoretically unanswerable.
I guess I'm debating whether winning a title is so much luck. "Luck"
is what we cannot explain. Sometimes, when you look closer, you can
explain more, so "luck" becomes smaller. Maybe those Rockets matched
up perfectly against all playoff teams, but were weak against
nonplayoff teams. If we do research to show that is true, we all of
a sudden say that it wasn't so much luck, but a crafty construction
of the coach and maybe they could do that against some of the
greatest teams of all time.
It's my opinion, I guess. I'm not saying that we shouldn't do
simulations. I'm saying that I would highly doubt them or not care
- > Is a player who hit .280 in the 1890 American Association (league average:
> .253) really just as good as someone who hit .280 in the 1966 National LeagueThey absolutely are not equal. The talent level of the National League in 1966 was FAR higher than the level of the AA in 1890. The available statistics (most notably the Stephen Jay Gould study, but also the recent Bill James study that he did for the Historical Baseball Abstract) support this. The fact that both leagues happened to hit in the mid-.250s does not change this fact.
> (league average: .256)? That's a ludicrous assertion on the face, and yet it
> seems to be the one you're arguing.
> this has been typical of your responses - you bait for a reply without
> defining your position. you call this "ludicrous" but you do not in any way
> state your side of this. so tell me - if that player is not as good, which
> player is better and what is your reasoning?..
> you are saying they are NOT equal, but totally avoiding the actual debate...
i've read james' first historical abtract (1985,1988) numerous times - its a great read. i've also read all of his single season abtracts prior to 1988 - they too were great reads. but while being a sabermetrician, he is not a statistician. i got this off a baseball stats analysts website (these are not my words) about his recent historical abstract:
"....Is it arrogance that leads James to feel as though he doesn't need to support his opinions? I have no idea, although I think that there's plenty of evidence in the new edition to indicate that the answer to that question is "yes." On the other hand, perhaps there's a more innocent explanation, like James' publisher demanding cuts of some of his analysis in order to make the work more accessible to the average, non-stathead fan. Whatever the reason, this problem makes the playing ratings portion of the new edition pretty much worthless, and puts a dent (maybe even a serious one) in James' reputation as an analyst.
and i found similar "opinions" about james elsewhere....
so having said that, and also having not read james' most recent diatribe about how players are better today than yesterday, are you going to summarize what he said for i and others here to read to support your position, or are you simply going to quote him as someone who supports your views without stating what he said?...
second, i haven't read gould's book but found numerous reviews of the book written on the internet. as far as i can tell from these reviews and written interviews of gould on the internet his basic premise about batting average is that (these are gould's words):
"...When you do that, you realize the following: The average batting average has never changed. It's always been around 260. It fluctuates back and forth, but it stays around 260. And that's not an absolute measure like running a mile or throwing a javelin; 260 is a balance between hitting and pitching. The fact that it's stayed 260 only shows that the balance has been maintained. I say it's been maintained as everyone has gotten better. Hitting's gotten better. Pitching's gotten better. Everything's gotten better. The balance remains the same. Now as everything gets better, the variation shrinks. That's all that happens...."
his contention is that the standard deviation shrinking with time is an indication of an overall increase in talent - for batting average. i read several interviews with him on the internet, and they were all done the year of the book or the year after (1996/1997 - the book being "life's grandeur"), in which he was asked about other baseball records and their possibility of being broken, and he was also quoted as saying the season home run record wouldn't be broken. well we've seen it broken 6 times in the past 5 years, by 3 different people...
we "baseball experts" all know how worthless batting average is as an indication of a baseball players true "worth". so tell me - what did gould say about slugging percentage? bonds just broke ruth's single season record held since the 1920s. is this an indication then that players were better in between when these marks were set? because no one got close to them and thus the standard deviation for that stats for the league was less variable and thus by gould's definition the league was better?...
tell me - did gould do a study of numerous aspects of baseball, and compare standard deviations thru time - or just batting average? everything i can see from reading reviews of his book was he only looked at batting average - is this true? if i'm wrong and he did, please inform me. but if he did just one, batting average, how do we know he he didn't use it as an example simply because it backed up his theory?...
> if you are asserting that the 1966 NL player is better, you are totallySo what you're saying is that there is no way to compare, then? Thanks for agreeing with me.
> forgetting to consider the logic of not just what would the 1890 hitter do in
> 1966, but what the 1966 hitter would do back in 1890 - under the conditions
> of the game as it was played in 1890. did they even use mits back then (if
> they did they were certainly of worse quality than the mits of 1966)? could
> the 1966 player even field with the type of mits they used back then? back in
> 1890 very few new balls were used in a single game, in 1966 after any ball
> went into the stands a new ball was put into play. based on that alone one
> could easily say that the 1966 .280 hitter would have a tougher time hitting
> under the conditions in 1890....
sure there is - the league batting avg was the same and the players hit the same - you know what i am going to say...they are the same (not knowing their OB%s and SLG%s, of course)...
was, quote "....the standard deviation has decreased and thus the variabilyty decreasing is indicative of overlal better play..." your only reason for saying the 1966 hitter was better than the 1890, or do you have other "evidence"...
fyi - as one who is a geologist that has the utmost respect for gould and his punctuated equilibria theories, he could stand a good lesson in sports. i notice he hasn't (and i guess won't at this point) done any similar analysis forhockey, basketball, or football...
back to basketball -
you can do all the standard deviations for basketball FG% you want, but if you want to prove your contention that the league has gotten better with time because variability has decreased, don't you think you'd have to look at many more or even all the different stats, not just FG%?...
> i guess the 1962-63 league overall free throw shooting percentage of .727 is
> also not equal to the 1998-99 league overall free throw shooting percentageWhich can be easily explained away by two things: one, free throw shooting as a skill has not significantly improved since the 1960s because there was no reason for it to do so,
> of .728 because of the "different" defenses too....
come again??? there was no reason for FT shooting to improve?? sorry - but there is every reason for FT shooting to improve, just as there is every reason for FG shooting to improve. the only way one wins games in the nba is by scoring as scoring is the parameter used to determine wins and losses, and if you improve your shooting %s (free throws and shot attempts from the field) you stand a great chance to score more than if you do not improve them - no matter how you improve them.....
the fact that FT% has not improved, rather neither increased nor decreased significantly since 1954-55, yet players have come and gone, is indicative of a similar process that has not changed (the different values for overall yearly FT% being random fluctuations but always near the 74.8% average), despite 2586 different players having played in the league in 47 years since 54-55...
and two, any small increases that were made in this particular skill over that time have been counterbalanced by selection bias; because defense and other non-shooting-related skills are valued more highly now than they were in the past,
"defense and other non-shooting skills valued more highly now than in the past"? a pure assumption not based in any fact whatsoever......lets see some evidence to back up this pure assumption...
Okay, some numbers. Here are the standard deviations of FT% and 2ptFG% since 1947. My
populations include all players with more than 30 attempts.
Season FT% 2FG%
1947 9.6% 4.3%
1948 9.2% 3.9%
1949 8.9% 5.2%
1950 8.3% 4.8%
1951 8.3% 5.6%
1952 8.3% 5.3%
1953 7.7% 5.0%
1954 8.2% 4.8%
1955 7.8% 4.8%
1956 7.9% 4.2%
1957 7.3% 3.9%
1958 8.8% 4.3%
1959 8.0% 4.1%
1960 9.6% 4.8%
1961 9.3% 4.8%
1962 8.8% 5.2%
1963 9.6% 5.4%
1964 9.0% 5.2%
1965 9.7% 5.1%
1966 10.0% 4.4%
1967 10.1% 5.5%
1968 10.4% 5.6%
1969 10.1% 5.4%
1970 7.9% 5.4%
1971 9.5% 5.2%
1972 10.5% 5.5%
1973 9.2% 6.2%
1974 9.0% 4.3%
1975 9.8% 4.2%
1976 8.9% 4.9%
1977 8.4% 5.2%
1978 8.3% 4.5%
1979 9.5% 5.0%
1980 9.2% 5.1%
1981 9.3% 5.9%
1982 9.3% 5.1%
1983 9.0% 5.3%
1984 9.0% 5.7%
1985 8.7% 5.5%
1986 10.4% 5.4%
1987 9.1% 5.1%
1988 9.1% 5.4%
1989 10.0% 5.0%
1990 10.0% 5.9%
1991 10.1% 5.2%
1992 9.0% 5.3%
1993 9.0% 4.9%
1994 9.6% 5.3%
1995 10.3% 5.5%
1996 10.3% 5.4%
1997 10.3% 5.8%
1998 11.0% 5.5%
1999 11.4% 5.7%
2000 10.3% 5.4%
2001 10.8% 5.9%
2002 10.7% 5.0%
Looks to me like a general increase in variation. Whether that constitutes "dilution"
is another matter. The increase is even more apparent if you only look at the top and
bottom 30 players in each category (a reasonable sample, no?).
if you see a general increase it is ever ever so slight - i see similar values from the early 1950s and the late 1990s and variations in between, with no general trend, but random fluctuations...
i would however love to see this for other stats to see if any general trends do appear...