Do younger rookies have longer careers? (MikeG, 2008)

[Crow]

page 1 of 4

Author Message
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Tue Apr 22, 2008 10:27 am Post subject: Do younger rookies have longer careers? Reply with quote
Can a guy who jumps into the draft at age 19 expect to play 3 years longer than he would if he'd wait until he's 22? Historically, no.

Utilizing this page:
http://www.basketball-reference.com/fc/ ... i?id=xvlyr

... I looked at players who, through age 24, had a career PER of 19.5 or better (There have been 87). It would be a non-random sample to just compare all players who have entered the league at 19, at 20, etc; since generally those coming in younger have more potential for greatness. By checking just those with early star power, we are comparing more similar players.

Eliminating those who only played in the ABA, and then selecting those whose careers have concluded, we have most of the alltime top-50, plus some early flameouts, plus some more less-heralded. I consider Webber, Francis, and Abdur-Rahim to have concluded.

Actually, I'm defining 'concluded' as a player's final 1000-minute season. Seasons of <1000 'don't count'. This affects 'career length' and 'min/yr' a bit. Here's the breakdown of 57 concluded careers, by age of entry:
Code:
SS# rk Yr1 age1 ageL YrLst yrs Min min/yr
10 51 1981 20 32.2 1994 13.2 32,821 2460
15 34 1973 21 32.7 1985 12.7 28,701 2344
25 46 1970 22 33.0 1981 12.0 30,006 2509
7 53 1966 23 33.7 1977 11.7 30,366 2557
SS# is sample size.
rk is average rank in the career PER of players <25 years old.
Yr1 is average rookie season. Note how rookies have gotten younger
age1 is age in rookie season, defined at b-r.com
ageL is last 1000-min-season age
YrLst is that year
Min is career, regular-season, NBA+ABA
min/yr, from rookie year (regardless of min.) thru last 1000-min year

Historically, though entering 3 years sooner, 20-year-old rookies only play 1.5 years more than their 23-year-entering counterparts. The progression is pretty smooth: Each year you cut off of college, you lose half a year at the end of your pro career.
Total minutes are even closer; perhaps because some young rookies do not even play 1000 minutes.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Tue Apr 22, 2008 11:02 am Post subject: Reply with quote
I guess I forgot to correct for shorter schedules in the olden days. Since there are uneven distributions of player-seasons before 82-game seasons, and I don't want to fart around with that, I'm just deleting players who started before 1960. Now it looks like this:
Code:
SS# rk Yr1 age1 ageL YrLst Min min/yr yrs
10 51 1981 20 32.2 1994 32821 2460 13.2
12 36 1979 21 33.1 1991 30350 2315 13.1
19 49 1975 22 33.6 1987 31623 2483 12.6
6 56 1968 23 34.2 1980 32896 2701 12.2

Now it seems there's almost no advantage to coming out early: Lose 3 college years to gain one pro season.
Moses came out at age 19, and is included in the 20's group.

Next: current careers contradict historic data.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Chicago76



Joined: 06 Nov 2005
Posts: 98


PostPosted: Tue Apr 22, 2008 11:53 pm Post subject: Reply with quote
I'd be curious to see your sample bifurcated into two groups: the guys entering around 83/84 and before and those entering after. There are a few reasons for this:

1-The pre-merger era had a decent number of players who came out early and played very well at in the divided league system. Neither league was as strong as the post-merger setup and a lot of these guys didn't fare nearly as well 3 or 4 years into the post merger era. Maybe they were too assured of their status as stars to continue to work hard and the increased competition got the best of them (a McGinnis for example).

2-The late 70s early 80s drug issues. Put money in a younger guys pocket in this era and his choices might be more suspect (the Haywood/Thompson). This is the argument people use for age limits today, but the fact is young guys are much more sophisticated in their choices today than they were 30 years ago. The money really began to take off at this point too.

3-Sneaker design. Mid-sole support and better footwear technology really started around 84/85 with Converse and shortly thereafter w/ Nike. Guys coming out in 83 or earlier are playing heavier minutes with inferior footwear than players of a the same age who played 4 years in college. If you look at injuries, they hit their peak right around the early/mid 80s as the game continued to get faster/bigger and sneaker technology had not yet caught up.
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Wed Apr 23, 2008 7:54 am Post subject: Reply with quote
Sneaker technology. I hadn't thought of that factor. But drugs (and other distractions) have continued to plague star players: Kemp, etc (rather not name names).

More money may = more distraction. Then I guess one can afford a personal manager, trainer, chef, etc. And in another thread, I'm seeing evidence that the mid-80's onward are the most competitive in league history. So that may skew players' career minutes a bit.

Larry Bird, 1980, was the last 'star' player to enter after age 22. Well, Andre Miller (2000). Nowadays, all stars enter before 22. Check out the current stars-by-24 who were 22-yr-old rookies: Grant Hill, Vince, Yao, and Wade. Yikes! Hardly an ironman group.

Yet of 18-19-yr-old (current) rookies (7), the least-durable has been McGrady (2389 min/yr). Even including their development years, they average 232 minutes more than those who rookied at 22.

The jury is still out on how these guys' careers will wind up. Of retired players who've rookied since '84, it seems 21 is the optimal age for entry. The samples are very small.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Chicago76



Joined: 06 Nov 2005
Posts: 98


PostPosted: Thu Apr 24, 2008 1:23 am Post subject: Reply with quote
Mike G wrote:
Sneaker technology. I hadn't thought of that factor.


It's kind of interesting when you look at minutes played and games missed over the last 30 years. As the game continued to get faster (not necessarily measured in pace, just movement) and guys on average got stronger, injuries generally trended up until the mid-80s. I forget what year it was, but Converse came out with a shoe with mid-sole support, then the Weapons a year or so later. Of course, Nike had the Air Jordan thing a year or so after that. Suddenly, injuries were down on average. That was the technology leap from Chucks and Keds to Cons and Nike was the big leap. Now we're just seeing little refinements, and not surprisingly, injuries have trended up.

Physics is catching up again. You can ask guys to run faster and be stronger, but medicine, training, and footwear can only help so much. If we knew then what we do now, who knows what would have become of the Waltons and Moncriefs of the world.
Back to top
View user's profile Send private message
gabefarkas



Joined: 31 Dec 2004
Posts: 1313
Location: Durham, NC

PostPosted: Mon May 05, 2008 10:26 am Post subject: Reply with quote
Mike G wrote:
I guess I forgot to correct for shorter schedules in the olden days. Since there are uneven distributions of player-seasons before 82-game seasons, and I don't want to fart around with that, I'm just deleting players who started before 1960.


Maybe I'm just having a brain-fart, but I'm not sure what you mean by this. Can you elaborate?
Back to top
View user's profile Send private message Send e-mail AIM Address
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Mon May 05, 2008 12:43 pm Post subject: Reply with quote
Oy, that's a very poorly constructed sentence. Well, I'm referring to shorter schedules (less than 82G/Yr) in the '50s and, to a lesser extent, in the '60s. This plays havoc with players' career-length, in both games and minutes.

'Uneven distributions' refers to the fact that younger rookies are more heavily represented in later decades. So, by including predominanty older rookies from the '50s -- who played fewer games and minutes per year, as well as tending to play fewer years -- the sample was skewed.

By eliminating those 10 old-timers (avg rookie age 21.8; career 10.1 yrs * 2286 min), I removed some older/briefer careers. So comparatively, younger entry looks even worse without this part of the sample.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Mon Apr 06, 2009 1:03 pm Post subject: Reply with quote
Resurrected Thread Warning!

This article came out a few weeks ago:
http://sportsillustrated.cnn.com/2009/w ... /?bcnn=yes

The aging of the NBA's prodigies
by Steve Aschburner
Quote:
The issue: Did the players who turned pro directly out of high school from 1995-2005 help themselves to four extra NBA seasons, or did they simply start drawing down early from a finite account of available minutes?

The question might actually be: Are marathon minutes even more costly to teenage players, such that their careers are even shorter?

Quote:
In Boston's second game this season, Garnett became the youngest player to participate in 1,000 NBA games, reaching that threshold at 32 years, 165 days.

Also speculations on Dwight, Kobe, Jermaine, McGrady.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
NickS



Joined: 30 Dec 2004
Posts: 384


PostPosted: Mon Apr 06, 2009 5:02 pm Post subject: Reply with quote
Mike G wrote:
Resurrected Thread Warning!

This article came out a few weeks ago:
http://sportsillustrated.cnn.com/2009/w ... /?bcnn=yes


That's a good article.

It's a subject that I've wondered about, myself and, while the article is mostly anecdotal in it's approach, it nevertheless feels substantial and has a number of interesting examples.
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Tue Apr 07, 2009 8:31 am Post subject: Reply with quote
When I did this study last year, there were a few younger entries who had recently concluded their careers; and several -- Kobe, KG, etc who seemed to be iron men. One of those 2 has perhaps hit a wall, though.

It will be interesting to revisit in a few years. It seems to have been suggested that a mandatory 2nd year between high school and NBA is being debated?
If so, does that mean it's now a consensus that the one-year delay is declared a success?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Wed Sep 30, 2009 10:26 am Post subject: Reply with quote
Revisiting the topic, and I'm still not aware of any other study of it.

I've looked at players who have been selected to one or more All-Star games, and who entered the NBA from 1992 to 2003. At what age did these players peak?

'Peak' is defined with equal weight given to a player's:
- Best PER
- Most WinShares
- - If Offensive WS peaked in a year other than total WS, the average of the two is used.
Example: Rashard Lewis' PER tops (at 20.7) in 2007. His WS tops (at 9.9) in 2008; but his OWS peaked sharply (8.3) in 2006. His 'overall' peak is thus fudged to 2007.

Tony Parker is similarly adjudged to have peaked in '08. His WS peaked in '07; PER was highest last year ('09); he is the only player in the study with either measure peaking so recently.

Sorted by their age at entry to the NBA, there's a distinct cutoff between players who first appeared by age 20, and those 21 or older. Players who entered at 18, 19, or 20 have an average peak at age 25.4; those who enter at 21 or older have peaked at 27.1 years, on avg.

This study involves 72 players. No player entering 2004 or later can be (I guess) supposed to have peaked.
http://www.basketball-reference.com/fc/ ... i?id=tCizM

Code:
Yr1 entry allstar YOB Peak Min PkYr PkAge
18 1998 McGrady 1979 2003 27462 6 23
18 1997 Kobe 1978 2006 34531 10 27
18 1997 Jermaine 1978 2004 22459 8 25
19 2002 Parker 1982 2008 20335 7 25
19 2002 Arenas 1982 2006.5 16175 5.5 23.5
19 1999 Lewis 1979 2007 26855 9 27
19 1997 Marbury 1977 2003 31891 7 25
19 1996 Garnett 1976 2004 39635 9 27
18.6 1998.5 2005.2 27417 7.7 25.3

20 1993 Shaq 1972 2000 39926 8 27
20 1997 Walker 1976 2002 31531 6 25
20 1994 Webber 1973 2000.5 30847 7.5 26.5
20 1999 Nowitzki 1978 2006 30693 8 27
20 1997 Shareef 1976 2000.5 28882 4.5 23.5
20 2000 Brand 1979 2006 24421 7 26
20 2000 BDavis 1979 2007.5 23826 8.5 27.5
20 2002 JJohnson 1981 2006 22908 5 24
20 2000 Artest 1979 2006.7 21037 7.7 26.75
20 2002 Kirilenko 1981 2004.5 17309 3.5 22.5
20 2003 Amar'e 1982 2006.5 14848 4.5 23.5
20 1998.8 2004.2 26020 6.4 25.4
25.4
- - - - - - -

Yr1 entry allstar YOB Peak Min PkYr PkAge
21 2003 Boozer 1981 2007.5 14019 5.5 25.5
21 2002 Gasol 1980 2008 20821 7 27
21 2001 Redd 1979 2006.5 18710 6.5 26.5
21 2000 Hamilton 1978 2004.2 24724 5.2 25.25
21 2000 Marion 1978 2006 28020 7 27
21 1999 Pierce 1977 2005 30525 7 27
21 1999 Carter 1977 2001 29271 3 23
21 1999 Stojakovich 1977 2004 24402 6 26
21 1998 Duncan 1976 2003 33138 6 26
21 1998 Billups 1976 2007 26876 10 30
21 1997 Iverson 1975 2007 36719 11 31
21 1997 Allen 1975 2001 35099 5 25
21 1996 McDyess 1974 2000.2 25049 5.2 25.25
21 1996 Stackhouse 1974 2001 28500 6 26
21 1996 Rasheed 1974 2001 34167 6 26
21 1995 Kidd 1973 2000 41155 6 26
21 1995 Juwan 1973 2001.7 34188 7.7 27.75
21 1995 GRobinson 1973 2001 25346 7 27
21 1994 Mashburn 1972 2003 22762 10 30
21 1992 KAnderson 1970 1996.5 25868 5.5 25.5
21 1992 Brandon 1970 1996 21545 5 25
21 1997.3 2002.9 27662 6.6 26.6


22 2003 Yao 1980 2005.5 15727 3.5 24.5
22 2003 Butler 1980 2008 17715 6 27
22 2001 Magloire 1978 2003 13730 3 24
22 2000 Francis 1977 2001 21632 2 23
22 2000 Szczerbiak 1977 2004 20052 5 26
22 1999 Jamison 1976 2003.7 29354 5.7 26.75
22 1998 Ilgauskas 1975 2004 20481 7 28
22 1997 Nash 1974 2007 29012 11 32
22 1996 Finley 1973 2000.2 37288 5.2 26.25
22 1996 Ratliff 1973 1999 19621 4 25
22 1995 Hill 1972 1997.7 28122 3.7 24.75
22 1994 Van Exel 1971 2000.2 28969 7.2 28.25
22 1994 Houston 1971 2002.2 28311 9.2 30.25
22 1994 Penny 1971 1996 23711 3 24
22 1994 Baker 1971 1998 25737 5 26
22 1993 Mourning 1970 2000 25975 8 29
22 1993 Sprewell 1970 1997 35270 5 26
22 1992 SSmith 1969 1998.5 28856 7.5 28.5
22 1992 LJohnson 1969 1996 25685 5 26
22 1992 DDavis 1969 2001 29606 10 31
22 1996.3 2001.3 25242 5.8 26.8

23 2003 Okur 1979 2006 15658 4 26
23 2001 KMartin 1977 2003.5 17806 3.5 25.5
23 1995 EJones 1971 1999 32778 5 27
23 1993 Gugliotta 1969 1998 23556 6 28
23 1993 Laettner 1969 1997 25760 5 27
24 1994 Cassell 1969 2003.7 29811 10.7 33.75
24 1992 Gatling 1967 1995.5 13760 4.5 27.5
25 2003 Ginobili 1977 2008 13237 6 30
25 1994 ADavis 1968 1999.2 26083 6.2 30.25
25 1992 Mutombo 1966 1997.7 36791 6.7 30.75
23.8 1996.0 2000.8 23524 5.8 28.6

27.1

_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
NickS



Joined: 30 Dec 2004
Posts: 384


PostPosted: Wed Sep 30, 2009 12:31 pm Post subject: Reply with quote
Nice work.

Mike G wrote:
'Peak' is defined with equal weight given to a player's:
- Best PER
- Most WinShares
- - If Offensive WS peaked in a year other than total WS, the average of the two is used.
Example: Rashard Lewis' PER tops (at 20.7) in 2007. His WS tops (at 9.9) in 2008; but his OWS peaked sharply (8.3) in 2006. His 'overall' peak is thus fudged to 2007.


Just one note, if you take AI's peak year to be 2000-01 (the year he won the MVP, and his second highest PER season) that would drop his peak age by 6 years, and the average peak age of his cohort by .3 year. It wouldn't make much difference but would make the relationship between the cohorts look slightly more linear (20/21/22 yo rookies would peak at 25.4/26.3/26.8 respectively).
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Wed Sep 30, 2009 1:22 pm Post subject: Reply with quote
If I were to step outside the parameters, I'd rank Iveron's 2003 as his peak, in eWins. Then, the 21-yr-old entries avg 26.4 as their peak age.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
NickS



Joined: 30 Dec 2004
Posts: 384


PostPosted: Wed Sep 30, 2009 1:31 pm Post subject: Reply with quote
Mike G wrote:
If I were to step outside the parameters, I'd rank Iveron's 2003 as his peak, in eWins. Then, the 21-yr-old entries avg 26.4 as their peak age.


Yes, I understand why it's important to have criteria that can be applied systematically even if the criteria is somewhat arbitrary.
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 865
Location: Washington, DC

PostPosted: Wed Sep 30, 2009 4:25 pm Post subject: Reply with quote
A guy over at RealGM (goes by TrueLAFan -- he's inexplicably a Clippers fanatic) once provided some analysis showing historically, total minutes is a better predictor than age of when a player will begin to decline. Alas, RealGM didn't archive some old conversations when it went to a new message board provider and the thread disappeared into the ether.

page 2 of 4

Author Message
tawtaw



Joined: 25 Jun 2008
Posts: 26
Location: Oregon

PostPosted: Thu Oct 01, 2009 11:29 pm Post subject: Reply with quote
That's really interesting MikeG. If the data is handy, can you tell if the players that entered the league early had more seasons near their peak than the players that entered at a later age?
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3427
Location: Delphi, Indiana

PostPosted: Fri Oct 02, 2009 4:36 am Post subject: Reply with quote
Along with this suggestion by kjb:
Quote:
.. total minutes is a better predictor than age of when a player will begin to decline.

... There are obviously various ways a player may be said to be 'in his prime', 'near his peak', etc. Taking a single 'best' year, everything afterward could be called 'decline'.

Meanwhile, I did include a total minutes (RS) column, just for reference. To date, players entering at 21 have averaged 1600 more minutes than those who entered at 20; and slightly more than the 18-19 group. There are more 'completed' careers among the 21's, and the 18-19-20 tend to be younger at present. But one could find a minutes per year, before/after peak, or whatever.

All this followup study did was:
- define 'elite player' differently (allstar vs high-PER)
- include active players, as well as recently-retired
- describe a career trajectory differently (by peak age vs by length)

And the results are strikingly similar. A career started younger peaks younger, as well as terminates younger.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
DJE09



Joined: 05 May 2009
Posts: 148


PostPosted: Fri Oct 02, 2009 9:05 am Post subject: Reply with quote
Mike,

Your study here is real nice. I'm not sure I would like to say anyone who was a rookie in 2002 or later has definitely peaked. Since this is a reasonable percentage of your 20 and unders, this would significantly increase their average age, eg. I could easily see healthy A'mare, TP, and Arenas posting "peak" seasons this year (along with quite a few other players). If those 3 players peak this season, your average age for <=20 players shifts to over 26 ... not such a big difference.

In a related observation, KD's list of top 10 player seasons (link here) contains 8 by players who rookied as 20 and unders - only TD and Wade represent the older cohort. Perhaps a shorter duration in College / quicker transition to NBA causes players to reach greater heights, rather than for a more sustained period? I wonder if the NBA has considered this aspect of putting off players for longer - whilst the 18/19 year olds are typically not "NBA ready" forcing them to 'get ready' in the NBA may produce 'better' players than them 'getting ready' somewhere else.

One thing neither your study, nor the SI article addressed, was if coming out earlier shortened or made a player's career worse (I know it would be very hard to establish this) than it would've otherwise been. Even if a player can only play a finite number of NBA minutes, if they are capable of starting them at 20, and end up retiring at 35 instead of 22 and 37, but have the same production etc, I know which option I would choose ... two more years as a multi-millionaire ex-nba player or two years or having to prove myself for no "payment" for "amature" competition ... One of the obvious problems of this is that each player is unique, there is no "typical" player development schema - we have no KG who went to college for a season or two to compare KG too (although one suspects such a player might bark less).

A study of that would have to have some nice way of managing 'injuries' and some how addressed the failures as well as the successes ... of course I have no idea how to do that Very Happy
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3427
Location: Delphi, Indiana

PostPosted: Fri Oct 02, 2009 9:27 am Post subject: Reply with quote
DJE09 wrote:
I could easily see healthy A'mare, TP, and Arenas posting "peak" seasons this year (along with quite a few other players).

If Arenas this year not only returns to his previous form but surpasses it, that's an unprecedented feat in NBA history (after missing 2 years). If all 3 do such a thing, it's a miracle.

If all these should do so, and none of the 21+ group should do so (Gasol?), the peak ages will shift relative to one another. Much more likely is that they'll remain about 1.7 years apart.

Quote:
Perhaps a shorter duration in College / quicker transition to NBA causes players to reach greater heights...?

More likely, it's just that better players tend to come out earlier.

Quote:
... Even if a player can only play a finite number of NBA minutes, if they are capable of starting them at 20, and end up retiring at 35 instead of 22 and 37..

I'd rather retire with good knees. And if salaries go up, you get more money with a later career.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
DJE09



Joined: 05 May 2009
Posts: 148


PostPosted: Fri Oct 02, 2009 9:49 am Post subject: Reply with quote
Mike G wrote:
me wrote:
... Even if a player can only play a finite number of NBA minutes, if they are capable of starting them at 20, and end up retiring at 35 instead of 22 and 37..

I'd rather retire with good knees. And if salaries go up, you get more money with a later career.

I'm not sure Reggie did this, and I was assuming that career length was basically the same - ie same number of seasons for the 20yr and 22 yr old, so the $ would basically be the same, yes? I mean salary correlates with years in the league, not age right?
For the Salary Argument to come into effect, you'd have to have some situation where the player who came out younger was earning for less years. Your chart seemed to show that the players peaked 'earlier' but it was (almost) commensurate with their younger age, so they were not playing less seasons by coming out earlier. In fact I thought the SI evidence was that they get more seasons, albeit it only a fraction of a year (between .25 and 0.5?) for each year they shave off college. So there was a financial advantage to coming out early.

There was a glut of players in the 84/85 draft which seems to have altered perceptions of a 'typical' NBA career length. Most players don't play even 12 seasons.

TP wasn't injured significantly last season, A'mare was on track to have his best season last season until the eye injury, which I can't really believe is related to micro-fracture surgery, and I think I can agree that it is unlikely agent zero will submit anything near a complete season, but if he does, I guess we are just going to disgree about his projected level. I don't think it is unlikely that all 3 have personal best seasons in 09/10.
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3427
Location: Delphi, Indiana

PostPosted: Fri Oct 02, 2009 2:48 pm Post subject: Reply with quote
DJE09 wrote:
..I don't think it is unlikely that all 3 have personal best seasons in 09/10.

Get back to me on this.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3427
Location: Delphi, Indiana

PostPosted: Mon Mar 08, 2010 5:37 pm Post subject: Reply with quote
As a follow-up to a study of elite NBA players and their age at entry vs their career length, I've done a somewhat larger survey of semi-greats.

The earlier one looked at players whose career PER was 19.5 or greater through age 24. Whether he's joined the league after 4 years of college or skipped straight from high school, I assume a player will have established himself by age 24 (exceptions being very-late bloomers, though none come to mind).

So now I've looked at guys -- careers completed -- with career PER less than 19.5 (thru age 24) but greater than 17.0 . A PER of 15.0 is defined as average. To avoid the super-expanded early-ABA era, only players from 1974 onward are included. Before that, some marginal guys got big minutes, with sometimes inflated stats.

Along with summary of the earlier survey, here's the new one.
SS# is sample size: the number of players in an age group.
(The under-21 entries in the 17-19.5 group are just 3 players: Cliff Robinson I, Antoine Walker, and Stephon Marbury. You might just ignore this line, because they aren't typical of anything.)
Average career, by age as a Rookie:
Code:
Thru age 24, career PER 19.5 or more
ageR ageL Min min/yr yrs ss#
20 32.2 32821 2460 13.2 10
21 33.1 30350 2315 13.1 12
22 33.6 31623 2483 12.6 19
23 34.2 32896 2701 12.2 6

Thru age 24, career PER 17 to 19.5
ageR ageL Min min/yr yrs ss#
19-20 28.7 27569 2636 10.3 3
21 30.9 23847 2138 10.9 12
22 31.7 23102 2126 10.7 35
23 33.6 26246 2230 11.6 16


Other than the anomalous 19-20 yr line (3 oddballs with mostly-bad teams), the advantage of being an older rookie is even more pronounced among middling stars.

More than half of this group (35/66) rookied at age 22. Relative to that standard:
- The age of last (1000 min.) season is almost one year less for players entering a year earlier (21).
- 23-year old rookies play to almost 2 years older.
http://www.basketball-reference.com/pla ... i?id=7WCAT

Minimum 3000 NBA/ABA minutes thru age 24.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3427
Location: Delphi, Indiana

PostPosted: Mon Mar 08, 2010 9:51 pm Post subject: Reply with quote
Another 80 players whose career PER at age 24 was between 15.5 and 17:
Code:
ageR ageL Min min/yr yrs ss#
19-20 27.5 14093 1566 9.0 2
21 29.3 20263 2153 9.3 10
22 31.1 21542 2097 10.1 55
23 32.1 22064 2182 10.1 13

_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3427
Location: Delphi, Indiana

PostPosted: Wed Mar 10, 2010 4:40 pm Post subject: Reply with quote
Some average players, and below
Code:
age 24 PER 14.5 - 15.5
ageR ageL Min min/yr yrs ss#
20-21 28.4 15551 1711 8.6 12
22 30.8 21036 2096 9.8 50
23 30.5 16457 1948 8.5 16

age 24 PER 13.5 - 14.5
ageR ageL Min min/yr yrs ss#
19-20 23.7 9648 1799 5.3 3
21 27.5 14768 1991 7.5 11
22 29.2 15475 1869 8.2 53
23 29.9 15467 2059 7.7 19

_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3427
Location: Delphi, Indiana

PostPosted: Fri Mar 12, 2010 7:37 am Post subject: Reply with quote
DJE09 wrote:
... I'm not sure I would like to say anyone who was a rookie in 2002 or later has definitely peaked. ... I could easily see healthy A'mare, TP, and Arenas posting "peak" seasons this year ...
Maybe next year.

This refers to AllStars entering the league 1992-2003, and my guesses as to their ages at which they 'peaked'.
Pared down to players currently active: Anyone think any of these players will surpass a previous best?
Age - Player (yr)
23 - Kirilenko'05, McGrady'03, Carter'01
24 - Arenas'07, Amar'e'07, Joe Johnson'06, Magloire'03
25 - Yao'06, Hill'98, Parker'08, Jermaine'04, Allen'01, Ratliff'99, Hamilton'04, McDyess'00
26 - Boozer'08, KMart'04, Brand'06, Okur'06, Peja'04, Duncan'03, Stack'01, Sheed'01, Kidd'00, Finley'00
27 - Redd'07, Artest'07, Jamison'04, Gasol'08, Butler'08, Lewis'07, Kobe'06, Dirk'06, Marion'06, Pierce'05, Garnett'04, Shaq'00
28 - BDavis'08, Juwan'02, Ilgauskas'04
30 - Ginobili'08, Billups'07
32 - Nash'07

Average peak age for these 43 -- 26.1
For those who rookied at <21 it's 25.2
For those entering over 21 it's 26.6
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
tsherkin



Joined: 31 Jan 2005
Posts: 247


PostPosted: Wed Mar 24, 2010 3:50 pm Post subject: Reply with quote
Kev, why don't you just ask John about that post? PM him, he responds to those after a few days. I was asking him questions about projecting RPG for players in different eras with rebound rate and he was pretty quick about it.

EDIT: Hmm, didn't notice there was a second page to this thread...

Anyway, I've started a thread up on the stats forum on RealGM to try and draw John into it and get him to re-post his analysis so that we can have a look at it over here.
Back to top
View user's profile Send private message
tsherkin



Joined: 31 Jan 2005
Posts: 247


PostPosted: Mon Mar 29, 2010 9:42 pm Post subject: Reply with quote
Here's John's summary:

Quote:
Player age has little to do with decline. It's the wear and tear on the body that predicts player dropoff. Call it the Indiana Jones Rule—“It’s not the years, honey, it’s the mileage.”

For most players and (especially) smaller players, the line seems to be at about 36000 regular season minutes. Jordan retired for the second time after about 36000 regular season minutes. He was still great, but I think his physical skills and overall play had started to erode in his final season and would have continued to if he’d kept playing. Gary Payton started to drop off after about 35000 regular season minutes. Same with Jerry West.

Big men seem to have about one or maybe two extra seasons...enough to get to about 38000 to 40000. But a lot of big men decline around the 36000 minute mark as well. The bottom line is that 36000 minutes is a lot of basketball, and the human body doesn't keep regenerating all its parts.

All players start to lose physical skills, probably earlier than 36000 minutes, but the smarter players can compensate for it. For a while. Once you pass 41000 for a big man, or 7000 or so for a smaller guy, it’s really hard or impossible to keep the level of your play up that high. There are a couple of outliers--Karl Malone, for example. had his skills drop off after 2000, and he had played about 44000 regular season minutes and never had a major injury at that time. Kareem was still pretty great in 1985 and 1986, good enough to finish in the top 5 of MVP voting. He had played about 48000 regular season minutes. I don't even know how to judge Wilt...he was still a top 5 player when he retired, and he'd played almost 48000 minutes. But those are the only real outliers in the last 40 or 50 years.


There's some further discussion on the topic as he discusses the thresholds, the outliers, etc, etc.
Back to top
View user's profile Send private message
mtamada



Joined: 28 Jan 2005
Posts: 373


PostPosted: Tue Mar 30, 2010 1:58 am Post subject: Reply with quote
It appears that both TrueLAFan at RealGM and MikeG here are using regular season minutes only, and ignoring playoff minutes? That might indeed be a correct procedure (but if so, then I could imagine that seasons played might be a helpful explanatory variable, as well as minutes played), but I'd like to see how the analysis compares if playoff minutes are included. Because if the pure minutes played argument is correct, then it would seem that playoff minutes ought to count too. (And maybe minutes played in the Olympics too, as the syndrome of more-injuries-after-playing-in-the-Olympics is often mentioned, although I don't know if anyone has done a systematic study.)

Basically current age, seasons played, minutes played (either regular season only, or including playoffs, whichever is better), and maybe even games played are all potential explanatory variables. TrueLAFan seems to have dismissed all the other explanatory variables and claims that only regular season minutes matter, but his evidence or justification for this seems pretty minimal.
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3427
Location: Delphi, Indiana

PostPosted: Tue Mar 30, 2010 6:43 am Post subject: Reply with quote
There are a couple of questions here:
1 - How young should we be drafting players?
2 - How hard should we be running them?

I haven't studied it systematically, but it's always seemed that novice coaches tend to overuse their star players. To give them more minutes, that is.

There's probably a more general relationship between the strength of will of the player and the coach. Larry Brown was no novice when he acquiesced to Iverson's insistence to play almost every minute.

MTamada may be right that playoff minutes should count, though it may be offset by a higher motivational level. Playing long minutes for perpetually bad teams might create more emotional wear and tear.

The studies I've undertaken have looked specifically at the relation between the age of entry and the length of the career. At every level of talent, it seems a career is hardly extended (or actually shortened) by entering younger than 21-22.

Summarized :
Code:
Thru age 24, career PER 19.5 or more
ageR ageL Min min/yr yrs ss#
20 32.2 32821 2460 13.2 10
21 33.1 30350 2315 13.1 12
22 33.6 31623 2483 12.6 19
23 34.2 32896 2701 12.2 6

Thru age 24, career PER 17.0 to 19.5
ageR ageL Min min/yr yrs ss#
19-20 28.7 27569 2636 10.3 3
21 30.9 23847 2138 10.9 12
22 31.7 23102 2126 10.7 35
23 33.6 26246 2230 11.6 16

Thru age 24, career PER 15.5 to 17.0
ageR ageL Min min/yr yrs ss#
19-20 27.5 14093 1566 9.0 2
21 29.3 20263 2153 9.3 10
22 31.1 21542 2097 10.1 55
23 32.1 22064 2182 10.1 13

thru age 24 PER 14.5 - 15.5
ageR ageL Min min/yr yrs ss#
20-21 28.4 15551 1711 8.6 12
22 30.8 21036 2096 9.8 50
23 30.5 16457 1948 8.5 16

thru age 24 PER 13.5 - 14.5
ageR ageL Min min/yr yrs ss#
19-20 23.7 9648 1799 5.3 3
21 27.5 14768 1991 7.5 11
22 29.2 15475 1869 8.2 53
23 29.9 15467 2059 7.7 19

The small-sample ages are included for completeness' sake.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
tsherkin



Joined: 31 Jan 2005
Posts: 247


PostPosted: Tue Mar 30, 2010 1:05 pm Post subject: Reply with quote
mtamada wrote:
It appears that both TrueLAFan at RealGM and MikeG here are using regular season minutes only, and ignoring playoff minutes?


Actually, in at least one of his posts he discusses playoff minutes in that thread.

spr6=underrated quoted truela from another thread saying:

Quote:
The max range seem to be around 36,000-40,000 combined regular season and playoff minutes.


He's not necessarily articulating it as clearly in the posts that he himself made in the thread, but he definitely referred to combined minutes as the drop-off point in the quoted portion. The original post was a long time ago and he was summarizing from memory.

EDIT: I guess his examples in the latest thread are a little off.

Malone had played just under 41,700 regular-season minutes after the lockout season, and an additional 6,170 playoff minutes, so nearly 48,000 combined minutes before the 99-00 season even begun... And then he promptly posted a 25.5, 9..5, 3.7 season on 58.2% TS. He'd still be pretty good in the two following years, averaging over 22 ppg, 8+ FTA/g and 4.3+ apg, but his defensive rebounding declined.

You'd think it was a sign of athletic decline, but traditionally steals are looked at as an athletic indicator and Malone recorded consecutive seasons of 2.5+ STL% from 01-02 through 02-03, the two highest STL% seasons of his career, and playing 80+ games both years and 2900+ minutes played.

He was an absolute freak...

[Crow]

page 3 of 4

Author Message
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Fri Apr 09, 2010 9:19 am Post subject: Reply with quote
A summary of the above summary:
In the 5 PER-based groups, weighted by sample size in each group,
relative to (214) players who entered at age 22,
players (70) entering younger averaged .33 fewer seasons and 900 fewer minutes in their NBA careers.

That, and less (or no) college career.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Thu Mar 03, 2011 8:00 am Post subject: Reply with quote
Decided to do an update, include playoff minutes in career assessment, and limit the study to players entering since 1990.
Once again, 3000 minutes by age 24 is the criterion, and a minimum PER of 13.5 (career) through that age. There are 112 such careers which have been completed.
http://bkref.com/tiny/1Z5hO

To get decent sample sizes, the 112 players are divided into thirds, by PER: 16.2 or above, 14.6 to 16.1, and 14.5 and under.
Within each group, players are separated by rookie age, again into 3 groups: under 22, 22YO, and >22.
Rookies have been predominantly 22 years of age, per b-r.com's definition (Feb.1 of season).
TG and TMin refers to totals of regular season and playoffs.
po% is % of players' career minutes in playoffs.
ss# is the sample size averaged in each line.
AgeR and AgeL are Rookie and Last playing age.
Code:
ss# PER24 Min24 AgeR AgeL yrs TG TMin po%
12 18.6 9493 20.5 31.9 12.4 839 26077 .061
17 17.7 6936 22.0 32.4 11.4 691 20960 .051
9 18.3 4736 23.1 34.8 12.7 814 25861 .054

12 15.1 6654 20.7 30.3 10.7 676 17620 .057
20 15.3 6514 22.0 32.8 11.8 790 23682 .059
3 14.7 3970 23.0 34.7 12.7 876 21832 .041

8 14.1 8505 20.5 30.3 10.8 654 18453 .044
26 13.9 4843 22.0 31.8 10.8 716 18214 .059
5 14.0 4107 23.0 29.0 7.0 354 8911 .004

For the younger rookies in each PER bracket, minutes thru age 24 are of course higher for those who entered earlier. But final career minutes are not necessarily so.
Only for below-avg PER, and a meager sample of 5, do we see that older rookies have had insubstantial careers. And no playoffs.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
acollard



Joined: 22 Sep 2010
Posts: 56
Location: MA

PostPosted: Thu Mar 03, 2011 11:04 am Post subject: Reply with quote
You highlight the fact that the average rookie year of each sample group is very different, are you worried that this could bias the data? I'm worried you're comparing apples and oranges (superstars in the 1960s and 70s vs. superstars in the 80s and beyond). How much do you think the different eras play into it vs. the actual starting age of the players?
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Thu Mar 03, 2011 11:36 am Post subject: Reply with quote
My latest post only compares players by rookie age in the last 20 years.
Earlier, I looked at players since 1974; not just superstars, but classed by PER in both studies.
In both cases, players are grouped by their PER through age 24, whether they'd been in the league for 1 year or 5 years.
There's no comparison of players of different eras.
Can you rephrase your question?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
acollard



Joined: 22 Sep 2010
Posts: 56
Location: MA

PostPosted: Thu Mar 03, 2011 12:38 pm Post subject: Reply with quote
I apologize, I should have waited until I had more time to flesh that out, and I did miss that the last post only uses players after 1990.

Basically, I have some worries about population bias in the samples, especially given the small number in each. My worry about population bias due to era is largely diminished by limiting to post 1990 players.

Now, however, I am worried about selecting against players who have played the longest (because they are still active). I don't know how this would affect the results, but it might, and it is something to consider. I was just wondering if it worried you, or if you had considered it.

Players in the late 90s-mid 00s came out of high school more than the early 90s. Good/healthy players in that era are still playing (excluded) but the hurt or bad ones are out of the league and still used in the study.

http://en.wikipedia.org/wiki/NBA_high_school_draftees

I think using a discrete period without many changes to the game and without many or any active players (1980-1990 or 1985-1995- there are about 5 active players drafted earlier than 1995) would eliminate most of my concerns. But then again, it could be fine.
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Thu Mar 03, 2011 1:44 pm Post subject: Reply with quote
acollard wrote:
...I am worried about selecting against players who have played the longest (because they are still active)...
Players in the late 90s-mid 00s came out of high school more than the early 90s. Good/healthy players in that era are still playing (excluded) but the hurt or bad ones are out of the league and still used in the study.
..

Good concerns, and thanks for clarifying.
For sure, I worry about sample selection. There are lots of players who aren't in the study because their first years were not very substantial: they never got to 3000 minutes before age 24.
Code:
Player G24 Min24 PER24 AgeR
Kenny Williams 260 2918 16.1 21
Eric Murdock 129 2915 18.4 23
Malik Sealy 161 2899 13.7 22
Lucious Harris 156 2860 13.5 23
Stanislav Medvedenko 204 2830 13.5 21
Derek Anderson 104 2817 15.7 23
Scott Burrell 116 2781 13.7 23
George Lynch 127 2715 14.8 23
Greg Minor 141 2706 14.2 23
Shandon Anderson 147 2668 13.6 23
Bison Dele 149 2652 16.4 22
...

There were about 70 careers of players with 200-2918 minutes by age 24 and with PER > 13.5.
The majority of such players did not have substantial/lengthy careers, and they are not included in the study.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Fri Mar 04, 2011 9:12 am Post subject: Reply with quote
OK, a quick look at Current players, their rookie ages, rookie minutes, years in the league, minutes per year.
Sorted by career regular-season minutes.
Code:
Min ss AgeR MinR Age yrs Min/Yr
35,000 + 8 20.1 2272 34.9 15.8 2607
30-35000 12 21.4 1447 33.9 13.5 2491
25-30000 17 20.5 1947 32.7 13.2 2171
20-25000 24 20.9 1824 30.9 11.0 2205

15-20000 38 21.2 1518 29.2 9.0 2058
10-15000 63 21.4 1205 28.4 8.0 1698
5-10000 101 21.7 1082 26.6 5.9 1461
< 5000 185 22.1 704 23.7 2.6 868

It appears that longevity has been a function of early drafting.
Or that the best and longest lasting players have been drafted younger.

Sorted by age in rookie season:
Code:
AgeR ss MinR From Age yrs G Min Min/Yr
18 9 447 2002.0 27.0 10.0 603 17205 1576
19 36 1264 2005.0 25.0 7.0 440 12554 1772
20 77 1360 2006.2 24.8 5.8 355 10134 1695
21 85 1321 2005.4 26.5 6.5 411 12031 1658

22 106 983 2006.2 26.8 5.8 342 8675 1329

23 83 922 2005.8 28.2 6.2 334 7485 1140
24 29 773 2007.6 27.4 4.4 222 4599 972
25 11 610 2007.8 28.2 4.2 230 4906 892
26+ 12 756 2008.8 29.3 3.2 163 3076 926

The average rookie age of players currently in the league is 21.67
The career-minutes-weighted average is 20.86
The Min/Yr-weighted avg rookie age is 21.48
Suggestions and comments strongly encouraged.
_________________
`
36% of all statistics are wrong

Last edited by Mike G on Fri Mar 04, 2011 9:56 am; edited 1 time in total
Back to top
View user's profile Send private message Send e-mail
EvanZ



Joined: 22 Nov 2010
Posts: 307


PostPosted: Fri Mar 04, 2011 9:28 am Post subject: Reply with quote
Mike, maybe I'm missing it in your numbers, but did you look at the average age of exit from the league? If that is the same between old and young draftees alike, then of course, younger draftees would on average have longer careers.

If there is a difference in the exit age, that probably tells you something as well.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Fri Mar 04, 2011 9:47 am Post subject: Reply with quote
Yes, scroll up and see AgeL
Quote:
AgeR and AgeL are Rookie and Last playing age.

Earlier studies used Age at last 1000-minute season.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
EvanZ



Joined: 22 Nov 2010
Posts: 307


PostPosted: Fri Mar 04, 2011 10:28 am Post subject: Reply with quote
Mike G wrote:


Code:
ss# PER24 Min24 AgeR AgeL yrs TG TMin po%
12 18.6 9493 20.5 31.9 12.4 839 26077 .061
17 17.7 6936 22.0 32.4 11.4 691 20960 .051
9 18.3 4736 23.1 34.8 12.7 814 25861 .054

12 15.1 6654 20.7 30.3 10.7 676 17620 .057
20 15.3 6514 22.0 32.8 11.8 790 23682 .059
3 14.7 3970 23.0 34.7 12.7 876 21832 .041

8 14.1 8505 20.5 30.3 10.8 654 18453 .044
26 13.9 4843 22.0 31.8 10.8 716 18214 .059
5 14.0 4107 23.0 29.0 7.0 354 8911 .004



I took the data from the above post and regressed YRS on AgeR + PER24 weighted by sample size. Here are the results:

Code:

Call:
lm(formula = YRS ~ AgeR + PER, data = ages, weights = ss)

Residuals:
Min 1Q Median 3Q Max
-7.8468 -1.6142 0.2966 1.6536 3.3306

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.51030 10.63867 0.518 0.623
AgeR -0.05907 0.45678 -0.129 0.901
PER 0.38268 0.21194 1.806 0.121

Residual standard error: 4.012 on 6 degrees of freedom
Multiple R-squared: 0.3553, Adjusted R-squared: 0.1404
F-statistic: 1.653 on 2 and 6 DF, p-value: 0.2680


Not statistically significant.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Fri Mar 04, 2011 10:35 am Post subject: Reply with quote
Is there an English translation?
(Numbers may be nouns.)
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
EvanZ



Joined: 22 Nov 2010
Posts: 307


PostPosted: Fri Mar 04, 2011 10:47 am Post subject: Reply with quote
Mike G wrote:
Is there an English translation?
(Numbers may be nouns.)


It may be true that longevity depends on AgeR and PER, but from the data you provided, there doesn't appear to be a statistically significant correlation.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Fri Mar 04, 2011 11:26 am Post subject: Reply with quote
In other words, going pro at a younger age does not add anything
(in minutes or years) to one's pro career (while curtailing or eliminating a college career) ?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
EvanZ



Joined: 22 Nov 2010
Posts: 307


PostPosted: Fri Mar 04, 2011 11:41 am Post subject: Reply with quote
Mike G wrote:
In other words, going pro at a younger age does not add anything
(in minutes or years) to one's pro career (while curtailing or eliminating a college career) ?


I don't know about minutes, but as for PER or years, apparently not.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3623
Location: Hendersonville, NC

PostPosted: Fri Mar 04, 2011 12:03 pm Post subject: Reply with quote
I think it's quite significant, if I'm an elite young player, that entering the NBA at age 19 does not give me any better or longer pro career than having a full college career and going pro at 22.

Whether I play 15 years and never look back, or fizzle after 5 years and forever wish I had a college degree, it can never hurt to have the college career. Can it?

Is such a tradeoff not significant?
_________________
`
36% of all statistics are wrong

page 4 of 4

Author Message
DSMok1



Joined: 05 Aug 2009
Posts: 602
Location: Where the wind comes sweeping down the plains

PostPosted: Fri Mar 04, 2011 12:12 pm Post subject: Reply with quote
EvanZ wrote:
Mike G wrote:
In other words, going pro at a younger age does not add anything
(in minutes or years) to one's pro career (while curtailing or eliminating a college career) ?


I don't know about minutes, but as for PER or years, apparently not.


That's huge.

A related question: how do the aging curve differ for early entrants vs. late entrants? In other words, are there two "aging" curves at play: 1 based on age/physical maturity, and the other based on time in the NBA/knowledge based on that?
_________________
GodismyJudgeOK.com/DStats
Back to top
View user's profile Send private message Send e-mail Visit poster's website
EvanZ



Joined: 22 Nov 2010
Posts: 269


PostPosted: Fri Mar 04, 2011 2:48 pm Post subject: Reply with quote
Since player contracts are typically multi-year, maybe this is a binary or discrete effect.

It would be interesting to look at the percentage of players that sign contracts after their rookie contract, and the proportion of each pool by age. And then players that sign 3rd contracts, etc.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3569
Location: Hendersonville, NC

PostPosted: Sat Mar 05, 2011 7:08 am Post subject: Reply with quote
DSMok1 wrote:
EvanZ wrote:
Mike G wrote:
... going pro at a younger age does not add anything
(in minutes or years) to one's pro career... ?


I don't know about minutes, but as for PER or years, apparently not.


That's huge.

This is unclear as to the effect of/on PER. Of course early entry does not affect a player's PER.
But higher PER is correlated with a longer career. This is true with every age of entry.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
EvanZ



Joined: 22 Nov 2010
Posts: 269


PostPosted: Sat Mar 05, 2011 8:06 am Post subject: Reply with quote
Mike G wrote:
DSMok1 wrote:
EvanZ wrote:
Mike G wrote:
... going pro at a younger age does not add anything
(in minutes or years) to one's pro career... ?


I don't know about minutes, but as for PER or years, apparently not.


That's huge.

This is unclear as to the effect of/on PER. Of course early entry does not affect a player's PER.
But higher PER is correlated with a longer career. This is true with every age of entry.


I ran the regression again, but just on PER alone.
Code:

Call:
lm(formula = YRS ~ PER, data = ages, weights = ss)

Residuals:
Min 1Q Median 3Q Max
-8.0389 -1.0915 0.4715 1.3894 3.1829

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.2499 3.0979 1.695 0.1340
PER 0.3818 0.1954 1.954 0.0916 .
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 3.703 on 7 degrees of freedom
Multiple R-squared: 0.3529, Adjusted R-squared: 0.2605
F-statistic: 3.818 on 1 and 7 DF, p-value: 0.09163


Technically, it's not statistically significant with the data you gave, but I think the issue is simply that you have lumped 112 data points into 9, so it appears to be a small sample size. If you do this for the full data set, it will most likely come out to be statistically significant. That could be true for the early entry, too, of course.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3569
Location: Hendersonville, NC

PostPosted: Sat Mar 05, 2011 10:04 am Post subject: Reply with quote
There is also likely greater significance using 'last 1000 minute season' to determine career length.
Code:
Thru age 24, career PER 19.5 or more
ageR ageL Min min/yr yrs ss#
20 32.2 32821 2460 13.2 10
21 33.1 30350 2315 13.1 12
22 33.6 31623 2483 12.6 19
23 34.2 32896 2701 12.2 6

Thru age 24, career PER 17.0 to 19.5
ageR ageL Min min/yr yrs ss#
19-20 28.7 27569 2636 10.3 3
21 30.9 23847 2138 10.9 12
22 31.7 23102 2126 10.7 35
23 33.6 26246 2230 11.6 16

Thru age 24, career PER 15.5 to 17.0
ageR ageL Min min/yr yrs ss#
19-20 27.5 14093 1566 9.0 2
21 29.3 20263 2153 9.3 10
22 31.1 21542 2097 10.1 55
23 32.1 22064 2182 10.1 13

thru age 24 PER 14.5 - 15.5
ageR ageL Min min/yr yrs ss#
20-21 28.4 15551 1711 8.6 12
22 30.8 21036 2096 9.8 50
23 30.5 16457 1948 8.5 16

thru age 24 PER 13.5 - 14.5
ageR ageL Min min/yr yrs ss#
19-20 23.7 9648 1799 5.3 3
21 27.5 14768 1991 7.5 11
22 29.2 15475 1869 8.2 53
23 29.9 15467 2059 7.7 19

The highest PER sample averages almost 13 years, over 30,000 minutes.
The lowest PERs average almost 8 years and about half the minutes.
Why doesn't the sample size increase the significance?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
EvanZ



Joined: 22 Nov 2010
Posts: 269


PostPosted: Sat Mar 05, 2011 11:36 am Post subject: Reply with quote
Can you post your entire spreadsheet as a csv?
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3569
Location: Hendersonville, NC

PostPosted: Wed Mar 16, 2011 5:50 am Post subject: Reply with quote
Well, I can't seem to figure that out in finite time. You should be able to copy and import this:
Code:
Player G24 Min24 PER24 AgeR AgeL TG TMin
Marbury 421 15956 19.0 19 31 878 32828
Abdur-Rahim 375 14237 19.7 20 31 836 29011
Walker 369 14186 17.5 20 31 957 33993
Miles 412 11433 13.9 19 27 446 11731
Iverson 274 11038 20.0 21 34 985 40789
Elliott 315 10800 14.3 21 32 827 27338
Webber 288 10678 21.3 20 34 911 33747
Kemp 383 10549 20.1 20 33 1139 32231
Finley 247 9396 16.1 22 36 1232 41904
Hughes 322 9345 15.1 20 31 767 23887
Robinson 242 9321 16.6 22 32 727 26438
Baker 246 9240 17.1 22 34 815 26486
Mercer 250 9125 13.8 21 28 443 13670
Sprewell 228 9045 14.6 22 34 975 37819
Stoudamire 222 9015 17.2 22 34 931 30602
Hardaway 241 8931 20.9 22 36 768 26351
Anderson 273 8920 17.1 21 34 894 26872
Weatherspoon 240 8792 16.2 22 34 933 28040
Thomas 357 8757 14.9 20 32 879 22881
Davis 341 8756 15.8 19 30 747 22291
Mashburn 246 8709 14.5 21 31 663 24611
Simmons 226 8375 15.8 22 28 458 13549
Abdul-Rauf 309 8370 14.8 21 31 601 16043
Bradley 283 8367 15.0 21 32 875 20088
Francis 214 8313 19.8 22 30 581 21854
Murray 311 8246 13.9 21 32 750 18739
VanExel 235 8157 15.3 22 34 956 31359
Szczerbiak 237 8144 16.2 22 31 705 21498
Johnson 215 8127 18.4 22 31 773 28082
Wallace 253 7990 15.4 21 35 1265 41786
Rice 233 7964 14.6 22 36 1055 37022
Reeves 226 7764 15.0 22 27 395 12071
Rider 229 7654 15.1 22 30 584 18604
Mourning 215 7603 20.8 22 37 933 28572
Divac 282 7425 17.2 21 36 1255 37566
Kittles 205 7396 16.0 22 30 561 18663
Payton 245 7341 14.5 22 38 1489 52599
Gill 230 7280 15.4 22 36 993 30277
Taylor 248 7217 14.8 21 30 539 13440
Brandon 304 6736 16.0 21 31 762 22491
Griffin 303 6728 14.3 19 24 303 6728
VanHorn 184 6683 17.6 22 30 632 19613
Marshall 268 6653 14.2 21 35 1004 25911
Houston 237 6587 14.4 22 33 902 30836
Cheaney 213 6579 13.7 22 34 833 22249
Owens 196 6449 15.8 22 31 611 18034
Blaylock 194 6400 13.6 22 34 943 33187
Wright 269 6343 13.5 21 33 794 18899
Smith 187 6192 15.5 22 35 1032 31758
Swift 292 6122 17.0 21 29 554 10925
Thomas 218 6101 13.6 22 32 655 17996
Jackson 161 5986 15.4 22 35 933 30344
Anderson 211 5959 15.6 22 34 849 26688
Hardaway 161 5878 18.7 23 36 923 32678
Trent 250 5859 16.7 21 29 534 10097
Davis 212 5857 15.5 22 37 1231 33268
Williams 193 5768 13.9 22 32 743 19402
Richardson 164 5735 17.1 23 32 645 19512
Smith 228 5646 13.6 22 24 234 5668
Gugliotta 159 5590 16.2 23 35 775 23891
Coles 244 5563 13.5 22 35 863 20514
Ellis 167 5506 16.2 22 32 640 18129
Rogers 226 5498 14.0 22 33 948 23590
Knight 184 5423 16.4 22 33 743 18426
Williamson 211 5420 14.6 22 33 888 19848
Ellison 176 5319 16.9 22 33 478 11661
Grant 182 5294 16.0 22 33 814 22973
LaFrentz 171 5279 17.1 22 31 598 15466
Laettner 151 5251 17.1 23 35 913 26918
Rose 227 5120 14.5 22 34 982 29809
Brown 193 5082 14.9 22 33 639 17625
Lenard 184 5055 14.1 22 32 596 16301
Douglas 154 5032 17.5 23 34 782 21672
Armstrong 245 4897 13.8 22 32 852 20154
Coleman 139 4809 18.5 23 37 820 27360
Fox 234 4713 13.8 22 34 1041 26250
Ceballos 254 4664 19.8 21 31 668 15925
Dickerson 132 4661 14.2 23 27 216 7572
Augmon 155 4617 14.8 23 37 1078 23130
Muresan 203 4612 19.1 22 28 310 6799
Dehere 226 4551 13.6 22 27 402 7305
Vaught 231 4518 14.1 22 32 704 17200
Alexander 221 4421 14.1 22 32 292 5783
Miller 189 4413 16.4 22 33 542 12065
Sura 207 4361 13.7 22 31 670 15827
Fizer 186 4352 13.7 22 27 289 6032
Murray 250 4289 15.2 21 32 665 12206
Fortson 191 4287 17.4 21 30 451 8937
Jones 134 4165 16.7 23 36 1035 35544
Jones 161 4158 13.9 23 33 535 12579
Day 147 4058 14.0 23 31 492 12314
Stewart 157 4052 13.6 23 28 274 5431
Wallace 198 3960 14.7 22 29 386 6187
Mobley 130 3952 14.6 23 33 773 28554
Pack 215 3855 14.5 22 34 585 11898
Campbell 212 3807 13.8 22 36 1150 28043
Ostertag 197 3767 14.0 22 32 845 16521
Barros 222 3711 14.8 22 36 880 19743
Sweetney 233 3610 15.1 21 24 244 3790
Nesby 123 3605 14.2 23 26 255 6658
Keefe 220 3582 13.8 22 30 681 11010
Outlaw 198 3511 14.2 22 36 936 21156
Boone 193 3437 14.3 22 25 268 4600
Rooks 119 3342 14.7 23 34 778 13812
White 168 3337 14.9 22 28 334 5913
Alexander 200 3312 13.7 22 31 316 4917
Roberts 146 3284 14.6 21 29 309 6046
Wells 148 3192 18.4 22 31 639 16232
Knight 182 3184 13.7 22 28 398 4718
MacLean 137 3161 17.0 23 31 319 6679
Traylor 224 3124 14.9 21 27 460 6489
Robinson 82 3002 26.3 24 37 1110 38492

The strong correlation seems to be between AgeR (rookie age) and AgeL (last playing age), when grouped by PER.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
EvanZ



Joined: 22 Nov 2010
Posts: 269


PostPosted: Wed Mar 16, 2011 7:46 am Post subject: Reply with quote
Thanks, I was able to import the data. As far as I can tell, the correlation is only between #years played and PER24, not rookie age:

Code:

Call:
lm(formula = YRS ~ PER24 + AgeR, data = players)

Residuals:
Min 1Q Median 3Q Max
-7.4338 -1.8009 0.4709 1.8589 6.2295

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.09211 6.80548 0.748 0.45593
PER24 0.37412 0.12764 2.931 0.00412 **
AgeR -0.03393 0.29667 -0.114 0.90917
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 2.891 on 109 degrees of freedom
Multiple R-squared: 0.0733, Adjusted R-squared: 0.0563
F-statistic: 4.311 on 2 and 109 DF, p-value: 0.01578


The same holds true for TMin:
Code:

Call:
lm(formula = TMin ~ PER24 + AgeR, data = players)

Residuals:
Min 1Q Median 3Q Max
-19561 -7179 409 5833 34127

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -395.8 22700.1 -0.017 0.986121
PER24 1714.6 425.8 4.027 0.000105 ***
AgeR -272.4 989.6 -0.275 0.783593
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 9643 on 109 degrees of freedom
Multiple R-squared: 0.1304, Adjusted R-squared: 0.1145
F-statistic: 8.174 on 2 and 109 DF, p-value: 0.0004922


and games played:
Code:

Call:
lm(formula = TG ~ PER24 + AgeR, data = players)

Residuals:
Min 1Q Median 3Q Max
-510.15 -184.69 11.47 172.14 799.90

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 356.804 618.795 0.577 0.5654
PER24 28.490 11.606 2.455 0.0157 *
AgeR -3.673 26.975 -0.136 0.8919
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 262.9 on 109 degrees of freedom
Multiple R-squared: 0.05268, Adjusted R-squared: 0.0353
F-statistic: 3.031 on 2 and 109 DF, p-value: 0.05237

_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir