Re: predicting success in the NBA

[Kestutis.Kveraga@dartmouth.edu (Kestutis Kveraga)]

--- You wrote:
> That just doesn't seem empirically true to me. A key to dribbling is 
> natural knee bend and great coordination, and it seems to me that people 
> with normally sized and proportioned bodies (including most NBA guards) are 
> more successful at it than gangly athletes. 

That brings up an interesting question: ARE most NBA guards, in fact, normally
proportioned? I noticed that Tinsley, though he is shorter than me, has a much
longer wingspan (over 6-7, as I recall from the pre-draft measurement data). I
don't think of myself as having unusually short arms, because my wingspan is
greater than my height, and I read somewhere that wingspan is typically about
the same as a person's height. So, is Tinsley representative of NBA guards, or
does he have disproportionally long arms *even for an NBA guard*?

> I don't know if its true, but 
> that's my strong overall impression. Sort of like how it is easier to use 
> shorter chopsticks than those that are 3-feet long. 

Well, all chopsticks are equally impossible to use for me...:)

> Besides, don't long 
> legs cancel out the benefit of long arms? ;-)

By 'long arms', I meant *relatively*, with respect to the rest of the body, as
they are long in the absolute sense in almost any NBA player. 

> Okay, I didn't realize this fact about the SAT but I'll make the following 
> point. The SAT and GRE probably aren't meaningfully useful for predicting, 
> say, future MIT professors or presidents/captains of industry or Nobelists. 
> In other words, many people with great test scores won't be these things.

Right, most won't be, but it's extremelly unlikely that *anyone* with low
scores (those who do get accepted) will ever be one of these. As I said
previously, it's also about identifying potential *failures*, not just
potential *greats*. It's not a linear function - above a certain cutoff point
for a bunch of given criteria, one's chances of being outstanding may be
unpredictable, but it still gives you valuable information, unless everyone in
the group of interest falls above the cutoff point. Let's say a battery of
tests on Moiso showed that, in most respects, he was way above the cutoff point
for 'success' in the NBA, but he scored very low on a psychological measure of
"motivation to succeed". Let's say that from past data we know that those
who've scored low on that measure fail in the NBA at much higher-than-average
rates. Do you still pick Moiso at 11 when you have that tidbit of information? 
This is all hypothetical, of course, and Moiso might make the All-NBA first
team next year. But companies regularly use such quantitative selection
criteria in picking applicants with some success. For example, AT&T always
looked for applicants who were in the top 15% of the class in college,
irrespective of the college. Did it have to do with intelligence? Maybe, but
why not pick  any applicant who's graduated from an elite college, a safer
statistical bet? It turns out they'd found that people who tend to be in that
top 15% have this characteristic of 'swimming to the top' in any organization,
the so-called "achievers", and they did very well at AT&T as well. Arguably,
this is the "achievement" characteristic that is necessary for a merely
talented player to become a star in the NBA, except the cutoff requirements for
playing in the NBA are much more difficult to meet  than for a
middle-management job at AT&T.  

> In terms of the overall probability of achievement, I think that's a more 
> appropriate analogy to predicting NBA success. First of all, you can be so 
> spectacularly dumb and still make good scores on the SAT and good grades at 
> an Ivy League school. 

It depends on your definition of "spectacularly dumb"....I've taught students
who were "spectacularly dumb" *in comparison* to the student body at Dartmouth,
yet they were still way above population average. It's a normally distributed
sample, of course, but it's way out there on the right tail of the population
distribution.... 

>Now, of course, Paul Pierce applied this very same scooping motion to hit a 
> half court shot in practice so maybe there is at least some truth to it. 
> ;-)  

It's funny, but you may be more right than you think. The said test allegedly
measures innate coordination and power, especially when applied to children,
youth, and untrained adults. Free (unguarded) shooting accuracy  statistically
depends on the % of available muscle power necessary to send the ball to the
rim (other things being equal). The more of your available power you have
apply, the greater the error in the output. I won't get into the details of it,
but it has to do with the order in which motor neurons are recruited to produce
movement, among other things. That's why slow movements are more accurate than
fast movements, and that's why half-court shots are harder to make than
3-pointers, which in turn are harder than 12-footers, which in turn are harder
than lay-ins (when all are equally practiced). Pierce, besides being highly
practiced, is also undoubtedly very coordinated and strong, and it wouldn't
surprise me in the least if he did well on the said test. 

> But in basketball a lot of success comes from the intangibles. For 
> example "court sense" is almost totally intangible as far as I'm concerned.

It's 'intangible' in the sense that nobody has figured out a good way to
quantitatively measure it. But, like any computation in the brain, it's
possible (if very difficult at this point) to measure it in principle. Bird's
famed ability to 'think a few moves ahead' on the court is nothing but a very
complex computation by the brain, taking the current positions of the players
on the court in relationship to his own location on it and extrapolating not
only where they're going to be a second or two hence, but also what they
possibly might be *thinking* of doing at that point, based on inputs such as
their movement rates, their tendencies, the game situation, their body language
etc. etc. It's 'intangible' just like the computations of the computer on which
I'm writing this post are 'intangible', except much more complex, which is
partly why we can't specify every step of the computation, like we can for the
computer. But it should be possible to evaluate its behavioral (or even neural)
manisfestations by employing a clever game-like test, kind of like they use to
evaluated race car drivers.  

> I think if someone compiled all the date it's predictive value would be 
> really great compared to "randomness", but not nearly accurate enough to be 
> that useful as a tool. You'd probably be better informed but still arrive 
> at your final player selection in the same apples-and-oranges fashion. 

If the chances of drafting the best available player were improved to, say 60%,
instead of 50% by the 'seat-of-the-pants' methods, would it be worth it? I
think so, the benefits of selecting the BPA 20% more often would accrue over
time, giving the teams that use it a significant advantage....
Kestas