Re: True shooting percentage on potentially assisted baskets?
- One thing that drives me crazy about assists is their subjectivity.
For instance, as we've pointed out here, assists are harder to get in
Europe than in the NBA. Their official scorers are tougher. That
should mean that their assists are more valuable, leading more
directly to baskets...
Sooo, if Ed is acting more as a Euro-like scorer, does that mean he is
exaggerating the value of assists? Potentially. More of the passes
that are being called assists are in the fuzzy area, rather than
I do think that assists have value, adding to a team's shooting
percentage (most teams, though exceptions exist, I'm quite sure). But
I'm guessing that Ed's study exaggerates that effect at least a little
I have considered ways of accounting for this subjectivity but never
put it into my production methods (it's all in R&D mode).
Author, Basketball on Paper
When basketball teams start playing Moneyball, this is the book
--- In APBR_analysis@yahoogroups.com, "dan_t_rosenbaum"
> Granted, assists are somewhat subjective and the data that you
> collected are non-random, so there are good reasons for not coming to
> firm conclusions based upon your work. But right now, I think the
> default is to assume that the true shooting percentage is the same
> for assisted and unassisted field goal attempts. And what evidence
> is there for that default assumption? As far as I can tell, there is
> no evidence. Thus, if I have to choose between your limited/non-
> random study and no evidence, I would have to go with your study.
> Would I like something more complete? Of course, but beggars cannot
> always be choosers.
> And of the two issues, I think the non-randomness is a bigger deal.
> Suppose there is no difference in true field goal percentage between
> phantom assisted field goal attempts (those you were not able to
> chart) and non-assisted field goal attempts. That would only knock a
> point or two off the difference in true field goal percentage that
> you measured.
> So perhaps the Raptors are more efficient on assisted field goal
> attempts and less efficient at defending assisted field goal attempts
> and so this is just a Raptors effect. Yes, I am uncomfortable
> assuming that this is not true, but I am more uncomfortable with the
> default assumption for which there is no evidence. For one thing
> your results are NOT simply due to sampling variation, because most
> of your results are highly statistically signficant. So the argument
> has to be that the Raptors on offense and defense are systematically
> different from other teams. I am more comfortable with the
> assumption that they are about the same than the assumption that the
> true field goal percentage is the same for assisted and unassisted
> field goal attempts.
> Perhaps as a practicing empirical economist, I simply am more used to
> making tenuous assumptions and working with messy data, because the
> compromises that we have to make working with NBA data are generally
> pretty minor to the assumptions economists often have to make in
> working with other types of data.
> Best wishes,
> --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer
> <edkupfer@r...> wrote:
> > dan_t_rosenbaum wrote:
> > >
> > > By my calculations the true shooting percentage is 57.1 percent
> > > on "assisted" true field goal attempts and 46.4 percent
> > > on "unassisted" true field goal attempts. In my opinion, this is
> > > huge difference and suggests that assists are quite valuable.
> > >
> > I have reservations about using my numbers to come to any firm
> > First, the sample dataset is definitely non-random, consisting
> mostly of
> > Raptors games. More importantly, the official scorers gave out
> about 15%
> > more assists than I did. I re-scored a few of these games, using
> the loosest
> > possible definition of assist that I could manage, and my assist
> > still fell short of the official.
> > In another study I had previously found some evidence of a home
> court bias
> > in assist scoring. If my impressions drawn from these studies above
> > correct, we have two reasons to be extremely wary of using assists
> > individual player ratings.
> > I urge caution in drawing conclusions from all this, but I also
> urge an
> > extremely conservative approach to incorporating assists into your
> > given that we have good reasons to think these number are inflated.
> > ed