RE: [APBR_analysis] Re: Scoring points vs Low FG%
- -----Original Message-----
From: Kevin Pelton [mailto:kpelton08@...]
Sent: Thursday, January 01, 2004 10:26 AM
>> --- In APBR_analysis@yahoogroups.com, "Franklin X" <xeifrank@y...>I still haven't finished reading DeanO's book, in fact I'm still
>> Where can I find your book at? and read a review?
>There's my review. Admittedly, you'd probably get just as much out
>of reading DeanO's chapter summaries at www.basketballonpaper.com.
on chapter one, but as someone familiar with his work at the
Journal of Basketball Studies and having read some of his
chapters in preliminary form, I will say that Dean Oliver has
done the best work of mathematically modelling basketball that
The one modification I would suggest to KevinP's review: early
on he lists DeanO's experience and qualifications, but doesn't
mention his work for (or with) the WNBA on Project Defensive
Scoresheet. KevinP does talk about it at length later on in
the review, but I think it's worth mentioning early on, because
it was a project noteworthy both for its innovation and its high
level (granted the WNBA isn't exactly the NBA or even NCAA Mens
B-ball, but it's the highest womens pro league in the country)
>As MikeG said in response to your questions, that's what we're hereThis was a question I was going to ask Dean privately but I'll
>to discuss. DeanO did outstanding work in his book modeling how
>efficiency changes with changes in workload, so to speak, but his
>models remain pretty theoretical. The fact is that most players
>stick to their roles pretty consistently, and we don't have any lab
>in which to experiment with Derek Fisher being a go-to guy on a
>night-in, night-out basis.
>The value of assists? To get a legitimate answer would probably
>require at least a season-long or league-wide study where volunteers
>tracked whether unmade shots would have been credited an assist or
>not. Ask DeanO how difficult that would be.
ask it publicly instead: I'm still in the chapter in which he
describes his notation for recording play-by-play info. Assisted
baskets get a double plus "++" whereas unassisted ones get a single
plus "+". Missed FGs get a minus "-".
What if we recorded "missed FGs which would've been credited with
an assist" as double minuses "--". And for that matter, extend a
similar treatment to FTs (e.g. the PG makes a great pass, the
shooter gets fouled and sinks two FTs, no assist is given, even though
for statistical modelling purposes, one ought to be). The doubling
notation wouldn't mesh well with DeanO's notation for FTs, so some
other sort of notation would be needed.
I suspect that I already know the answer to my question: the
information could be encoded easily enough, the hard part would be,
in the frenzy of a game, deciding which shots or FTs were assist-worthy
or not, and having time to write down the identities of the passers.
But if we recorded those "missing assists" (the ones for both missed
FG attempts, and the ones associated with FTs), we'd have enough
information to get a pretty good estimate of the value of an assist
on average, and of a playmaker's contributions.
While I'm at it, here's an even more impossible request: multiple
assists for some baskets ... I think ice hockey does this. It's
rare enough in basketball so that this is largely irrelevant, but
every once in awhile there's a play where a good pass leads to an
easy pass for a dunk: and the guy who made the easy pass gets
the assist, but the play was really started by the first good pass.
Multiple responsibility for turnovers is another item on my wish
list: usually it's the passer's fault, but sometimes it's the
receiver's fault and I'm not sure that official NBA stats properly
assign the responsibility, especially in cases where it's a legitimately
mixed responsibility. Having an Olden or an Olowakandi (or an
Olumide Oyedeji -- stonehanded "oh-oh" players all) on a team
has to have a negative effect on a PG's Assist:TO ratio.
This was one of the great pleasures of doing DeanO's defensive
scoresheets: when assigning responsibility for causing a turnover,
we were free to list two players if appropriate (e.g. the player
who deflected the pass, and the one who picked up the deflection).