re: nba stats site
- in response to steve's questions, the stats come from databases we
have built with different sources and our own charting
the on court versus off court numbers are shown in a 48 minute net
which may be confusing.
Pierce = 3073 min on court, +1.7/48 = +109 pts
Pierce = 856 min off court, -8.4/48 = -150 pts
net = -41 for team
Battie = 1671 on court, +5.0/48 = +174
Battie = 2257 off court, -4.6/48 = -216
net = -42 for team
note there are rounding errors in minutes, plus/minus to one
decimal...the nets would be the same otherwise, Battie's actual +/-
on was +175
- I think I'm probably more skeptical of this type of information than
many of the other posters here, if only because I've seen fairly
similar breakdowns for the Sonics and wasn't convinced they held a
great deal of value.
That said, this is a tremendous resource. I really like the
opportunity to see the magnitude of the effect an unbalanced player
has on his team's offense or defense (or rebounding). For example,
the Sonics had two completely opposite point guards after the trade
last season in Brent Barry and Kevin Ollie. It's easy to know that a
unit with Barry will be superior offensively and one with Ollie
better defensively, but how much? How much did an offensive non-
factor like Ansu Sesay hurt the Sonics' offense? How much did Reggie
Evans help them on the boards? There's alot of information there to
answer these questions.
A couple of things I would do if I had this data:
1. Study the "hot hand" effect with a significantly larger sample.
Assumedly, you can track shot order for every player in the league
2. Do more with assisted/unassisted field goals. That was one of the
hidden benefits of the data I mentioned before. I'm fairly obsessed
with investigating assisted and unassisted field goals, and I think
they go a long ways towards giving players who take a significant
percentage of their team's shots the credit they are due.
Unrelated: This is the first time I've heard of the measure for
distribution mentioned earlier. Would this be a way to track
consistency for players? That is, could we use what percentage of
their points were scored/rebounds grabbed/etc. in each game? This
would seem to account for the biggest problem of most measures of
variability in evaluating consistency in basketball - the wide
differences in the size of the base (total points, rebounds, etc.).
Does that make sense?
----- Original Message -----
From: "Kevin Pelton" <kpelton08@...>
Sent: Saturday, October 11, 2003 1:41 AM
Subject: [APBR_analysis] Re: nba stats site
> I think I'm probably more skeptical of this type of information than
> many of the other posters here, if only because I've seen fairly
> similar breakdowns for the Sonics and wasn't convinced they held a
> great deal of value.
I think the best way to look at these numbers is to remember that they are
only indirectly _individual_ stats -- they are primarily _team_ stats,
showing how well the team performs in various situations (eg with and
without a certain player on the floor). What made me suspicious of +/-
analysis that came before was it was presented without the context that the
82games site provides. I especially like the 5-man unit numbers, which will
probably be the focus of my next round of calculations.
- --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer
> I especially like the 5-man unit numbers, which willThis reminds me of what I found to be the single most valuable bit
> probably be the focus of my next round of calculations.
of information from this kind of a statistical breakdown. I'll put
it in capitals so you don't accidentaly miss it:
THE STARTING LINEUP MATTERS!
Previously, I'd been of the opinion that the starting five really
didn't matter much. I recall getting a little tired of JohnH dogging
coaches for starting players (Johnny Newman comes to mind here) -
the really problem was that guys like that were playing at all. My
opinion could be best summed up as, it's who plays the most that
truly makes a difference, right?
Looking at the lineup analysis, however, it's obvious that the only
group that really plays together all that much is the starting five.
Guys rotate in and out so much in the late first/third quarters and
second/fourth quarters that no single lineup gets all that much time
together, even on a team with a fairly set rotation.
What this means to me is that while it might not be crucial to get
your five best players on the court at the start, it is imperative
to have a lineup that works well together and complements each
other. If guys on the bench don't fit as well, that's not much of a
problem, but the starters better be a cohesive unit.
- --- In APBR_analysis@yahoogroups.com, "Kevin Pelton" <kpelton08@h...>
> This reminds me of what I found to be the single most valuable bitI wonder how many games Detroit would have won if Rick Carlisle hadn't
> of information from this kind of a statistical breakdown. I'll put
> it in capitals so you don't accidentaly miss it:
> THE STARTING LINEUP MATTERS!
stubbornly started Michael Curry all season long. He had the third
worst Roland rating in the league, showing that whatever his defensive
contributions were, they definitely didn't make up for his complete
lack of offense.
Detroit's starters (who were their most used lineup - and the third
most used lineup in the league) were 28-36 by the site's calculations
while Detroit's overall record was 50-32!
Some of the discrepency comes from Detroit's excellent bench (Barry
and Williamson were among the league leaders in Roland ratings), but
the numbers definitely suggest that Curry was pretty worthless, at
least in a lineup that also featured another all-defense no-offense
player in Ben Wallace (and a third player in Cliff Robinson who wasn't
much better in those regards).
Joe Dumars was smart to trade Curry to prevent Larry Brown from even
getting the chance to give him playing time, and now it looks like
Curry might be starting in Toronto (whose new head coach of course was
the architect of the Pistons defense the past two seasons).