RE: [APBR_analysis] Re: possessions
- Good points, the other thing though that I wonder about is that
it's common sense and can be shown mathematically that, all other
things being equal, an underdog team will prefer a slow-paced game
and a superior team will prepare a fast-paced one. DeanO even had
an article about this on his website several years ago.
What is harder to say is, outside of the theory, how large is
this effect in actual practice. It sounds like DeanO's found
empirical evidence that it is a small effect.
From: schtevie2003 [mailto:schtevie@...]
Sent: Friday, September 24, 2004 9:53 AM
Subject: [APBR_analysis] Re: possessions
--- In APBR_analysis@yahoogroups.com, "Dean Oliver" <deano@r...> wrote:
> --- In APBR_analysis@yahoogroups.com, "Michael Tamada" <tamada@o...>
> > If we wish to measure game pace, as I mentioned before, I think
> DeanO's "possessions" (i.e. major possessions) are less useful than
> looking at plays/minor possessions. This NBA definition, by failing
> to count those blocked shots and other missed FGAs which miss the rim
> as plays/minor possessions, shares some of that disadvantage.
> A side note on this comment...
> Many coaches believe that controlling tempo is key to winning.
My guess if these coaches were sat down, and a real discussion were had on definition of
terms (centering around the word "control") my guess is that there would not be such a
belief. What I think coaches think of as control is actually an ex post facto rationalization
of fortuitous random events dovetailing with a given game plan. For example, suppose
the coach of a "running team" thinks it is important to "control the tempo" by running, and
it happens in a given game that an unexpectedly large number of rebounds bounce long,
leading to quick outlet passes and fast breaks, then in the post-game press conference he
says that the game was in hand because the tempo of the game was controlled. But it
wasn't; the ball just bounced in a way which given his game plan led to success.
To the extent to which that belief is true, however, is the extent to which the phrase
"controlling tempo" is eqivalent to "sticking to one's game plan". Where one's offensive
and defensive strategies are picked to maximize one's chances of winning, and the
counterpart to each strategy is an expected average duration of each possession.
> looked at pace many different ways, I have found that none of our
> current definitions generally suggest that pace matters much to a team
> winning/losing. Not plays, not possessions. Teams that average a lot
> of plays or possessions don't seem to increase their odds against
> teams that average few by playing a game with more.
I think the problem of such investigation is that one should not expect to see any clear
correlation. Consider this thought experiment. Suppose Team A and Team B are of equal
strength (i.e. playing a whole lot of games against each other would lead to each team
expecting to win half) but that Team A likes to run (has talent suited to such a strategy)
and Team B doesn't. What should one expect that the investigation of pace versus success
would show, using the definitions of pace at hand? I think that the answer is no strong
correlation, if any. Consider the games were Team A was on net luckier, i.e. more long
rebounds, more Team B errant passes leading to Team A fast breaks. Here we would
expect a positive correlation. But even in this best case scenario it would be weak, for the
counterpart to the unexpected success of A is a slower pace for B, for fishing the ball out
of one's basket will slow the pace of B's game. Next consider what happens if A plays
unexpectedly poorly, by taking hasty shots in the half-court? This implies a faster pace,
less success for A and more success for B.
I think the upshot is that after one takes into account all aspects of possible variation in
game variables, one should expect no clear direction from the empirical record comparing
pace and game outcomes.
Which is not to say that pace doesn't matter, but that pace itself, at this level of analysis,
tells you little about the way the game was played or what role "luck" played in the
I am working on
> another approach that may reflect the perception of pace more. I
> believe it may be a way that shows a relationship, but isn't
> necessarily something a coach can control. (In general, this commonly
> held view about controlling tempo is a real pain to prove or disprove.)
> Dean Oliver
> Author, Basketball on Paper
> "Excellent writing. There are a lot of math guys who just rush from
> the numbers to the conclusion. . .they'll tell you that Shaq is a real
> good player but his team would win a couple more games a year if he
> could hit a free throw. Dean is more than that; he's really
> struggling to understand the actual problem, rather than the
> statistical after-image of it. I learn a lot by reading him." Bill
> James, author Baseball Abstract
Yahoo! Groups Links
- --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer <edkupfer@r...> wrote:
> schtevie2003 wrote:Thanks for the data. I am not sure if the latest "points per possession" data are just for
> > --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer
> > <edkupfer@r...> wrote:
> >> Okay, how about this: team pace as the distribution of shot clock
> >> time it
> >> takes to get off the first attempt of the possession. If we ignore
> >> possessions in which no attempt took place, and if we count shooting
> >> fouls as attempts, the distribution of shot clock times for the
> >> 03-04 Net's looks like this:
> >> Time RelFreq CumFreq
> >>> 03 2.7% 2.7%
> >>> 06 1.7% 4.5%
> >>> 09 7.2% 11.7%
> >>> 12 12.2% 23.9%
> >>> 15 23.3% 47.2%
> >>> 18 26.0% 73.2%
> >>> 21 16.2% 89.4%
> >>> 24 10.6% 100.0%
> >> That doesn't tell us too much, so let's compare them to other teams.
> >> First, the 03-04 Raptors, which, to my eyes, were the slowest team
> >> in the history of the NBA. The rightmost column is a random sample
> >> of 44 NBA games.
> >> NJ TRN X
> >>> 03 2.7% 2.1% 1.9%
> >>> 06 4.5% 3.6% 3.6%
> >>> 09 11.7% 10.8% 11.8%
> >>> 12 23.9% 22.5% 26.3%
> >>> 15 47.2% 44.7% 50.0%
> >>> 18 73.2% 70.6% 73.9%
> >>> 21 89.4% 87.9% 89.7%
> >>> 24 100.0% 100.0% 100.0%
> >> The Nets don't look that much different than average by this measure
> >> of pace, which is consistent with other measures.
> >> ed
> > A couple of impressions from the data given.
> > I think the Nets do look different from average.
> They are. I should've said so. A chi-square test on the raw frequencies
> shows a significant difference between the Nets and the NBA sample.
> Statistical significance may not equate to practical significance though --
> I see you address that below.
> > First (and let's stipulate that the NBA numbers are representative)
> This deserves examination, just to make sure. I analysed a second sample of
> 45 games, and performed a X^2 test. The samples were not significantly
> different .
> Here's my problem: parsing these numbers out of PbP logs is extremely time
> consuming for me. I use Excel to do it -- not the best way. So while I'd
> love to mine everything I can out of the logs, I'm limited to what my time
> and basic programming skills have to offer.
> > they get 0.8% more fast breaks (shots within 3 seconds) than the average
> > team. This is about one per game. And let's suppose (reasonably, I
> > that these shots earn 0.5 points more on average than all others.
> Here are the average number of points per possession when the first shot
> attempt was taken in X seconds.
> Time of points/
> 1stAtt poss
> :00-:03 1.43
> :03-:06 1.46
> :06-:09 1.26
> :09-:12 1.18
> :12-:15 1.19
> :15-:18 1.16
> :18-:21 1.12
> :21-:24 0.99
> For some reason, the quick attempt is not quite as effective as the 3-6
> second attempt (the difference is not significant). Go figure. In any case,
> I think we're not just talking about fast breaking teams, but fast pace
> teams -- to me, that is a team that pushes the ball upcourt quickly, not
> necessarily taking an attempt quickly. I'm willing to include 3-6 second
> attempts in a "fast pace" definition.
> I'm afraid I can't finish this post right now. If you want to reassess your
> suppositions in light of the numbers above, I'd like to see it. I'm also
> willing to do a bit more work on this over the weekend if you have anything
> specific you'd like me to check.
the Nets or reflect the supposed NBA average. As for a reassessment of suppositions, I
feel comfortable with my general initial assessment - especially if one allows that a real
fast break could take more than 3 seconds and still be considered such. Just eye-balling
the data, one expects about 0.3 more points per possession when one shoots within the
first 6 seconds and than when one shoots afterwards. My guess of 0.5 still seems about
right to me when one includes turnovers in the mix - as the expectation of these should
be a key factor in deciding when to shoot, and these become more likely the longer one
gets into the shoot clock.
So, what this new data tells me (assuming it is for the Nets) is that the fast break is likely
even more important for their net superiority (pun intended) that the one fifth I initially
surmised (given that the Nets take 1.7% more first shot attempts within the first 6 seconds
and that the 0.5 additional points per possession - turnovers included - conjecture is
And regarding specific things that one should check, if one wishes to look at the value of
Jason Kidd, it would be great to have the points per possession (turnovers included) and
compare that to the NBA average. But then again, more data is always better.