- --- In APBR_analysis@yahoogroups.com, "aaronkoo" <deano@r...> wrote:
> --- In APBR_analysis@yahoogroups.com, "schtevie2003"

<schtevie@h...>

> wrote:

declines

> > > You are basically assuming that offensive performance declines

> with

> > > time left on the shot clock. What I've found is that it

> > > with <3 s on the clock but is not significantly impacted before

web

> > > then. That implies that teams have pretty constant offensive

> > > performance up to that point and that the assumptions in that

> > > page are pretty accurate up to the point of <3 s on the clock.

is

> >

> > You are correct that I am effectively assuming that offensive

> > performance declines with time left on the shot clock. But the

> real

> > assumption is that teams aspire to optimal play - that is

> maximizing

> > their offensive productivity - given this assumption, the theory

> > dead solid that offensive performance is expected to decline as a

give

> > function of elapsed shot clock time.

>

> The real assumption is that teams aspire to win, which is different

> than maximizing offensive productivity. If slowing the game can

> you an extra 8% and maximizing offensive productivity gives you an

high

> extra 1%, which do you go with? I'm not sure how steep or flat you

> think the curve is for offensive productivity as a function of time

> left on the clock. I'm not even convinced that productivity is

> with 21-24s on the clock when turnover rates can be high.

It is hard to keep an eye on the forest for all the trees in these

strings, so let me restate the larger "pace" argument:

Except for milking the clock when ahead at the end of the game and

possibly the "shoot slow if you suck and hope you are lucky", I see

no way, theoretically, for a team to try and control game pace

without costing itself points and the likelihood of victory.

That said, I agree with the above statement that the real assumption

is that teams aspire to win, but disagree with the statement that it

is necessarily different than maximizing offensive productivity -

except in the two realized cases above.

As to how steep or flat I think the curve is for offensive

productivity as a function of time left on the clock, I don't know

either, as I have only a general formula with hypothetical variables

and lack true data to plug in. That said, I will shuffle through

some old paper and offer up hypothetical slope values that the good

folks here can comment on.

**************************************

> > > So, yes, I do think 8% is within the realm of possibility. 8%

at

> ain't

> > > a lot to gain and can be offset if your team doesn't play well

> a

additional

> > > slow pace (they defy the more general study above, which is

> > > definitely possible). And, yes, throwing away fast breaks

> doesn't

> > > make sense.

> >

> > I think that 8% improvement is a very large number in the context

> of a

> > possibly costless competitive advantage - an expected 4

> wins

understanding

> > per season for an average team in the league if I am

> the

a

> > statistic correctly?

>

> Not quite. That extra 8% came from when they were an underdog and

> big underdog. In normal games, you aren't going to get 8%. Over

the

> course of the season, there are a few games where this matters.

Can you give a graphical argument as to why the gain is greatest when

> Winning 4 seems a little high but I haven't done the calc.

the expected loss is greatest. If I suppose a bell shaped curve

centered around this expected losing margin, then tweak the system

with a slow down strategy and this shifts the distribution such that

a greater amount of probability mass is above zero. If this is the

argument (and I am not saying it is) then suppose a more equal but

still superior opponent. Wouldn't I expect a greater gain in the

probability of victory if a thicker part of the bell curve (assume

same moments) was being pushed past zero?

**********************

> > And I too remember some atrocious pre-shot clock college games.

alter

> But

> > let's pose the question this way. Suppose two equal and average

> NBA

> > teams are playing, and one has a 10 point lead. Pick a time X

> before

> > the end of the game at which point you think it is optimal to

> > one's offensive strategy so that the first priority is

controlling

> > pace.

I'd

> >

>

> Obviously it depends on assumptions about how a team's offensive

> strategy affects its productivity. If I assume that it doesn't, I

> get one answer. If I assume that it does, I get another. Though

> like to get the answer either way (and think I can -- I have my old

Yeah sure it depends on assumptions, but that is the game...Just

> simulator sitting around somewhere), I don't know if we have a good

> way to make that assumption.

>

> DeanO

trying to establish a hypothetical tradeoff to anchor our

expectations. - --- In APBR_analysis@yahoogroups.com, "schtevie2003" <schtevie@h...>

wrote:> It is hard to keep an eye on the forest for all the trees in these

Forest is about the same. I think we may have different senses of

> strings, so let me restate the larger "pace" argument:

>

> Except for milking the clock when ahead at the end of the game and

> possibly the "shoot slow if you suck and hope you are lucky", I see

> no way, theoretically, for a team to try and control game pace

> without costing itself points and the likelihood of victory.

>

the magnitude of "cost". Cost of 1% vs value of 8% would need to be

weighed. In some cases, the cost will be greater than the benefit,

I'm sure.

> Can you give a graphical argument as to why the gain is greatest

when

> the expected loss is greatest. If I suppose a bell shaped curve

It's in the book. But here's a try. Basically the bell curve

describes the point difference between the 2 teams. Centered at +5

for the favorite, spreading the distribution (which is all that

taking a high risk/slow pace strategy does) causes the tail of the

distribution go more across that 0 line (when the underdog wins). If

it's centered at +1 (small favorite), a fair amount of the curve is

already across the 0 line, so spreading the curve doesn't make much

difference.

Is that clear?

>

I can add it to the list.

> Yeah sure it depends on assumptions, but that is the game...Just

> trying to establish a hypothetical tradeoff to anchor our

> expectations.