- --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer

<igorkupfer@r...> wrote:> <snip>

vs .457), as

>

> The following is a comparison of the two FTA coefficients (.4

> shown by the differences between offensive possessions and defensive

larger

> possessions,by season. The smaller coefficient outperforms the

> exactly half the time, although with a larger standard error. The

difference

> is small enough (to me) as to not warrant using the extra decimal

places.

> What is needed is some empirical data, counting actual possessions

in a

> game, and seeing how close we are.

I've got some of these. I'll try to pass them on after I'm done

traveling (again).

My general comment to Ben when he first passed on his note to me was

that he may be right that 0.45 works better to account for FT

possessions. However, there are minor errors in the rest of the

possession formula associated with team ORs and other little things

that get lumped into that multiplier on FTA. Or you can get more

complex with your possession formula, something BobC has done.

I share the sentiment that it doesn't matter too much. Soon, we will

be able to just count possessions for each team and do the simple

division to get pts/poss. Still, the formula estimate will remain

useful (in historical work, in individual possession calcs), but the

difference between 0.4 and 0.45 won't make a huge difference in

estimates. Plus, the multiplier changes with rules and enforcement

of such. It was lower in the 3-to-make-2 era, for example, and is

higher in the 1-and-1 world of college hoops.

DeanO *I have noticed that many people have used the following formula to*

estimate possessions:

FGA + (FTA)(.4) + TO - OReb

The times fouled shooting factor should be .457; not .4.

Or you can get more complex with your possession formula, something BobC has done.

the proper formula is:

FGA + FTA/2 + TO - oreb - techFTA - (AOSF FGM) - flagrants

(denominator 2 larger prior to 81-82 when 3 to make 2 bonus was in affect), but i add 0.33 to account for the techs and flagrants...

and not more complex, just more exact, simply because it works in a simulation, which does not take into account technicals and flagrants, but does account for AOSF FGM (act of shooting fouls FGM, i.e. traditional 3 pt plays)...

your average for 97-98 (the last year i updated the sim) was 92.4 poss/48 min/team. the sim uses 91.5 on average and the stats come out pretty much exact (i.e. avg pts/g/team right on the money, well within 1%)....

if you ever run the B-BALL sim for a full season (takes about 40 seconds on a pentium to run an 1189 game season) it lists each team's poss/g in the final stats file...

bob chaikin

bchaikin@...- I agree with the general sentiment -- if 0.4 vs. 0.44 would only

change the result between two players are teams with microscopic

differences.

Also, in the specific example that brought up this thread, there's a

couple things missing -- times fouled on 3-pointers, flagrant fouls

and breakaway fouls -- that would further adjust the coefficient.

--- In APBR_analysis@yahoogroups.com, "Dean Oliver" <deano@r...>

wrote:> --- In APBR_analysis@yahoogroups.com, igor eduardo küpfer

defensive

> <igorkupfer@r...> wrote:

> > <snip>

> >

> > The following is a comparison of the two FTA coefficients (.4

> vs .457), as

> > shown by the differences between offensive possessions and

> > possessions,by season. The smaller coefficient outperforms the

possessions

> larger

> > exactly half the time, although with a larger standard error. The

> difference

> > is small enough (to me) as to not warrant using the extra decimal

> places.

> > What is needed is some empirical data, counting actual

> in a

was

> > game, and seeing how close we are.

>

> I've got some of these. I'll try to pass them on after I'm done

> traveling (again).

>

> My general comment to Ben when he first passed on his note to me

> that he may be right that 0.45 works better to account for FT

will

> possessions. However, there are minor errors in the rest of the

> possession formula associated with team ORs and other little things

> that get lumped into that multiplier on FTA. Or you can get more

> complex with your possession formula, something BobC has done.

>

> I share the sentiment that it doesn't matter too much. Soon, we

> be able to just count possessions for each team and do the simple

the

> division to get pts/poss. Still, the formula estimate will remain

> useful (in historical work, in individual possession calcs), but

> difference between 0.4 and 0.45 won't make a huge difference in

> estimates. Plus, the multiplier changes with rules and enforcement

> of such. It was lower in the 3-to-make-2 era, for example, and is

> higher in the 1-and-1 world of college hoops.

>

> DeanO - --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
> the proper formula is:

affect),

>

> FGA + FTA/2 + TO - oreb - techFTA - (AOSF FGM) - flagrants

>

> (denominator 2 larger prior to 81-82 when 3 to make 2 bonus was in

> but i add 0.33 to account for the techs and flagrants...

The problem is that your last 3 terms aren't available in most stat

>

books so we end up estimating them. I just end up rolling them into

the multiplier on FTA, but there are other ways.

DeanO