|
APBRmetrics The statistical revolution will not be televised.
|
View previous topic :: View next topic |
What is the appropriate multiplier for free throws? |
.4 |
|
0% |
[ 0 ] |
.44 |
|
88% |
[ 8 ] |
.45 |
|
11% |
[ 1 ] |
|
Total Votes : 9 |
|
Author |
Message |
Kevin Pelton Site Admin
Joined: 30 Dec 2004 Posts: 979 Location: Seattle
|
Posted: Thu Jan 27, 2005 10:59 am Post subject: More standardization |
|
|
Trying to explain what we're doing to other people constantly reminds me how confusing it is to have multiple definitions. So would it be possible for us to agree on a free-throw multiplier? |
|
Back to top |
|
|
jkubatko
Joined: 05 Jan 2005 Posts: 702 Location: Columbus, OH
|
Posted: Thu Jan 27, 2005 11:14 am Post subject: Re: More standardization |
|
|
admin wrote: | Trying to explain what we're doing to other people constantly reminds me how confusing it is to have multiple definitions. So would it be possible for us to agree on a free-throw multiplier? |
Could we have people present cases for each of the above? Perhaps John Hollinger could make the case for 0.44, Dean Oliver for 0.4, and ??? for 0.45. _________________ Regards,
Justin Kubatko
Basketball-Reference.com |
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 787 Location: Toronto
|
Posted: Thu Jan 27, 2005 11:46 am Post subject: Re: More standardization |
|
|
jkubatko wrote: | Could we have people present cases for each of the above? Perhaps John Hollinger could make the case for 0.44, Dean Oliver for 0.4, and ??? for 0.45. |
It's important to remember that there's an essential tension in everything we do between simplicity and accuracy. Almost always, a move towards one brings us away from the other.
On the simplicity side, there's 0.4. That's a single significant digit to remember.
On the other side, there's 0.44. I think different methods have all supported the accuracy of this number.
I did a little test. Using the same player's numbers, I changed only the multiplier. Here's how his stats changed:
Code: | POSS ORTG FLR% DPOSS DRTG STP%
0.4 761 118.5 53.3% 869 103.6 51.1%
0.44 774 115.6 53.9% 878 102.6 50.6%
0.5 794 111.5 54.8% 891 101.2 49.9% |
Those changes are huge. We need to decide on a number. I'm going to go with the most accurate -- there is no need for simplicity. _________________ ed |
|
Back to top |
|
|
Dan Rosenbaum
Joined: 03 Jan 2005 Posts: 541 Location: Greensboro, North Carolina
|
Posted: Thu Jan 27, 2005 11:46 am Post subject: |
|
|
Ed Küpfer wrote: | Here's how the types of FTs break down:
Code: | FTAs
TYPE 1 2 3 TOTAL
Away from Play 35 4 0 39
Clear Path 28 0 0 28
Delay Technical 4 0 0 4
Double Technical 1 0 0 1
Elbow 0 4 0 4
Flagrant Type 1 0 77 0 77
Flagrant Type 2 0 12 0 12
Hanging Technical 12 0 0 12
Illegal Defense 341 0 0 341
Inbound 0 24 0 24
Loose Ball 3 1002 0 1005
Non Supported Technical 2 0 0 2
Offensive 0 8 0 8
Personal 8 6628 3 6639
Punching 2 0 0 2
Shooting 2473 18578 303 21354
Taunting Technical 6 0 0 6
Technical 446 0 0 446
TOTAL 3361 26337 306 30004 |
The FTAs that represent the end of a possession -- Shooting, Loose Ball, Away from Play, and Personal -- sum to 13259, or 44.2% of the total. Nice to have that FTA coefficent confirmed. |
I think this data represents all of the free throws this season up through January 21st. This is just fabulous to have these data.
I think the number of possessions is equal to:
two-shot away from play divided by two = 4/2 = 2
two-shot elbow divided by two = 4/2 = 2
two-shot inbound divided by two = 24/2 = 12
two-shot loose ball divided by two = 1002/2 = 501
two-shot offensive divided by two = 8/2 = 4
two-shot personal divided by two = 6628/2 = 3314
three-shot personal divided by three = 3/3 = 1
two-shot shooting divided by two = 18,578/2 = 9289
three-shot shooting divided by three = 303/3 = 101
Add this all up and you get 13,226, which is 44.0808% of the 30,004 total free throws.
Note that a couple of these categorizations are a little fishy, but ...
40 percent of 30,004 is 12,002
43 percent of 30,004 is 12,902
44 percent of 30,004 is 13,202
45 percent of 30,004 is 13,502
50 percent of 30,004 is 15,002
So even if we haggle a bit over a few of these foul categories, we are going to end up pretty close to 0.44. Like others, I have also found 0.44 to work best in terms of balancing offensive and defensive team possessions. |
|
Back to top |
|
|
HoopStudies
Joined: 30 Dec 2004 Posts: 706 Location: Near Philadelphia, PA
|
Posted: Thu Jan 27, 2005 1:07 pm Post subject: Re: More standardization |
|
|
Ed Küpfer wrote: |
It's important to remember that there's an essential tension in everything we do between simplicity and accuracy. Almost always, a move towards one brings us away from the other.
On the simplicity side, there's 0.4. That's a single significant digit to remember.
On the other side, there's 0.44. I think different methods have all supported the accuracy of this number.
... I'm going to go with the most accurate -- there is no need for simplicity. |
In all cases, there are 3 considerations in making this kind of decision:
1. Reality or accuracy
2. Simplicity
3. Consistency
These are listed in the importance I personally place on them. The 0.44 is more accurate. A few people have shown that. And it's not really much more complex or inconsistent since adding a digit ain't hard and Hollinger has had .44 out for a while. I'm certainly planning on using it and do use it in some things I've done, even though I've been using .4 for nearly 20 years.
The only realm in which it has a problem is in the possession calculation. I've been tracking real possessions this year pretty carefully, even knocking off the fake ones that come out of play by plays at the end of quarters. And if you just use 0.44*FTA you seem to overestimate those real possessions because of slight errors in the other part of the equation. This is a minor point. I've been dealing with various formulas for possessions for years. Renormalizing them to some mean is not a huge deal.
Anyway, I do think the .44 is the best factor to use, but it has some cascading effects in terms of fixing other things.
I'll try to post some of the study on possessions so someone else can polish the job. Just not this week. _________________ Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers. |
|
Back to top |
|
|
gabefarkas
Joined: 31 Dec 2004 Posts: 1313 Location: Durham, NC
|
Posted: Fri Jan 28, 2005 5:01 pm Post subject: Re: More standardization |
|
|
Ed Küpfer wrote: |
Those changes are huge. We need to decide on a number. I'm going to go with the most accurate -- there is no need for simplicity. |
they are indeed huge, but they are huge for every player's stats for which the multiplier is applied (ie, somewhat evenly across the board). |
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 787 Location: Toronto
|
Posted: Fri Jan 28, 2005 5:07 pm Post subject: Re: More standardization |
|
|
gabefarkas wrote: | they are huge for every player's stats for which the multiplier is applied (ie, somewhat evenly across the board). |
They are internally consistent, but if someone tells me that some player is hitting 1.02 points per possession, I don't want to be asking about which FT coefficient was used so I can compare it to my numbers. I mean, if it's going to cause some confusion among us knowledgable types, imagine the confusion among the less statistically savvy. _________________ ed |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|