View previous topic :: View next topic |
Author |
Message |
HoopStudies
Joined: 30 Dec 2004 Posts: 705 Location: Near Philadelphia, PA
|
Posted: Fri Aug 19, 2005 7:43 pm Post subject: |
|
|
Confusion over actual poss data is why I posted actual detailed possession data for Seattle in many games this past year. Those are defined exactly as I do in the book. Those are somewhere in a thread here. (I'm still in Brazil and still trying to relax, usually successfully, so I'm not looking now.)
Yeah, both methods overestimate possessions, the simple version more. Just as Bill James had many different versions of runs created, I expect many different versions of possession estimates. I use the more technical version because it's programmed into everything I use. Others use the other form and I just correct the bias approximately. Not a big deal.
The technical form did come from an analysis of actual game data and it showed a better fit, though that was several years ago with a smaller data set. With Roland, I did figure out that the missing information to really make poss estimates better is team offensive and defensive rebounds (not dead ball rebounds). In a quick cut, we got a FT multiplier of around .45, which was good because it goes with other studies. I think we used the more technical approximation (without the 1.07, which I always disliked) because it makes better use of info. But that is the study that should be done in greater detail...
All in all, though, it doesn't make a huge difference which formula you use. The errors are fairly uniform across teams/players. _________________ Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers. |
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 783 Location: Toronto
|
Posted: Fri Aug 19, 2005 8:41 pm Post subject: |
|
|
jkubatko wrote: | Here's a question: What percentage of made free throws come on the *last* free throw attempt? |
FTAs in parentheses. Not all games in each season used -- 2002 sample is especially reduced. 2005 includes some playoff games, the other seasons are only reg season.
Code: | YEAR 1 of 1 1 of 2 2 of 2 1 of 3 2 of 3 3 of 3 TECH
2002 76.1% 74.9% 79.6% 78.0% 89.5% 88.1% 79.9%
(1897) (9595) (9612) (59) (57) (59) (826)
2003 72.2% 72.9% 78.2% 82.6% 85.5% 86.1% 80.1%
(3728) (19386) (19336) (138) (138) (137) (1551)
2004 71.7% 72.7% 77.1% 77.0% 87.6% 81.3% 80.6%
(4549) (24717) (24691) (217) (217) (219) (1667)
2005 70.5% 72.7% 76.7% 76.4% 79.4% 82.9% 81.4%
(5249) (26503) (26642) (195) (194) (193) (1706)
|
_________________ ed |
|
Back to top |
|
|
Dan Rosenbaum
Joined: 03 Jan 2005 Posts: 540 Location: Greensboro, North Carolina
|
Posted: Fri Aug 19, 2005 10:44 pm Post subject: |
|
|
Ed Küpfer wrote: | jkubatko wrote: | Here's a question: What percentage of made free throws come on the *last* free throw attempt? |
FTAs in parentheses. Not all games in each season used -- 2002 sample is especially reduced. 2005 includes some playoff games, the other seasons are only reg season.
Code: | YEAR 1 of 1 1 of 2 2 of 2 1 of 3 2 of 3 3 of 3 TECH
2002 76.1% 74.9% 79.6% 78.0% 89.5% 88.1% 79.9%
(1897) (9595) (9612) (59) (57) (59) (826)
2003 72.2% 72.9% 78.2% 82.6% 85.5% 86.1% 80.1%
(3728) (19386) (19336) (138) (138) (137) (1551)
2004 71.7% 72.7% 77.1% 77.0% 87.6% 81.3% 80.6%
(4549) (24717) (24691) (217) (217) (219) (1667)
2005 70.5% 72.7% 76.7% 76.4% 79.4% 82.9% 81.4%
(5249) (26503) (26642) (195) (194) (193) (1706)
|
|
Given Ed's data above (and a few corrections to equalize some of the free throw discrepancies), here is the percentage of made free throws that end a possession.
2002: 45.0%
2003: 45.5%
2004: 45.6%
2005: 45.4%
Here is the ratio of possessions to free throw attempts, which again is nice confirmation that 0.44 is the right multiplier when converting free throw attempts to possessions.
2002: 43.7%
2003: 43.9%
2004: 44.3%
2005: 44.1% |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 3442 Location: Delphi, Indiana
|
Posted: Sat Aug 20, 2005 6:50 am Post subject: |
|
|
Ed Küpfer wrote: |
Code: | YEAR 1 of 1 1 of 2 2 of 2 1 of 3 2 of 3 3 of 3 TECH
2002 76.1% 74.9% 79.6% 78.0% 89.5% 88.1% 79.9%
(1897) (9595) (9612) (59) (57) (59) (826)
2003 72.2% 72.9% 78.2% 82.6% 85.5% 86.1% 80.1%
(3728) (19386) (19336) (138) (138) (137) (1551)
2004 71.7% 72.7% 77.1% 77.0% 87.6% 81.3% 80.6%
(4549) (24717) (24691) (217) (217) (219) (1667)
2005 70.5% 72.7% 76.7% 76.4% 79.4% 82.9% 81.4%
(5249) (26503) (26642) (195) (194) (193) (1706)
|
|
3-point shooters are better FT shooters than and-1 scorers (unsurprisingly).
The 2nd (and 3rd) FT% are about .060 higher than 1st FT% (significant difference).
Technical FT shooters are hardly better than the average 3-pt shooter who gets fouled (surprisingly).
Great work, Ed. I assume the '1 of 1' column does not include the 'tech' numbers ? |
|
Back to top |
|
|
HoopStudies
Joined: 30 Dec 2004 Posts: 705 Location: Near Philadelphia, PA
|
Posted: Thu Aug 25, 2005 10:01 am Post subject: Re: Possession Estimators |
|
|
jkubatko wrote: | Kevin Broom recently e-mailed me to ask why the league efficiency on the Basketball-Reference.com team pages was 103, but the league rating on the player pages was 106. League efficiency and league rating are both estimates of points per 100 possessions. "Efficiency" is John Hollinger's term, while "rating" is Dean Oliver's term. Kevin went to 82games.com and recorded the actual possessions per game for each team over the last three years. I then computed estimated possessions using Hollinger's formula and Oliver's formula:
Code: |
Hollinger = FGA + 0.44*FTA - ORB + TO
Oliver = FGA + 0.4*FTA - (ORB/(ORB+(oppTRB-oppORB)))*(FGA-FG)*1.07 + TO
|
Using the data collected by Kevin, the root mean square error (rmse) of Hollinger's estimates is 3.91 possessions per game, while the rmse of Oliver's estimates is 1.49 possessions per game. It is interesting to note that Hollinger's formula always produces an overestimate of team possessions per game (errors ranging from -6.21 to -2.37, with error calculated as actual minus estimated). Oliver's formula produced an overestimate in more than 90% of all cases, with errors ranging from -3.79 to 0.31.
|
Note that when I use the data I recorded this year very precisely, my formula with 1.07 actually underestimates real possessions by an average of 1 per game (if I remember correctly). I did a quick check and I think I found that a multiplier of 1.03 worked best when not trying to actually use team rebound data which works a lot better. _________________ Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers. |
|
Back to top |
|
|
oprice
Joined: 25 Aug 2005 Posts: 2 Location: North Carolina
|
Posted: Thu Aug 25, 2005 1:34 pm Post subject: |
|
|
Does anyone here have a good possession estimator for college ball (NCAA Men's Div 1)? I believe Ken Pomeroy has suggested FGA + 0.46FTA - OREB + TO. I guess the main differences between the college and pro game (for the purpose of poss. estimating) are the 1-and-1 foul situations and the team/dead ball rebound designations. Speaking of, shouldn't the NBA suck it up and adopt the NCAA method of keeping track of team and dead ball rebounds?
And regarding the FTA coefficient, it would seem to me to be dependent upon a team's FT%, at least in college ball, as teams that shoot a higher percentage will shoot slightly more FTs per possession. So that coefficient will be inversely proportional to FT%. Has anyone put any thought into this? |
|
Back to top |
|
|
kenpom
Joined: 28 Mar 2005 Posts: 7 Location: Cheyenne, WY
|
Posted: Thu Sep 01, 2005 9:02 am Post subject: |
|
|
I suppose this is a good opportunity to stop lurking and contribute something. (If only there was an "ACBR" message board, I would be a regular there. Kevin?)
I looked at 30 somewhat random college games from 2005. The average FTA/poss was .474. It's a small sample, but the 95% confidence range was .463-.485.
It seems to me that a theoretical multiplier could be computed if we know how many of each type of foul occur, so I recorded the types of fouls that occurred before the bonus was in effect in those 30 games.
415 (59.7%) Non-shooting
214 (30.8%) Shooting, two-shot
59 (8.5%) And-one
6 (0.9%) Shooting, three-shot
1 Technical, two-shot
A flaw in my method was that I didn't document offensive fouls.
I was planning to start using .475 for the time being. But some of my back-of-the-envelope theoretical calculations indicate it should be .46-.47. Heck, why don't we just round up to .5 and call it good.
As far as team FT% affecting the multiplier, there are only about three one-and-one situations per team per game, so I don't think it has enough of an effect for me to sweat about.
Ken |
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 783 Location: Toronto
|
Posted: Thu Sep 01, 2005 9:17 am Post subject: |
|
|
kenpom wrote: | I suppose this is a good opportunity to stop lurking and contribute something. (If only there was an "ACBR" message board, I would be a regular there. Kevin?) |
I'd like to address this right here. Speaking personally, my interest here is in the analytic side of stats, and with that in mind, hoops is hoops is hoops and college ball certainly deserves a place in an analytic setting like this, along with international ball. The methods we use for pro hoops should have applications elsewhere, and if they don't, that would be interesting too. My primary interest is in pro hoops, and all the work I do involves that, but I would love to read what others are doing in other areas. I figure stats dorkdom is universal. _________________ ed |
|
Back to top |
|
|
HoopStudies
Joined: 30 Dec 2004 Posts: 705 Location: Near Philadelphia, PA
|
Posted: Thu Sep 01, 2005 3:26 pm Post subject: |
|
|
Ed Küpfer wrote: | kenpom wrote: | I suppose this is a good opportunity to stop lurking and contribute something. (If only there was an "ACBR" message board, I would be a regular there. Kevin?) |
I'd like to address this right here. Speaking personally, my interest here is in the analytic side of stats, and with that in mind, hoops is hoops is hoops and college ball certainly deserves a place in an analytic setting like this, along with international ball. The methods we use for pro hoops should have applications elsewhere, and if they don't, that would be interesting too. My primary interest is in pro hoops, and all the work I do involves that, but I would love to read what others are doing in other areas. I figure stats dorkdom is universal. |
I personally second this. I do work now on basketball from HS to college to international and the women's side. I've felt it inevitable that the APBR would become ABR, Association for Basketball Research. Which makes it abrmetrics and a lot easier to say... _________________ Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers. |
|
Back to top |
|
|
Dan Rosenbaum
Joined: 03 Jan 2005 Posts: 540 Location: Greensboro, North Carolina
|
Posted: Thu Sep 01, 2005 4:03 pm Post subject: |
|
|
I third the motion. ABRmetrics sounds good to me. We are studying not basketball - not just professional basketball. |
|
Back to top |
|
|
Ben
Joined: 13 Jan 2005 Posts: 264 Location: Iowa City
|
Posted: Thu Sep 01, 2005 5:16 pm Post subject: |
|
|
I think many people in APBR were drawn to NBA not so much out of love for the NBA over college ball, but because the quality of the data is so much better in the NBA. I think you find the same thing in social science where Americans probably seem more provincial than they actually are because there is so much data available for the U.S. |
|
Back to top |
|
|
mtamada
Joined: 28 Jan 2005 Posts: 375
|
Posted: Fri Sep 02, 2005 12:29 am Post subject: |
|
|
HoopStudies wrote: | I've felt it inevitable that the APBR would become ABR, Association for Basketball Research. |
True, although after growth, it'll have to split again, into college specialists and pro specialists.
Quote: | Which makes it abrmetrics and a lot easier to say... |
The old rec.sport.basketball newsgroup called their stuff (including both the statistical and non-statistical stuff) "hoopology". I'd go for "hoopstats" personally. "Abrmetrics" is equally descriptive and sounds less flippant, but it's also to me too derivative of "sabrmetrics". Plus, there actually is no "ABR" at the moment, as far as I know. |
|
Back to top |
|
|
Kevin Pelton Site Admin
Joined: 30 Dec 2004 Posts: 969 Location: Seattle
|
Posted: Fri Sep 02, 2005 12:35 am Post subject: |
|
|
I think we can be inclusive of all types of basketball stats research/discussion without needing to change the name. This is a self-serving argument, because it would be a pain in the behind to change the name at this point.
In addition to the many things mentioned here (and I'm a great fan of Ken's work), there is of course the WNBA, which is also professional basketball but still distinctly different. John Maxwell and I have done some WNBA work and he reminded me of my error when I intially subtitled this forum your place for NBA discussion as opposed to the all-inclusive basketball discussion.
Of course anyone who likes baseball and football can just go elsewhere as far as I'm concerned.
Last edited by Kevin Pelton on Mon Oct 24, 2005 12:46 am; edited 1 time in total |
|
Back to top |
|
|
kenpom
Joined: 28 Mar 2005 Posts: 7 Location: Cheyenne, WY
|
Posted: Fri Sep 02, 2005 3:35 am Post subject: |
|
|
I didn't expect the discussion to go this way. It is nice to know some college talk is welcomed. But I'll probably continue to spend most of my time here reading - and dreaming of the day when there's an 82games.com for the college game. |
|
Back to top |
|
|
Ed Küpfer
Joined: 30 Dec 2004 Posts: 783 Location: Toronto
|
Posted: Fri Sep 02, 2005 11:13 am Post subject: |
|
|
admin wrote: | Of course anyone who likes baseball and football can just go elsewhere as far as Im' concerned. |
Bah. That's too bad. I've been doing some work on predicting football outcomes using a method similar to the one I used in the Predicting the future thread. Ah well. _________________ ed |
|
Back to top |
|
|
|