View previous topic :: View next topic |
Author |
Message |
mateo82
Joined: 06 Aug 2005 Posts: 211
|
Posted: Fri Jan 19, 2007 9:37 am Post subject: Hollinger power ranking |
|
|
anyone care to give a crack at the methodology behind hollinger's new power ranking? At first glance it seems reasonable enough for me, but I'd have to hear why he chose the values he did.
Code: | RATING = (((SOS-0.5)/0.037)*0.67) + (((SOSL10-0.5)/0.037)*0.33) + 100 + (0.67*(MARG+(((ROAD-HOME)*3.5)/(GAMES))) +
(0.33*(MARGL10+(((ROADL10-HOMEL10)*3.5)/(10)))))
SOS = Season Win/Loss percentage of team’s opponents, expressed as a decimal (e.g., .500)
SOSL10 = Season Win/Loss percentage of team’s last 10 opponents, expressed as a decimal (e.g., .500)
MARG = Team’s average scoring margin
MARGL10 = Team’s average scoring margin over the last 10 games
HOME = Team’s home games
HOMEL10 = Team’s home games over the last 10 games
ROAD = Team’s road games
ROADL10 = Team’s road games over the last 10 games
GAMES = Team’s total games |
|
|
Back to top |
|
|
kjb
Joined: 03 Jan 2005 Posts: 864 Location: Washington, DC
|
Posted: Fri Jan 19, 2007 9:41 am Post subject: |
|
|
My biggest question is why he gives as much emphasis as he does to the last 10 games. I vaguely recall an Ed Kupfer post saying that giving extra credit to recent games does not improve results. |
|
Back to top |
|
|
mateo82
Joined: 06 Aug 2005 Posts: 211
|
Posted: Fri Jan 19, 2007 9:47 am Post subject: |
|
|
I don't have a problem with that, that's a large part of what a power ranking is, how teams are currently playing. if it were simply an effort to measure team performance for the current year that would be one thing, but it seems like an effort to make a more objective power ranking. |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 3554 Location: Hendersonville, NC
|
Posted: Fri Jan 19, 2007 10:07 am Post subject: |
|
|
Extra weight to recent games may give you an edge, if there is a trend: The Suns are on a roll; the Spurs are losing steam. But if it's due to a player or two being out, it's probably more valuable to know when the player(s) are due back. If you know they're coming back tomorrow, you would do better to discount the recent games. _________________ `
36% of all statistics are wrong |
|
Back to top |
|
|
deepak
Joined: 26 Apr 2006 Posts: 664
|
Posted: Fri Jan 19, 2007 11:37 am Post subject: |
|
|
Is there any good way to account for injuries in this sort of formula? |
|
Back to top |
|
|
asimpkins
Joined: 30 Apr 2006 Posts: 244 Location: Pleasanton, CA
|
Posted: Fri Jan 19, 2007 11:56 am Post subject: |
|
|
I think giving more credit to the last 10 games makes it a little more entertaining at least, even if it doesn't mean too much.
On the other hand, I remember reading some good articles at the end of last year about how the Pistons had cooled off considerably since their hot start and how Miami was finishing very strong -- and that even though their records were still very different, the teams were playing much closer to each other. Of course, those trends continued and Miami beat Detroit in the playoffs. This formula would pick up on something like that. |
|
Back to top |
|
|
Gary C
Joined: 14 Apr 2006 Posts: 69
|
Posted: Fri Jan 19, 2007 12:40 pm Post subject: |
|
|
At the 41 game mark, "Last 10 games" gets replaced by "Last 25 games" as the article notes. It's just an attempt to credit teams who are getting better or worse over the course of a season objectively. As with any objective measure, it will sometimes be misleading without context. |
|
Back to top |
|
|
jkubatko
Joined: 05 Jan 2005 Posts: 702 Location: Columbus, OH
|
Posted: Fri Jan 19, 2007 12:49 pm Post subject: |
|
|
Gary C wrote: | At the 41 game mark, "Last 10 games" gets replaced by "Last 25 games" as the article notes. |
No, it's most recent 25 *percent* of their schedule. For example, after 60 games we would look at their last 15 games. _________________ Regards,
Justin Kubatko
Basketball-Reference.com |
|
Back to top |
|
|
mateo82
Joined: 06 Aug 2005 Posts: 211
|
Posted: Fri Jan 19, 2007 2:16 pm Post subject: |
|
|
I was hoping someone could explain the formula, such as why he divides by 0.037 and so forth. Sorry, I have a BA, this stuff is not obvious to me. |
|
Back to top |
|
|
deepak
Joined: 26 Apr 2006 Posts: 664
|
Posted: Fri Jan 19, 2007 2:24 pm Post subject: |
|
|
mateo82 wrote: | I was hoping someone could explain the formula, such as why he divides by 0.037 and so forth. Sorry, I have a BA, this stuff is not obvious to me. |
Just a wild guess, but it looks like he's standardizing the SOS and SOSL10 terms by subtracting from 0.5 (average win%) and dividing by 0.037 (perhaps an estimate for standard deviation of win% league wide). |
|
Back to top |
|
|
THWilson
Joined: 19 Jul 2005 Posts: 164 Location: phoenix
|
Posted: Fri Jan 19, 2007 2:35 pm Post subject: |
|
|
deepak_e wrote: | mateo82 wrote: | I was hoping someone could explain the formula, such as why he divides by 0.037 and so forth. Sorry, I have a BA, this stuff is not obvious to me. |
Just a wild guess, but it looks like he's standardizing the SOS and SOSL10 terms by subtracting from 0.5 (average win%) and dividing by 0.037 (perhaps an estimate for standard deviation of win% league wide). |
I also find it non-obvious, and there is no explanation provided, so it would be cool if John swings by to explain. Standard deviation in winning percentage is currently on the order of 0.14 while standard deviation for winning percentage faced is closer to 0.013... |
|
Back to top |
|
|
cherokee_ACB
Joined: 22 Mar 2006 Posts: 157
|
Posted: Fri Jan 19, 2007 4:16 pm Post subject: |
|
|
mateo82 wrote: | I was hoping someone could explain the formula, such as why he divides by 0.037 and so forth. Sorry, I have a BA, this stuff is not obvious to me. |
I believe he's trying to convert the opponents winning percentage into an expected point differential (i.e., compute the inverse pythagorean). +0.037 is a good estimation of the increase of pythWin% for every +1.0 in point differential.
Last edited by cherokee_ACB on Fri Oct 10, 2008 2:02 pm; edited 1 time in total |
|
Back to top |
|
|
Dan Rosenbaum
Joined: 03 Jan 2005 Posts: 541 Location: Greensboro, North Carolina
|
Posted: Mon Jan 22, 2007 1:09 am Post subject: |
|
|
Hollinger's new power rankings will be great for Chicago. In a week when they lose three games in the 4th quarter because they can't score, but win the fourth game by 40 over some hapless opponent, they will be able to feel good because they are moving up in the Hollinger power rankings.
It would be a lot of work, but it would be so much better if these ratings re-weighted points per possession within a game by the probability that a point actually changes the outcome. I don't think we learn a lot from those teams that are effective turning 25 point routs into 40 point routs vs. those teams that turn 25 point routs in 10 point wins that were never in doubt.
I am not claiming that margin of victory has no informational content, but all points are not created equal. Last year if you did something like I mention above, San Antonio did not look nearly as impressive. And that was a pretty good predictor of how San Antonio did in the playoffs. |
|
Back to top |
|
|
Mike G
Joined: 14 Jan 2005 Posts: 3554 Location: Hendersonville, NC
|
Posted: Mon Jan 22, 2007 7:08 am Post subject: |
|
|
Wow, Dan. You sound like one frustrated Bulls fan.
I think the assumption of pt-diff models is that over the course of a season, some 25-pt games will turn into 10-pt games, and others will turn into 40-pt games, but there won't be a significant tendency from one team to the next. A team with a deep bench may tend toward the blowout. Doesn't this mean they're really a better team than the team without a bench?
The Spurs are an interesting choice for your example. They're kind of famous for giving long rest to key players: Duncan averages 34 minutes, Ginobili 28 ... And this year, they're on a pace to win 8 games less than their Pythex suggests.
What I'd really like to see is a model that answers the perennial question: Would da Bears beat da Bulls? _________________ `
36% of all statistics are wrong |
|
Back to top |
|
|
Harold Almonte
Joined: 04 Aug 2006 Posts: 616
|
Posted: Mon Jan 22, 2007 8:16 am Post subject: |
|
|
With good bench or not, it doesn't change that a 25-40 pts. margin in the 4th. quarter is low worth-garbage time with a far win probability (it must be adjusted for that kind of development game no matter same quality of opponents). That's no a real competition, the game passed the first three quarters. Points per possession is a relative valid team measure. |
|
Back to top |
|
|
|