View previous topic :: View next topic |
Author |
Message |
petrel
Joined: 26 May 2008 Posts: 2001 Location: Atlanta, GA
Back to top |
Posted: 05/29/08 9:39 pm ::: WNBA Hollinger Power Rankings |
Reply |
|
I'm going to try to keep a running count of the Hollinger Power Rankings for the WNBA over the season. These are based on a formula by basketball statistician John Hollinger. Among the factors used to calculate a team's performance are:
a) margin of victory
b) strength of schedule
c) home vs. away games
d) performance overall vs. performance over last 10 games played
As the WNBA season is short, as the season progresses extra weight will be placed on the final ten games played as opposed to the final quarter of games played for the NBA rankings.
Note that this early in the season there will be a lot of fluctuation as teams settle down. Since less than 10 games have been played, "d)" above is not a factor. Strength of schedule will dominate the ratings with small sample sizes.
Rankings as of 5/29/2008
1 Fever 106.922
2 Comets 105.261
3 Liberty 103.873
4 Monarchs 103.460
5 Mercury 103.312
6 Dream 103.298
7 Sky 102.617
8 Lynx 101.890
9 Sparks 98.172
10 Shock 97.899
11 Mystics 97.258
12 Silver Stars 96.357
13 Sun 96.147
14 Storm 95.153
Why are the Comets ranked #2 and the Dream ranked #6? Because they're tied for the toughest schedule played, and the Comets have sucked less egregiously. Why are the Sparks, Silver Stars, Sun and Storm at or near the bottom? Because they've played the four weakest schedules so far this year.
As the season progresses, the cream will rise to the top, and the catfish will fall to the bottom.
--Pet
|
|
UTexRulz23
Joined: 17 Jun 2006 Posts: 2611 Location: Austin, Texas
Back to top |
Posted: 05/29/08 9:42 pm ::: |
Reply |
|
K, I'm a huge Comets fan, but that's some crazy shit right there!
_________________ The Houston Comets disbanding will always be the biggest black cloud in the history of the WNBA. Pitiful.
|
|
braveniler58
Joined: 30 May 2007 Posts: 10537 Location: Arizona
Back to top |
Posted: 05/29/08 10:11 pm ::: |
Reply |
|
Awesome, I was thinking about doing this myself but you beat me to it. Interesting that the Mercury are 0-3, yet just fifth on the list. Wonder what it would be like had we won some more?
_________________ Phoenix Mercury fan for life!
|
|
Shoots&Scores
Joined: 19 Mar 2007 Posts: 1544 Location: Poolside while you shovel snow!
Back to top |
|
recycledcomets
Joined: 04 Mar 2008 Posts: 294 Location: todd mission texas
Back to top |
Posted: 05/30/08 9:40 am ::: |
Reply |
|
Thanks for doing this . . . I just learned about Hollinger (on this board - thanks again, reb!!) and have been curious but my schedule is limited during the season . . . all those games to drive to eats up my time! I'll keep watching your updates to see if I can understand the Hollinger system better . . . thanks again, petrel!
_________________ Houston Comets? History . . .
Now playing in the league with other fondly remembered teams: The Cleveland Rockers, Charlotte Sting, Miami Sol, Portland Fire, Orlando Miracle, Utah Starzz.
|
|
jfhst18
Joined: 15 May 2006 Posts: 308
Back to top |
Posted: 05/30/08 11:05 am ::: |
Reply |
|
Could you spell out your algorithm? I think you'll agree these rankings look kind of funny and it would be interesting to know the exact factor weights you are assigning.
Very interesting project...thanks for your contribution...and will look forward to following your work.
|
|
psheehy
Joined: 07 Feb 2006 Posts: 2972
Back to top |
Posted: 05/30/08 11:11 am ::: |
Reply |
|
The key to know is how you are ranking strength of schedule.
_________________ Coach Agler's 2014 Off Season Task List
1) Think about "what if?" he had gotten that lottery ball in 2015....
2) Remember to watch some NCAA games
3) Invest in mystical restorative ointments for hips, knees, backs and well everything.
|
|
petrel
Joined: 26 May 2008 Posts: 2001 Location: Atlanta, GA
Back to top |
Posted: 05/30/08 11:59 am ::: |
Reply |
|
The full formula is here.
http://insider.espn.go.com/nba/insider/columns/story?columnist=hollinger_john&page=Rankings-Intro&action=login&appRedirect=http%3a%2f%2finsider.espn.go.com%2fnba%2finsider%2fcolumns%2fstory%3fcolumnist%3dhollinger_john%26page%3dRankings-Intro
I'm using Hollinger's formula in the given form. I'm led to believe that Hollinger changes his formula for the NBA by giving extra weight to the last 25 percent of games as the season closes rather than the last 10 games. However, the WNBA plays fewer games and there is no need to make such an adjustment.
Strength of schedule (SOS) will be determined by the combined win-loss of opponents at the time that the rankings come out (not at the time the games are played). If a team plays another team twice, the opponent's win-loss is double-counted. As the year progresses, SOS evens out as every team gets the chance to play every other team. However, SOS will still reflect the strength of one conference over another.
For example, take the Storm, which I believes has one of the bottom four ranked strength of schedule. Looking at Storm games gives us:
Chicago 1-2
Sacramento 1-3
@Phoenix 0-3
@San Antonio 2-1
Combined W/L = 4-9, or .308 or 30.8 percent.
I believe the Storm's combined margin of victory is 1.75 points, which isn't that great for a winning team. Add it all up and it puts the Storm at the bottom.
However, I can't emphasize enough - with very few games played, expect a lot of wildness in the rankings before everyone gets games on the map and the ratings reflect reality a bit better.
--Pet
|
|
StevenHW
Joined: 25 Jul 2005 Posts: 10987 Location: Sacramento, California
Back to top |
|
TDAO
Joined: 11 Jun 2005 Posts: 555
Back to top |
Posted: 05/30/08 3:25 pm ::: |
Reply |
|
Certainly, some weirdness is to be expected early in the season, especially because strength of schedule depends so much on the team itself. That is, Atlanta's opponents have been prety good overall (7-5) but get really good because they've gone 4-0 against the Dream.
That said, I'm not sure about your numbers. When I plug Atlanta and Seattle, to take two extreme examples, into Hollinger's formula, I get 97.7 as a rating for the Storm and 91.9 for the Dream.
Any idea why the discrepancy?
|
|
petrel
Joined: 26 May 2008 Posts: 2001 Location: Atlanta, GA
Back to top |
Posted: 05/30/08 3:34 pm ::: |
Reply |
|
TDAO,
No idea. I'll go ahead and recheck the spreadsheet when I rerun the numbers tonight.
--Petrel
|
|
petrel
Joined: 26 May 2008 Posts: 2001 Location: Atlanta, GA
Back to top |
Posted: 05/31/08 12:05 am ::: |
Reply |
|
For a few moments there, I thought that all was well with my Hollinger Power Rankings, until I found a fatal flaw...when the Sun beat the Liberty today, the Liberty moved up and the Sun moved down!
TDAO suggested that something was wrong in my rankings, but the formula was exactly as Hollinger entered it. I had to do some work to figure out where the problem was, but I found it -- it was in the definition of "strength of schedule" (SOS)
Someone had already pointed this out -- the way I was calculating SOS was by taking the percentage of opponent wins/opponent wins + losses. A team got credit for playing opponents with good winning percentages -- but the problem was that the good winning percentages of the opponents came from losing games to those same opponents. In short, teams were getting rewarded for losing by fattening up their opponent's win-loss.
So how do you define SOS? SOS, it seems, has a specific meaning as an acronym. I used the RPI formula for SOS:
SOS = 2/3 * opponent's win percentage + 1/3 * opponent's opponent's win percentage.
This formula seems to have done the trick...somewhat. The current Hollinger Power Rankings after tonight's games:
Fever 104.239
Liberty 102.496
Sparks 102.178
Lynx 102.077
Sky 101.877
Comets 101.711
Dream 100.648
Silver Stars 100.183
Mercury 100.159
Monarchs 100.045
Shock 99.451
Storm 98.749
Sun 98.475
Mystics 98.387
...and the strengths of schedules of each team. Remember that strength of schedule is an important part of the rankings.
Dream 0.652
Liberty 0.611
Sky 0.609
Fever 0.588
Mercury 0.583
Comets 0.576
Silver Stars 0.525
Monarchs 0.516
Mystics 0.479
Shock 0.469
Lynx 0.461
Sun 0.452
Storm 0.441
Sparks 0.421
(* * *)
Also as a bonus: the pythagorean projections of each team. Pythagorean projection gives a winning percentage based only on total points scored and total points against.
Fever 0.905
Sun 0.883
Lynx 0.853
Sparks 0.786
Shock 0.755
Silver Stars 0.690
Storm 0.596
Sky 0.462
Liberty 0.363
Mercury 0.291
Mystics 0.219
Comets 0.125
Monarchs 0.117
Dream 0.024
|
|
psheehy
Joined: 07 Feb 2006 Posts: 2972
Back to top |
Posted: 05/31/08 12:14 am ::: |
Reply |
|
math is a pain in the rear...
beware excel...sometimes you will make one tiny little typo and it cascades through five sheets of dependant formulas...it can take forever to track it down...
looks more on target now, but as TDAO points out, Hollinger's formulas are not particularly predictive early in the season. check out APBRMetrics forum for the real math/stat hardcores... a great place to get advice from the pros (hollinger and dean among others post and lurk there) on basketball number analysis and to pose new ideas.
_________________ Coach Agler's 2014 Off Season Task List
1) Think about "what if?" he had gotten that lottery ball in 2015....
2) Remember to watch some NCAA games
3) Invest in mystical restorative ointments for hips, knees, backs and well everything.
|
|
TDAO
Joined: 11 Jun 2005 Posts: 555
Back to top |
Posted: 05/31/08 2:42 am ::: |
Reply |
|
Hmm ... I don't think that's it. The link you provided makes it pretty clear that Hollinger defines strength as schedule as simply the record of opponents (which is funny given the whole premise underlying the power rankings is that records are a worthless indicator and only differential should be used, something I don't exactly agree with).
Atlanta's differential is so historically bad (-17.2) that, using the formula as I have it, their SOS would have to be 1.17 (opponents winning more than 100% of their games) to produce the above-average rating you have for them.
Here is the formula as I have it:
=(((D2-0.5)/0.037)*0.67)+100+(0.67*(C2+(((F2-E2)*3.5)/(E2+F2))))
where D2 is strength of schedule
C2 is differential
E2 is home games
and F2 is road games
How does that compare to what you have?
|
|
Shoots&Scores
Joined: 19 Mar 2007 Posts: 1544 Location: Poolside while you shovel snow!
Back to top |
|
petrel
Joined: 26 May 2008 Posts: 2001 Location: Atlanta, GA
Back to top |
Posted: 05/31/08 7:01 am ::: |
Reply |
|
psheehy: yep, I too know the pains of Excel. I pray that there isn't any problem, and I'll keep hunting down the formulas and checking down the VLOOKUPS over the season.
I just signed up for the APBRmetics messageboard. Thanks for pointing me to it!
TDAO: here's what I'm getting:
You have
=(((D2-0.5)/0.037)*0.67)+100+(0.67*(C2+(((F2-E2)*3.5)/(E2+F2))))
I have:
A + B + 100 + C + D
Where
A = ((SOS-0.5)/0.037)*0.67
B = ((SOS10 -0.5)/0.037)*0.33
C= 0.67*(Margin+((Road-Home)*3.5)/Games
D= 0.33*(Margin10 + ((Road10-Home10)*3.5)/10
So we'll take a look at my favorite team, the Dream.
SOS = 2/3 (13/17) + 1/3 (29/68) = 0.509804 + 0.142157 = 0.65196.
This is the same value of SOS10. We can talk about the true meaning of SOS later; for now we'll assume that these values are the correct ones.
Margin = -17.25, the average margin of loss for the Dream. Margin10 is the same.
Road=2 (against the Sun and Fever)
Home=2 (against the Shock and Sparks)
Road10 and Home10 are the same.
Games = 4
A = ((0.65196 - 0.5)/0.037)*0.67 = 2.75122
B = ((0.65196 - 0.5)/0.037)*0.33 = 1.35332
C= 0.67 * (-17.25 + (2-2)*3.5)/4 = -2.88398
D=0.33 * (-17.25 + (2-2)*3.5)/10 = -0.56925
2.75122 + 1.35332 + 100 - 2.88398 -0.56925 = 100.6513, which is what I have as the Dream's Hollinger rating.
There are two ways we could be getting non-matching answers:
a) We use different SOS and SOS10 ratings, or
b) IN "D", there is a factor that says "Divide by 10", when maybe it should be "divide by the mininum of current # of games played, or 10".
Looking forward to your reply on this:
--Pet
|
|
TDAO
Joined: 11 Jun 2005 Posts: 555
Back to top |
Posted: 05/31/08 1:14 pm ::: |
Reply |
|
petrel wrote: |
b) IN "D", there is a factor that says "Divide by 10", when maybe it should be "divide by the mininum of current # of games played, or 10". |
Yes, I think that's it. That means you're overvaluing SOS relative to differential.
|
|
Slovydal
Joined: 17 Nov 2004 Posts: 12205 Location: Indianapolis, IN
Back to top |
|
petrel
Joined: 26 May 2008 Posts: 2001 Location: Atlanta, GA
Back to top |
Posted: 05/31/08 2:06 pm ::: |
Reply |
|
Okay. After making the change to divide D by the minimum of either "# of games played" or 10, we get the following results:
Fever 104.697
Sparks 103.475
Lynx 103.014
Liberty 102.372
Comets 101.625
Sky 101.556
Silver Stars 100.059
Monarchs 99.969
Dream 99.795
Mercury 99.530
Shock 99.507
Storm 98.818
Sun 98.432
Mystics 98.127
|
|
TDAO
Joined: 11 Jun 2005 Posts: 555
Back to top |
Posted: 05/31/08 3:21 pm ::: |
Reply |
|
Alright, I hate to keep coming back to this, but I really think I've got it now. Somewhere along the way, you lost a set of parentheses in the SOS formula. You're dividing differential by number of games, but only the home-road discrepancy (multiplied by 3.5 points per extra home game) should be divided by games.
So you have the formula as:
C= 0.67*(Margin+((Road-Home)*3.5)/Games
but I think it should be:
C= 0.67*(Margin+(((Road-Home)*3.5)/Games))
Same with the formula for differential in the last 10 games.
|
|
petrel
Joined: 26 May 2008 Posts: 2001 Location: Atlanta, GA
Back to top |
|
Shades
Joined: 10 Jul 2006 Posts: 64255
Back to top |
|
petrel
Joined: 26 May 2008 Posts: 2001 Location: Atlanta, GA
Back to top |
Posted: 05/31/08 9:35 pm ::: |
Reply |
|
Hooray: First, the Hollinger Power Rankings for 5/31/2008:
Sparks 112.234
Fever 108.079
Lynx 107.595
Shock 105.972
Sun 105.518
Storm 103.098
Sky 101.251
Liberty 100.344
Silver Stars 97.859
Mercury 97.424
Monarchs 95.282
Comets 95.237
Mystics 93.164
Dream 86.863
Next, RPI Strength of Schedule
Dream 0.652
Mercury 0.636
Fever 0.618
Sky 0.614
Liberty 0.605
Comets 0.583
Monarchs 0.514
Silver Stars 0.513
Mystics 0.510
Shock 0.505
Sun 0.441
Lynx 0.439
Storm 0.433
Sparks 0.410
Finally, Pythagorean Projections for each team based on their current points scored and points against:
Sun 0.883
Lynx 0.863
Sparks 0.835
Fever 0.795
Shock 0.782
Storm 0.781
Sky 0.462
Silver Stars 0.363
Liberty 0.363
Mercury 0.234
Monarchs 0.211
Mystics 0.173
Comets 0.132
Dream 0.024
|
|
Richyyy
Joined: 17 Nov 2005 Posts: 24484 Location: London
Back to top |
Posted: 05/31/08 9:46 pm ::: |
Reply |
|
Heh. Pythagoras currently expects the Dream to win 0.8 of a game over the course of the season. Ouch.
|
|
OldSchoolBB
Joined: 05 Jun 2007 Posts: 207 Location: Michigan
Back to top |
Posted: 05/31/08 10:00 pm ::: |
Reply |
|
petrel wrote: |
Hooray: First, the Hollinger Power Rankings for 5/31/2008:
Sparks 112.234
Fever 108.079
Lynx 107.595
Shock 105.972
Sun 105.518
Storm 103.098
|
So how are the Shock (5-1) ranked below the Fever (3-2) when the Shock have a better record and have beaten the Fever twice?
|
|
|
|