Power Rankings..other statistical musings.

[dwmyers]

These are my latest power rankings. If Pittsburgh wins tonight, they'll likely dispalce the Giants as the #5 team on my chart. The power rankings as displayed here are based first on winning percentage and next on point spread. It's a purely scoring based ranking.

Global Statistics:
Games  Home Wins Winning_Score Losing_Score Margin
129        72        27.33         15.26     12.06


Rank  Team    Spread  AvSco  AvAll  Wins Loss Ties 
--------------------------------------------------
1     NE      23.11   39.44  16.33   9  - 0  - 0  
2     IND     14.75   30.50  15.75   7  - 1  - 0  
3     DAL     11.25   33.12  21.88   7  - 1  - 0  
4     GB       6.50   24.25  17.75   7  - 1  - 0  
5     NYG      5.12   25.00  19.88   6  - 2  - 0  
6     TEN      5.12   20.62  15.50   6  - 2  - 0  
7     DET      1.88   25.00  23.12   6  - 2  - 0  
8     PIT     13.29   26.29  13.00   5  - 2  - 0  
9     JAC      0.50   19.38  18.88   5  - 3  - 0  
10    CLE     -0.75   28.38  29.12   5  - 3  - 0  
11    WAS     -1.00   19.00  20.00   5  - 3  - 0  
12    BAL      0.71   17.71  17.00   4  - 3  - 0  
13    TB       2.22   18.22  16.00   5  - 4  - 0  
14    SEA      3.25   20.88  17.62   4  - 4  - 0  
15    SD       3.12   23.62  20.50   4  - 4  - 0  
16    NO      -1.62   21.62  23.25   4  - 4  - 0  
17    KC      -2.75   15.50  18.25   4  - 4  - 0  
18    CAR     -3.00   17.12  20.12   4  - 4  - 0  
19    BUF     -3.25   16.25  19.50   4  - 4  - 0  
20    HOU     -2.56   22.56  25.11   4  - 5  - 0  
21    MIN      1.50   20.75  19.25   3  - 5  - 0  
22    PHI      0.12   19.50  19.38   3  - 5  - 0  
23    ARI     -2.12   19.62  21.75   3  - 5  - 0  
24    CHI     -4.62   18.00  22.62   3  - 5  - 0  
25    DEN    -12.62   15.75  28.38   3  - 5  - 0  
26    OAK     -3.12   19.00  22.12   2  - 6  - 0  
27    CIN     -5.75   24.75  30.50   2  - 6  - 0  
28    ATL     -6.75   14.38  21.12   2  - 6  - 0  
29    SF     -10.25   13.00  23.25   2  - 6  - 0  
30    NYJ     -7.67   17.67  25.33   1  - 8  - 0  
31    MIA     -9.75   20.75  30.50   0  - 8  - 0  
32    STL    -15.00   12.38  27.38   0  - 8  - 0

I have other codes that calculate things like median point spread as opposed to average point spread. Medians are supposed to be more robust than averages. Does that really get you anything? I'd say yes, since median point spreads (alone) at the end of the season accurately predicted the AFC and NFC winners in 2002 and 2004 and the winner of the Super Bowl.
Summaries of the results are:

Scoring: 
1 pt for AFC winner.
1 pt for NFC winner.
1 pt for Super Bowl winner.

Record means prediction based on record -and- average point spread.
Median means prediction based on median point spread alone.

2001

Predicted

Record) STL* versus PIT
Median) STL* versus PIT

Actual

NE* versus STL

Record 1 pt, Median 1 pt

2002

Predicted

Record) PHI* versus OAK
Median) TB*  versus OAK

Actual

TB* versus OAK

Record 2 pts, Median 4 pts

2003

Predicted

Record) NE* versus STL
Median) KC versus PHI

Actual

NE* versus CAR

Record 4 pts, Median 4pts

2004 

Predicted

Record) PIT* versus PHI
Median) NE* versus PHI

Actual

NE* versus PHI

Record 5 pts, Median 7 pts

2005

Predicted

Record) IND* versus SEA
Median) IND* versus NYG

Actual 
PIT* versus SEA

Record 6 pts, Median 7 pts


2006

Predicted

Record) SD* versus CHI
Median) BAL versus CHI (even)

Actual

IND* versus CHI

Record 7 pts, Median 8 pts.

I suspect some combination of record and median point spread, using some folk rules, will be more accurate still.


David.