NFL Power Rankings: Week 17
By Keith Myers
Some big news at the bottom of the rankings, as the Colts are out of the bottom spot for the first time all season. Who would have thought that Dan Orlovsky would have such a difference. Actually, this has as much to do with what’s going on in Tampa than anything, as that team has completely quit.
At the top, the 49ers remain #1, almost entorely on the strength of there Special teams and turnovers. Once the off season gets here and I have time to import the last decades worth of stats into my power rankings database, we’ll get that straightened out. The 49ers aren’t a dominant football team, and should not be able to sit atop these rankings based soling on turnovers, especially since we’ve already shown on this blog that turnovers aren’t the be-all end all stat that people think they are.
The Seahawks managed to move up 1 spot to #11 despite losing. This was because of the continues success of their running game, and also because of the huge numbers the Falcons defense gave up that hurt their overall power score. #11 seems a but high for the Seahawks at this point, but as I said above, this formula is still flawed this early stage in it’s development.
Rank |
Last
Team
Yds
ST
Pt Dif
TO
Power
1
1
San Francisco
1.0
3.4
9.6
26
85.71
2
2
Green Bay
0.8
-0.6
13.1
22
83.82
3
3
Houston
2.3
1.1
6.9
7
75.17
4
4
New England
-0.3
0.9
9.5
14
68.63
5
7
Pittsburgh
2.9
1.2
6.3
-8
66.65
6
10
Detroit
0.2
-3.5
6.1
13
65.80
7
5
New Orleans
1.1
1.8
12.0
-4
65.63
8
8
Baltimore
0.9
-2.3
6.9
1
61.86
9
6
Dallas
0.7
-0.9
2.6
6
60.64
10
9
Chicago
0.6
5.1
0.5
2
56.98
11
12
Seattle
-0.1
-0.3
0.6
7
55.19
12
11
Atlanta
-0.4
2.0
2.1
6
55.07
13
13
Cincinnati
0.1
2.6
2.0
1
54.14
14
19
Carolina
0.6
-4.6
0.3
2
53.58
15
15
Tennessee
0.0
3.2
0.4
1
52.19
16
14
Miami
1.0
1.3
1.0
-7
51.66
17
20
NY Giants
-0.2
-1.7
-1.5
5
50.58
18
18
Philadelphia
1.2
-0.4
2.9
-13
49.63
19
24
Buffalo
-0.8
1.4
-2.3
4
46.77
20
16
NY Jets
-0.7
2.3
1.1
-2
46.51
21
17
San Diego
-0.1
-2.8
1.1
-7
44.38
22
21
Oakland
0.2
-2.8
-4.1
-4
43.09
23
22
Arizona
0.3
1.1
-2.6
-12
40.41
24
26
Minnesota
-0.3
-1.9
-7.0
-3
38.49
25
25
Jacksonville
-1.6
-2.5
-6.2
3
36.44
26
27
Cleveland
-1.7
0.2
-5.7
0
35.02
27
23
Denver
-0.4
0.5
-5.1
-11
34.62
28
29
Kansas City
-1.3
-0.4
-8.6
-3
31.88
29
28
Washington
-0.7
1.4
-3.7
-15
31.65
30
30
St. Louis
-2.3
-1.0
-13.8
-3
20.96
31
32
Indianapolis
-1.8
-7.2
-12.1
-10
18.14
32
31
Tampa Bay
-2.2
0.3
-12.4
-14
15.10
All stats except turnovers are per-game stats.
Mathematical explanation:
Here’s the formula I decided to go with this week: Power = A*Yds+B*ST+C*PD+D*TO+50
A,B,C and D are coefficients that I can use to properly weight each of the 4 variables. For this week, A=5.5, B = .75, D=.8, and the others are all 1.
Yds = yards index = (YpC-YpCA)+(YpP-YpPA)
ST = special teams index = ((KR-KRA)+2(PR-PRA))/4
PD = point differential
TO = turnover differential
YpC = yards per carry, YpCA = yards per carry against
YpA = yards per pass attempt, YpPA = yards per pass attempt against
KR = kick return average, KRA = kick return against average
PR = punt return average, PRA = punt return against average
Why this particular formula? I wanted to move away from total yards because there’s too many things that go into those totals. Yards per carry and yards per pass attempt tell a much clearer story of the proficiency of an offense or defense. These stats also have a much higher correlation to wins and losses than do total yards.
The special teams index takes into account punt and kick returns. Punt returns are given twice the weight in the formula for a few reasons. First, the values are naturally smaller and secondly, initial results from my statistical work indicate that punt return yards have a larger impact in the results of the game.
I’ve put turnovers back into the formula, and they will remain there. I had taken them out in light some initial results that suggested quite strongly that turnovers had no correlation to wins and losses. I’ve been continuing to work with the turnover data and have uncovered some additional relationships between the variables that indicate that turnovers do indeed belong in this model. I’ll be posting some of those results in next day or so.
Point differential seems self explanatory, especially in light of the fact that point differential correlates very strongly to wins and losses even across an entire season.
The coefficients A through D are there because I will be tinkering with how much weight to give each of the 4 variables in this model. Weighing the natural magnitude of each variable (Yds has a maximum of 2.7 this week while TO has a maximum of 13) along with the statistical importance of each variable (Yds having the strongest correlation to wins except for PD) will constitute the bulk of the tweaking that is left to do to the formula.
The +50 at the end the formula just shifted every result so that the numbers were between 0 and 100. I will be switching to a more precise and accurate way to normalize the raw results in the next few weeks. For now, this works and does not change the order, nor the relative distance between each team.
At some point, I would like to implement a penalty component as well, but that will likely have to wait until next season.