I've been manipulating statistics for a while, and have come up with some reasonable power rankings.

Methodology:

For all of the stats that go into this, I've normalized the range. The best team for the category is given a 1, the worst a 0, and all other teams get a score between 0 and 1, scaled linearly. All stats only factor in games already played - there's no look-ahead.

The most important stat is winning percentage (WPCT). It counts for 50% of the ranking.

The second stat is the complicated one. I'll relative adjusted margin of victory (RAMOV). I developed a formula for predicting the outcome of the game. It compares the average margin of victory of the two teams in other games (so it can "predict" already played games). However, the actual margins of victory are scaled to reduce the effects of blowouts. The first 10 points of victory count full, the next 10 count half, the next half again, and so on (well, anything above 40 counts 1/16th). So a 32 point win counts 10 + 10/2 + 10/4 + 2/8 = 17.75. The home team gets a two point swing. So if team A, averaging a 3 point win, plays at home against team B,averaging a 4 point loss, the expected result is a 9 point win for team A. If, in reality, team A won by 6, it goes into the RAMOV number as -3 for team A and 3 for team B. This is averaged, but it doesn't really matter since all teams have now played the same number of games. This counts as 25% of the ranking.

Adjusted strength of schedule is the third stat (ASOS). It's the average winning percentage of opponents in games other than the game against the team. Alternatively, it's the total number of wins by opponents less the number of losses by the team divided by the number of games played by the opposition less the number of games played by the team. This counts 15%.

Strength of victory (SOV) plays a minor role. It's like strength of schedule, but only factoring in teams that the team beat. There's no adjustment, since it'd be constant factor depending on the winning percentage of the team. This counds 5%.

The last stat is points allowed per game (PAPG). Defense wins games in the playoffs, so this is important. Or maybe it's included because it pushes Denver ahead of Baltimore, which is my subjective opinion of the rankings. You decide. This counts 5%.

The normalized values for each stat are multiplied by the percentage, resulting in a ranking number. This is normalized in the same way to give the team score.

STAT MIN MAX MEAN STDEV MEDIAN
WPCT .125 1.00 .500 .222 .500
RAMOV -5.75 4.25 0 2.637 -.625
ASOS .393 .625 .500 .056 .500
SOV .188 .563 .400 .085 .403
PAPG 29.75 12.25 20.63 4.18 20.75
RANKING FACTOR  .428 .256 .401

 RANKFAC WPCT RAMOV ASOS SOV PAPG
1 IND 1.000 1.000 2.38 0.607 0.531 21.63
2 NE 0.849 0.750 4.25 0.571 0.417 14.25
3 DEN 0.792 0.750 1.50 0.589 0.479 12.25
4 CHI 0.790 0.875 4.25 0.393 0.375 12.50
5 BAL 0.786 0.750 3.00 0.536 0.458 13.88
6 NYG 0.732 0.750 1.63 0.554 0.417 18.00
7 SD 0.665 0.750 3.25 0.429 0.313 16.75
8 JAC 0.634 0.625 3.50 0.464 0.400 14.25
9 NO 0.566 0.750 -0.63 0.464 0.375 19.88
10 KC 0.560 0.625 0.75 0.500 0.475 21.13
11 DAL 0.494 0.500 3.63 0.446 0.344 20.50
12 PHI 0.465 0.500 1.63 0.482 0.375 20.00
13 SEA 0.451 0.625 -1.88 0.500 0.375 22.13
14 MIN 0.440 0.500 -0.75 0.518 0.438 16.88
15 CAR 0.433 0.500 -0.88 0.518 0.500 20.38
16 CIN 0.403 0.500 -0.75 0.500 0.406 21.00
17 ATL 0.399 0.625 -1.50 0.429 0.325 20.50
18 PIT 0.380 0.250 3.75 0.518 0.438 22.00
19 WAS 0.371 0.375 0.25 0.536 0.458 23.75
20 TB 0.351 0.250 -0.25 0.625 0.500 21.63
21 STL 0.348 0.500 -1.38 0.482 0.375 24.63
22 CLE 0.330 0.250 1.13 0.571 0.375 21.38
23 BUF 0.299 0.375 -1.38 0.518 0.375 20.38
24 NYJ 0.247 0.500 -3.50 0.464 0.281 24.13
25 DET 0.207 0.250 -1.38 0.518 0.500 25.38
26 HOU 0.184 0.250 -1.38 0.500 0.438 24.50
27 MIA 0.173 0.250 -0.63 0.411 0.563 19.75
28 GB 0.150 0.375 -2.13 0.429 0.208 25.00
29 SF 0.110 0.375 -5.75 0.500 0.417 29.75
30 TEN 0.052 0.250 -4.38 0.500 0.313 27.00
31 OAK 0.051 0.250 -3.13 0.446 0.188 20.50
32 ARI 0.000 0.125 -3.25 0.482 0.375 24.50