The Math Behind the Power Rankings

I thought it might be interesting to see the math that goes into the power rankings.  As the season goes along the roster score loses value and the actual win total gains value.  The expected win percentage as found in your advanced standings tab stays the same.  I ran rankings again this morning and they came out as below.  Currently the roster is 37.5%, the wins total is 42.5%, and the expected winning percentage is 20%.  By this point in the season the percentages will change by about 2.5% a week (i.e. when I run them again on Saturday it will be 35%/45%). Anyway, I hope you find what is below interesting.  A couple of fun ones to look at are Oakland and OKC.  In the case of Oakland, the roster ranking predicts their run differential quite well, but somehow they are killing it record wise.  In the OKC case, their two most recent trades have radically reduced their roster score, but by then they had already banked wins so their overall score is a bit deceiving. 

Team	Score	roster	wins	Expected
Jacksonville	2.625	2	3	3
Columbus	2.925	5	2	1
Honolulu	4.525	6	3	5
Pawtucket	4.725	1	6	9
Oakland	7.9	13	1	13
San Francisco	8.725	12	9	2
Pittsburgh	9.2	14	6	7
Colorado	9.675	8	11	10
Tacoma	12	7	15	15
Milwaukee	12.15	19	9	6
Richmond	12.2	21	5	11
Boston	12.25	9	11	21
Montreal	12.375	3	18	18
Los Angeles	12.4	22	6	8
Las Vegas	13.625	24	9	4
Salem	13.875	16	11	16
San Juan	15.075	4	23	19
Chicago	16.275	20	15	12
Baltimore	16.825	17	18	14
Kansas City	17.825	26	11	17
Hartford	18.4	18	18	20
Texas	19.975	10	25	28
Florida	20.6	11	27	25
Philadelphia	21.85	29	15	23
Scottsdale	23.5	15	27	32
Oklahoma City	24.075	31	18	24
Wichita	24.85	32	18	26
Montgomery	25.95	28	26	22
Cincinnati	26.825	30	23	29
New Orleans	27.05	25	27	31
Toledo	27.6	23	31	29
New York	28.275	27	30	27