Linear Rank 01-05	Rank 1998	Measures in FCUs 01-05	College Football Team (won-loss: opponent strength / best win / worst loss) thru 01-04	AP Rank (BCS) 01-04	ESPN Rank 01-04
1	2	1044	FLORIDA ST (12-0 783/983/0)	1 (1)	1
2	21	1014	NEBRASKA (12-1 768/927/878)	3 (3)	2
3	13	999	MICHIGAN (10-2 810/1000/801)	5 (8)	5
4	45	990	MICHIGAN ST (10-2 799/947/885)	7 (9)	7
5	29	983	VIRGINIA TECH (11-1 717/865/1044)	2 (2)	3
6	11	979	KANSAS ST (11-1 694/878/1066)	6 (6)	6
7	31	961	MARSHALL (13-0 581/844/0)	10 (12)	10
8	4	948	WISCONSIN (10-2 701/938/649)	4 (7)	4
9	37	946	ALABAMA (10-3 821/934/792)	8 (4)	8
10	15	941	PENN ST (10-3 786/932/769)	11 (11)	11
11	1	914	TENNESSEE (9-3 770/998/887)	9 (5)	9
12	38	910	MISSISSIPPI ST (10-2 698/801/887)	13	12
13	49	900	SOUTHERN MISSISSIPPI (9-3 697/854/899)	14	13
14	3	882	FLORIDA (9-4 773/862/894)	12 (10)	14
15	18	880	MIAMI FLA (9-4 760/845/854)	15	15
16	36	860	OREGON (9-3 732/821/754)	19	18
17	108	853	ILLINOIS (8-4 765/1051/708)	24	25
18	9	847	TEXAS A&M (8-4 745/896/766)	23 (14)	20
19	75	844	LOUISIANA TECH (8-3 679/998/792)
20	86	842	UTAH (9-3 715/873/740)
21	27	835	ARKANSAS (8-4 724/878/687)	17	19
22	24	833	PURDUE (7-5 782/938/832)	25
23	12	832	GEORGIA (8-4 727/864/697)	16	16
24	14	826	TEXAS (9-5 754/962/688)	21 (15)	23
25	10	825	GEORGIA TECH (8-4 745/780/801)	20	21
Home field advantage = 52 FCUs for games up through 01-04

1. For all 200+ College Football rankings
2. NCAA College Men's Basketball Measures and Rankings

Notes:

• Home field advantage is worth extra FCUs (Football Competency Units). These are added to the home team's FCUs.
• An advantage of 70 FCU's gives a team "2-to-1 on" odds of winning.
• Based on Rasch analysis of published results of Division 1AA and 1A games available at time of computation.
• Intended solely as the demonstration of a measurement method.
• Game scores were provided by a variety of Media and Internet sources, including ABC, NBC, CBS, ESPN, AP, Yahoo and the websites of the schools themselves. AP and ESPN rankings are provided by those organizations.

The Health Care Finance Administration (HCFA) has recently approved Rasch "paired comparison" methodology for the identification of misvalued procedures, see Total Physician Work. Leon L. Thurstone perceived the necessity that measures be independent of sampling and testing specifics. Measures based on paired comparisons meet this criterion. Bradley and Terry proposed this model but from a descriptive, rather than measurement, perspective. Consequently, they noticed, but did not capitalize on, the linearity of the measures produced. Georg Rasch perceived the linearity of measures produced from ordinal data by logit-linear models.

Since mathematical expression of such models obscures their immediate utility and ease of application, they are illustrated here by a simple paired comparison analysis of NCAA game outcomes. Obvious advantages include:
(a) all teams are ranked, not just the favored few
(b) all teams are located on a linear measurement scale
(c) all teams are given equal attention
(d) the data are easy to obtain, even for non-experts
(e) rare, idiosyncratic results can be indentified, and, even if left unremedied, produce minimal distortion.
(f) it is noticeable, in these data, that the conventional ranking services pay too much attention to won-loss record - and even particular wins and losses (the immediate details), and not enough to strength of schedule (the big picture).

Thurstone, L.L. The method of paired comparisons for social values. Journal of Abnormal and Social Psychology, 1927, 21, 384-400.

Bradley, R.A. and Terry, M E. Rank analysis of incomplete block designs I: The method of paired comparisons. Biometrika, 1952, 39, 324-345.

Go to Top of Page
Go to Winsteps & Facets home Page