Developing Statistical College Basketball Rankings

In past years, I have enjoyed making a college basketball 'Top 50' during the course of a season. However, it was essentially an opinion poll, based off of final score, and perceived strength of opponent. In essence, there was no way of validating it any higher than the AP Top 25. This season, I have decided to go from a near 100% opinion-based poll, to a near 100% statistical-based poll. There are a couple of constraints that will prevent this from being as timely, accurate, and predictive as Ken Pomeroy's (http://kenpom.com/index.php) rankings: a.) I am not a computer scientist and have no idea how to develop code to make this a more automated process, and; b.) I am not a statistician, so it is above my head as to how to develop a unique methodology or statistical testing to take this to the next level. So I'll do the best I can with data that is freely available and my trusty sidekick OpenOffice Calc.

There are five key factors I use in developing the strength of a team both offensively and defensively. For offense, they are:

1. Points Per Possession, or Offensive Efficiency
2. Effective Field Goal %
3. Free Throws Attempted Per Field Goal Attempted
4. Offensive Rebounding %
5. Turnover %

Conversely, the five key factors on defense are:

1. Opponent Points Per Possession, or Defensive Efficiency
2. Opponent Effective Field Goal %
3. Opponent Free Throws Attempted Per Field Goal Attempted
4. Opponent Offensive Rebounding %
5. Opponent Turnover %

For reference, see http://kenpom.com/blog/index.php/weblog/entry/help_with_team_page for a description of these stats. A big thanks goes to Team Rankings for making all of these statistics freely available (http://www.teamrankings.com/ncb/stats/). Without that site, I couldn't embark on this endeavor without shelling out some cash. Offensive and defensive metrics were given a 75/25 adjustment based off strength of schedule. For example, say Team A has the #12 offense but #100 strength of schedule; while Team B has the #40 offense with the #4 strength of schedule. Team A would receive an 'Adjusted Offense Rank' of 34 [(0.75*12)+(0.25*100)], while Team B would Receive an 'Adjusted Offense Rank' of 31 [(0.75*40)+(0.25*4)]. A slightly worse offense with a noticeably tougher schedule, yields a better adjusted offense.

Before I post my Top 50, keep in mind the ultimate caveat, in that it is early in the season and the sample size is small. With that said, I weighed the current metric, with recruiting rankings (http://247sports.com/Season/2013-Basketball/CompositeTeamRankings), and a blend of the final 2012-13 final rankings from Ken Pomeroy (http://kenpom.com/index.php?y=2013) and Team Rankings (http://www.teamrankings.com/ncaa-basketball/ranking/overall-power-ranking-by-team?rating_date=2013-04-09). I have currently placed a heavier weight towards last years final rankings (about 48%), while recruiting received the lowest weight (about 17%). This season's metric rank receives the remaining weighting (about 35%). There is no rhyme or reason for those values other than some trial and error on my part. The lower the score in the Weighted Metric, the better. As the season progresses, those weights will be modified, ultimately until the point that this season's metric rank accounts for the entire 100%. I have developed a "Bad Loss Correction" to adjust a team's weighted rank when they lose to a team perceived to be worse than them. For now, I will keep my methodology behind that private, as I want to test it for a while to see how well it works.

To give you an idea of the direction a team is headed, I am posting this year's metric rank to the right of my current blended rank. This will give you an idea of whether a team is rated too high or low because of how the finished last year. For example, I have Kentucky rated #22, but they are playing as the #1 overall team this season, despite their loss to Michigan State. One last thing, current rankings are through games played on Sunday, 11/18/2013. Due to time constraints, I will likely update this only once per week - probably every Monday to coincide with the latest AP rankings (http://collegebasketball.ap.org/poll). Alright, here are my current rankings:

Rank	TEAM	Weighted Metric	2013-14 Metric Rank	AP Rank
1	Louisville	8.26	19	3
2	Kansas	9.23	16	2
3	Arizona	10.095	2	5
4	Duke	10.69	22	6
5	Ohio State	11.39	4	8
6	Michigan	12.984	9	15
7	Pittsburgh	14.475	6	36
8	Indiana	20.045	50	27
9	Marquette	22.78	33	25
10	UCLA	23.745	5	22
11	Michigan St	23.77	17	1
12	Gonzaga	23.865	8	13
13	Oklahoma St	24.105	12	7
14	Iowa State	24.275	7	21
15	Wisconsin	26.685	36	12
16	Florida	27.053	21	16
17	Missouri	27.375	15	38
18	Memphis	27.41	42	11
19	Minnesota	29.065	3	39
20	Syracuse	29.6	67	9
21	Oregon	29.63	30	17
22	Kentucky	31.24	1	4
23	Baylor	31.955	31	20
24	Georgetown	33.435	39	47
25	BYU	34.915	11	44
26	Colorado	36.1	18	37
27	Illinois	36.325	41	NR
28	VCU	36.525	13	10
29	Saint Louis	36.765	28	40
30	Cincinnati	36.925	38	42
31	Villanova	37.205	20	33
32	St Marys	38.305	37	NR
33	Virginia	38.43	24	28
34	New Mexico	38.94	35	19
35	Creighton	39.435	27	23
36	Oklahoma	39.55	14	NR
37	Iowa	42.795	26	26
38	NC State	43.58	69	NR
39	Butler	46.17	64	NR
40	California	46.335	52	NR
41	San Diego St	46.935	49	NR
42	Wichita St	49.765	70	14
43	Connecticut	50.83	65	18
44	Arizona St	52.105	23	32
45	Arkansas	56.345	47	NR
46	U Mass	63.13	34	43
47	S Mississippi	65.04	58	NR
48	Colorado St	65.415	87	NR
49	Alabama	65.545	83	NR
50	LSU	66.525	59	NR

[Note: The following AP ranked teams are not in my Top 50...#24 North Carolina (61st), #29 Harvard (75th), #30 Boise State (56th), #31 Belmont (78th), #34 Xavier (54th), #35 Tennessee (55th), #41 Florida State (62nd), #45 Indiana State (79th), and #46 Notre Dame (60th)]