I promised a review of the 2006 picks, so here's a partial one. Last year, the Crashing the Dance oracle correctly selected 29 of 34 at large teams. There was a lot of teeth gnashing about some of the selections last year (it seemed like more than the usual teeth gnashing, at least anecdotally), so I guess I don't feel so bad.

However, it is a great opportunity to review what we're doing here and to see how we can improve. I won't be able to cover all hits and misses here, but I'll give the data points and one example of how wrong the committee was.

First, the teams that CTD let in but the committee left out:

Missouri State (I had them as a 6 seed)

An obvious mistake by the committee. Compare their resume to (for example) that of Arizona, who was selected at-large (and placed in the #8 seed line?):

	MSU	U of A
W-L	20-8 (71%)	19-12 (61%)
RPI (Total/Conf/NonConf)	21st/6th/50th	24th/9th/52nd
Schedule strength	46th	7th
Road +/-	3	-2
Last 10	8-2	6-4
W vs. 1-25 RPI	0	1
+/- vs. 1-50 RPI	-4	-5
+/- vs. 1-100 RPI	-3	-1
L vs. 101+ RPI	0	3

(Note, my RPI numbers may not exactly match other sites due to home/road/neutral classification differences.)

The lesson seems to be that it is more important to have a very difficult schedule than to actually win against a moderately difficult one. The RPI (as imperfect as it is) is set up to balance playing a good schedule and winning games.

Arizona's one top 25 RPI win was against Kansas in their first game of the season, and they didn't have a top 50 win after January 1.

The others are (sorry, no time for analysis in this post):

• Cincinnati (8 seed)
• Michigan (9 seed)
• Creighton (10 seed)
• Hofstra (12 seed)

The teams that the committee selected but CTD left out were:

• California
• Alabama
• Seton Hall
• Utah State
• Air Force (this one was roundly criticized - I'll take a closer look later)

Last but not least, these were the big seed misses:

• UNCW (CTD had 12; actually was 9)
• NC State (had 7; was 10)
• Bradley (had 8; was 13)
• Indiana (had 9; was 6)
• Wichita State (had 10; was 7)

Here are the nitty gritty details for the teams in question for you to draw your own conclusions. I'll try to add more analysis as I have time. One quick note is that the five teams selected by the committee but not by CTD each had 3 losses against RPI 101+ teams, while the four of the five CTD did select had 1 or fewer such losses (the other had 2). Perhaps the committee doesn't put as much weight on bad losses as people think. The look at Missouri State vs. Arizona also supports this.

Remember that the poll is the coaches poll (I didn't estimate the top 25 last year, so it was as of the previous Monday) and that NC here stands for non-conference, not neutral court.

I let in, they left out			Rank			Conf		NC	Road	L10	1-25	26-50	51-100	101+
School	Conf.	W-L	Sagarin	Poll	RPI	RPI	+/-	RPI	+/-	W	W	+/-	+/-	L
Missou. St	MVC	20-8	25		21	50	5	6	3	8	0	-2	1	0
Hofstra	CAA	24-6	64	37	30	29	11	61	4	8	0	1	1	1
Creighton	MVC	19-9	46		39	45	5	24	-2	6	2	0	1	2
Mich.	B10	18-10	38		47	66	-1	38	-1	3	3	-2	5	1
Cincinnati	BE	18-12	30		40	54	-1	29	-4	5	2	0	4	1
I left out, they let in			Rank			Conf		NC	Road	L10	1-25	26-50	51-100	101+
School	Conf.	W-L	Sagarin	Poll	RPI	RPI	+/-	RPI	+/-	W	W	+/-	+/-	L
Cal	P10	20-10	52		57	40	7	125	1	6	2	0	2	3
Alabama	SEC	17-12	57	39	56	35	3	103	-4	6	3	-1	1	3
Seton Hall	BE	18-11	58		58	79	1	37	-1	5	2	2	-1	3
Utah St	WAC	22-8	69		46	71	7	16	1	7	1	0	3	3
Air Force	MWC	22-6	54	31	51	51	7	56	1	7	0	-1	3	3
Big seed misses			Rank			Conf		NC	Road	L10	1-25	26-50	51-100	101+
School	Conf.	W-L	Sagarin	Poll	RPI	RPI	+/-	RPI	+/-	W	W	+/-	+/-	L
UNCW	CAA	25-7	48	35	27	19	15	94	2	9	0	1	3	2
N.C. St	ACC	21-9	23	22	49	37	3	73	1	5	1	2	5	1
Bradley	MVC	20-10	37		34	39	5	41	-3	7	2	2	1	3
Indiana	B10	18-11	24	43	35	28	2	77	-2	5	4	3	0	2
Wich. St	MVC	23-8	36	35	23	25	10	55	4	7	2	-2	2	0

The fight for #1 seeds is getting interesting. The seed points (Pts column; see the S-Curve list for full list) show very little difference between the top six teams (through games of Feb 5).

			Rank			Conf		Non	Road	L10	1-25	26-50	51-100	101+
School	W-L	Pts	Sagarin	Poll	RPI	RPI	+/-	RPI	+/-	W	W	+/-	+/-	L
UCLA	19-2	15.99	2	2	1	10	7	1	1	8	5	1	5	0
Wisc.	22-2	15.95	5	4	4	5	7	11	4	9	4	1	5	0
Pitt	20-3	15.64	7	6	2	9	7	5	4	9	2	1	7	0
Fla.	21-2	15.63	3	1	13	2	8	44	2	10	2	3	6	0
Ohio St	20-3	15.62	4	3	5	1	7	24	1	9	2	4	1	0
UNC	20-3	15.56	1	5	3	13	4	4	2	8	6	0	3	1

The seed points is just that - a point value given based on the seed a team has (or is predicted to have), where the better the seed, the more points are given. A #1 seed is worth 16 points, a #2 seed is worth 15 points, and so on. (If we need to assign seed points to a team not selected for the field, they get a value of 0. That is why we chose a descending point value; 16 points (#1 seed) is better than 8 points (#9 seed), which is better than 0 points (watching on TV).)

We assign each team selected for the predicted field of 65 a seed point value by comparing their profiles to teams from previous years. The seed point value comes from combining (sort of like an average) value from the most similar teams. We then rank the teams by seed points to determine the overall S-Curve.

For example, UCLA has 15.99 seed points, meaning that they have a profile that essentially matches #1 seeds from previous years. (Remember 16 points is the maximum.) Six teams have at least 15.56 points, meaning they all have profiles that, as of today, are most like that of #1 seeds from the past.

The teams are bunched together, but they do have different profiles, which makes it difficult to differentiate.

School	Pro	Con
UCLA	Non-conf schedule (RPI #1) 5 W vs. RPI Top 25	Conf RPI #10 Road record (3-2)
Wisconsin	Road record (5-1) 4 W vs. RPI Top 25	Non-conf RPI #11
Pittsburgh	Road record (6-2) Non-conf schedule (RPI #5)	2 W vs. RPI Top 25
Florida	10-0 Last 10 #1 in polls	Non-conf schedule (RPI #44) 2 W vs. RPI Top 25
Ohio State	Conference record (8-1, RPI #1)	Non-conf schedule (RPI #24) 2 W vs. RPI Top 25
North Carolina	Non-conf schedule (RPI #4) 5 W vs. RPI Top 25	Bad losses (Gonzaga RPI #53, NC State RPI #103) Conference record (RPI #13, lowest of the 6 teams)

Of course, only four teams will end up as #1 seeds, so the smallest thing can have a large impact. Before last weekend, North Carolina was at the top of the S-Curve with 15.97 seed points. After losing to NC State, they dropped to sixth with 15.56. In most years, 15.56 points would easily be a #1 seed. If the Heels beat Duke Wednesday, they may well jump back up to a #1; only .08 seed points separate teams 3 through 6.

And don't forget, if Texas A&M continues their strong performance, they could join the big boys at the top. This one could go down to Selection Sunday.