We’ve looked at Massey’s Compiled Rankings over the last several weeks as a different way to come up with projected seeding for the NCAA tourney. The underlying assumption is that the consensus of the various polls and computer formulas will come close to the actual seedings doled out by the NCAA Selection Committee.
I have come up with several reasons to doubt that underlying assumption, but first let’s go ahead and look at this week’s compilation. Click here for a frequency graph and here is the table:
|
|
NCSU |
UNC |
BC |
|||
|
|
2/19 |
2/26 |
2/19 |
2/26 |
2/19 |
2/26 |
|
SEED |
|
|
|
|
|
|
|
1 |
|
|
|
|
|
|
|
2 |
|
|
|
|
|
|
|
3 |
53% |
|
10% |
50% |
|
|
|
4 |
23% |
16% |
27% |
38% |
20% |
9% |
|
5 |
13% |
47% |
40% |
|
33% |
22% |
|
6 |
|
16% |
13% |
|
17% |
25% |
|
7 |
|
9% |
|
|
|
16% |
|
8 |
|
|
|
|
10% |
13% |
It will be interesting to see what happens with BC’s actual seeding. A loss to UVa followed by a win in Raleigh has certainly spread their rankings around a lot. We’ll have to see if they converge any after the regular season concludes.
As I was looking over the compiled rankings, I was curious where Massey’s got their RPI numbers from. That’s when I noticed that it looks like they got the RPI rankings Sunday morning, before Saturday’s games were included in the calculations. I changed the RPI values for the table and graph, but certainly didn’t bother to check the other 31 rankings in the compilation. The obvious concern is that if there are mistakes with the input data, then any conclusions reached are compromised.
My next area of concern came when I looked at a previous SFN entry comparing four year’s worth of RPI rankings vs NCAAT seed:
|
NCAAT |
RPI |
Average |
|
Seed |
Range |
RPI |
|
1 |
1-7 |
3 |
|
2 |
4-32 |
9 |
|
3 |
1-20 |
11 |
|
4 |
9-25 |
16 |
|
5 |
13-28 |
19 |
|
6 |
12-37 |
22 |
|
7 |
17-47 |
30 |
History says that State is sitting somewhere around a 7 or 8 seed while the compiled rankings put State at around a 5 seed. Over the last four years, no team with a #34 ranking has gotten a 5 seed and only one team ranked 31+ has gotten a six seed.
The difference between historical trends (though only four years) and the compiled rankings made me start wondering how much the NCAA Selection Committee moves teams from their expected seeding based strictly on the RPI ranking. Here’s a summary of how much the first six seeds in the NCAA tourney varied from their RPI ranking:
|
|
Seeding |
|
|
|
|
|
|
Change |
2002 |
2003 |
2004 |
2005 |
|
Better Seed |
-6 |
1 |
|
|
|
|
-5 |
0 |
|
|
|
|
|
-4 |
0 |
1 |
|
|
|
|
-3 |
1 |
0 |
0 |
|
|
|
-2 |
2 |
5 |
1 |
2 |
|
|
-1 |
2 |
2 |
8 |
8 |
|
|
|
0 |
10 |
13 |
8 |
8 |
|
Worse Seed |
1 |
7 |
3 |
6 |
4 |
|
2 |
1 |
|
0 |
2 |
|
|
3 |
|
|
1 |
|
Of the 96 teams summarized in the table, 82% were seeded within +/- 1 seed of what would be expected based on their RPI calculation. There are still a fair number of teams that were seeded higher than we would “expect�, but let me ask you a question: What has State done over the last six weeks that would lead an impartial observer to seed them higher than “normal�?
