2000-01 NCAA Seeding Correlation

2000-01 NCAA Tournament Seedings & Results
(data prior to NCAA Tournament)

The table below shows the predicted rank of the top 44 teams (11 seeds) which were selected to the NCAA Tournament. The purpose is to compare various polls and ratings systems to see which method is most accurate in predicting the seeds the NCAA committee determined. Each rating system (RPI, Sagarin, & Massey Rating Systems along with the Associated Press sportswriters and ESPN/USA Today Coaches poll) was taken and the rank of each team was assigned to the teams which made the tournament. This data was then regressed against the average ranking a team of that particular seed would be expected to earn.

So for example, a number one seed would be expected to be one of the top four teams in the tournament. They thus would have an average ranking of (1 + 2 + 3 + 4)/4 = 10/4 = 2.5. This approximation is used because it is generally unknown how the NCAA committee ranked teams within each seed.

The data was regressed and a measure of the error between the predicted and actual rank was determined by using the r-squared function (rsq). For those unfamiliar with this function, rsq can vary between 0 and 1.0, with the higher value indicating a better fit. Because each team is being regressed against an average ranking (and thus there are natural errors introduced), the very best regression that can be achieved is not an r-squared of 1.0 (which corresponds to zero error) but a lower value, shown at the bottom of the table. The r-square of each model was normalized to that reduced value.

The plot below the table shows the difference between the predicted and actual seedings. A perfect fit would be a straight line through the center of the graph from the bottom left to the top right.

NCAA Seed	NCAA Seed #	Team		RPI #	AP #	ESPN/ USA #	Massey Ranking	Sagarin Ranking	Result
1	2.5	Stanford		2	2	2	2	2	2.00
1	2.5	Duke		1	1	1	1	1	1.00
1	2.5	Illinois		6	4	6	5	5	5.20
1	2.5	Michigan State		3	3	3	3	3	3.00
2	6.5	North Carolina		4	6	5	4	6	5.00
2	6.5	Arizona		8	5	4	6	4	5.40
2	6.5	Iowa State		13	10	9	13	13	11.60
2	6.5	Kentucky		10	9	10	8	8	9.00
3	10.5	Boston College		5	7	7	7	11	7.40
3	10.5	Maryland		22	11	11	14	10	13.60
3	10.5	Mississippi		7	14	13	12	15	12.20
3	10.5	Florida		17	8	8	9	7	9.80
4	14.5	UCLA		11	15	18	15	17	15.20
4	14.5	Indiana		16	20	21	19	18	18.80
4	14.5	Oklahoma		15	13	14	10	16	13.60
4	14.5	Kansas		12	12	12	11	9	11.20
5	18.5	Ohio State		34	26	26	25	26	27.40
5	18.5	Cincinnati		31	27	31	30	28	29.40
5	18.5	Syracuse		18	17	16	34	42	25.40
5	18.5	Virginia		33	16	15	17	14	19.00
6	22.5	Southern California		21	35	36	24	22	27.60
6	22.5	Wisconsin		19	25	24	23	21	22.40
6	22.5	Notre Dame		25	19	19	50	40	30.60
6	22.5	Texas		9	18	17	16	20	16.00
7	26.5	Iowa		26	24	25	40	29	28.80
7	26.5	Arkansas		41	28	54	29	27	35.80
7	26.5	Penn State		24	41	42	35	38	36.00
7	26.5	Wake Forest		30	23	22	20	12	21.40
8	30.5	Georgia		27	56	39	49	41	42.40
8	30.5	Georgia Tech		39	39	35	38	33	36.80
8	30.5	California		35	50	54	28	23	38.00
8	30.5	Tennessee		14	31	30	21	19	23.00
9	34.5	Missouri		47	56	54	54	49	52.00
9	34.5	St. Josephs		36	22	23	18	24	24.60
9	34.5	Fresno State		20	30	27	26	25	25.60
9	34.5	Charlotte		48	37	37	36	35	38.60
10	38.5	Creighton		23	36	35	27	34	31.00
10	38.5	Georgetown		42	21	20	44	37	32.80
10	38.5	Butler		29	46	43	22	39	35.80
10	38.5	Providence		28	33	38	41	32	34.40
11	42.5	Oklahoma State		49	43	46	51	48	47.40
11	42.5	Georgia State		32	32	32	45	58	39.80
11	42.5	Temple		38	40	54	33	30	39.00
11	42.5	Xavier		44	46	46	31	31	39.60

rsq	0.9739	top 6 seeds		0.526	0.801	0.742	0.664	0.674	0.824
rsq	0.9922	top 11 seeds		0.662	0.718	0.701	0.633	0.682	0.808
	Ideal		Normalized	RPI #	AP #	ESPN/ USA #	Massey	Sagarin	Average
					*	*

* Note - Others Receiving Votes were included, teams without any votes were given the next open rank number. When multiple vote numbers were present, the lowest rank was given.

ChartObject Predicted vs. Actual
NCAA Seeding
(2000-01 Season)

Preliminary Conclusions

1.) In general, it appears that the polls (ie AP and ESPN/USA Today etc.) do a better job at predicting the order that the teams will be seeded than the mathematical models. (Perhaps the polls are subtly influenced by the NCAA committee, I don't know).

2.) All the mathematical models do about an equal job in terms of correlating with the at-large teams which got into the tournament. (I cut off the at-large teams at the 11 seed). Of the models, the Sagarin is the best slightly, followed by the RPI and Massey. When looking at the top 6 seeds (ie approximately the top 25 in the country), the RPI is by far the worst predictor of seedings.

3.) Of all the models, only the RPI showed a substantially positive increase between correlating with the top 11 seeds as compared to how it correlated with the top 6 seeds. The Sagarin model also made a positive change, however it didn't increase as dramatically. The reasons for this are likely two-fold. First, unlike the other models, the RPI is not designed to explicitly determine the strongest team. In fact, the RPI is really more a measure of schedule strength. So it should not really be expected to zero in on the top seeds and thus the correlation with the top seeds is relatively poor.

However, it is known that the RPI is used by the NCAA committee when looking to determine which bubble teams to invite to the tourney. The relatively better correlation with the top-11 seeds (which does include these bubble teams) suggests that having a good RPI rating does in fact help in this particular area. This is evidenced by the uptick in correlation when considering the top-11 seeds. Granted, the overall correlation of the RPI to the NCAA field is still poor, however it does seem to suggest that there is an influence on bubble teams.

4.) Probably the most significant result is that if you average the RPI rating (which was designed to be more a measure of schedule strength), the Sagarin and Massey Ratings (which are designed to be a measure of team strength) and the polls (which are a reflection of pollsters impressions of team strength), you get a far more accurate estimate of the seedings than any one method by itself. This lends support to the notion that the committee takes into account many different factors when deciding on which teams make the field. In a way it is reassuring that a seed correlates well with a conglomeration of all these different factors, rather than skewed by one particular measure.

A second aspect of these models to consider is how well they did in terms of predicting which teams would make the tournament and which ones wouldn't. Below is a table listing the highest rated teams which did not make the NCAA's field.

Cut-Off Teams		RPI #	AP #	ESPN/ USA #	Massey	Sagarin
	Southern Mississippi	53	54		43	43
	Alabama	51	34	29	47	36
	Pepperdine				52	56
	Richmond	43	53	50	53	53
	Wyoming				55
	Texas El-Paso			45	56
	UC Irvine			51	57
	Mississippi State	40			58	45
	Villanova	45			60
	Seton Hall		48
	Purdue					50
	Tulsa					51
	Auburn	57				52
	South Carolina	60				54
	Utah	50				55
	Connecticut	54				57
	Wyoming	58				59
	New Mexico	52
	Pittsburgh	56

Lowest Team Left Out		40	34	29	43	36
Number of Teams Included		All	55	53	All	All
Number of Cut-Off in top 40		1	1	1	0	1
Number of Cut-Off in top 44		2	1	1	1	2
Number of Cut-Off in top 50		4	2	2	2	4
Number of Cut-Off in top 53		7	3	4	4	7
Number of Cut-Off in top 60		12	4	4	9	12

		RPI #	AP #	ESPN/ USA #	Massey	Sagarin

Preliminary Conclusions

1.) The results show that the AP, ESPN and Sagarin all had Alabama as the highest ranked team which did not make the tournament. The RPI tabbed Mississippi State and Massey Southern Mississippi.

2.) All systems did a good job in not having many highly ranked teams NOT make the tournament field. Of the top 44 ranked teams (top 11 seeds), three systems had only 1 team not make the field with the Sagarin and RPI having only 2.

3.) The results seem to show that the mathematical models tend to have more higher ranked teams not make the field as the bar is lowered (as compared to the polls), however this is an iffy conclusion as it is around this area (ie top 50-55 teams) that the 'Also Receiving Votes' category found in the polls no longer covers the field of potential teams.

Last Updated March 14, 2001

Return to Kentucky Wildcat Basketball Page or RPI Page.

Please send comments to Jon Scott