The Dance Card

As the season winds down, discussions of the NCAA tourney bubble will surely heat up. Talking heads will guarantee that 75 different teams are “in�. Terms like quality wins, bad losses, finishing strong, and stumbling down the stretch will be daily discussion points on the internet and in the main-stream media.

Past experience has led me to conclude that a good percentage of what I read or hear is not worth the time it consumes. But when I was searching for articles on the NCAA Selection Committee, I stumbled upon something far different than the normal silly chatter or bracketology.

Two professors and self-professed basketball addicts performed a parametric study to develop a formula that predicts which teams will receive at-large bids to the NCAA tourney. They called their formula “The Dance Card� and appear to be updating their ranking on a weekly basis. Here is a short overview:

Accuracy of the Dance Card
The Dance Card has never missed on more than three spots in any season. Over the entire 12-year period, the Dance Card correctly predicted 385 of the 410 available at-large Tournament slots (or 93.9%)…

The Dance Card can only be as accurate as the Selection Committees are consistent; it is an estimate of the Selection Committees’ (not the authors’) decision criteria. The high level of accuracy and consistency of the model is strong evidence that the Selection Committees (which differ in composition each year) are actually quite consistent from year to year.

Description of the Dance Card
…The Dance Card formula suggests that only six pieces of information about each team are highly important in determining whether it gets an at-large Tournament bid:

– RPI (Ratings Percentage Index) Rank
– Conference RPI Rank
– Number of wins against teams ranked from 1-25 in RPI
– Difference in number of wins and losses in the conference
– Difference in number of wins and losses against teams ranked 26-50 in RPI
– Difference in number of wins and losses against teams ranked 51-100 in RPI

-Interesting Side Note: The Dance Card did not predict NC State’s selection in 2003 or 2005 (using the new RPI formula). Using the old RPI formula, the Dance Card would have predicted NC State to receive an at-large bid in 2005. The authors have decided to use the old RPI formula for this year’s Dance Card since it appears that the Selection Committee did not really use the new formula last year.

If we assume that the Dance Card authors’ have discovered the key parameters used by the Selection Committee, then there are several items worthy of discussion.

STRENGTH OF SCHEDULE
I find it interesting that the conference RPI rank shows a correlation and the individual team’s strength of schedule apparently does not. When you consider that the larger conferences (conference size not just strength) play a un-balanced schedule, it would appear that the Selection Committee would be giving some teams more credit and others less credit than they deserve.

Before we get into a large rant on this point, it is important to remember that the overall schedule strength is inherently included in the RPI calculation. But then again, the conference strength is inherently included as well. I just found it strange that conference strength by itself turned out to be an important factor.

RECORD OVER THE LAST 10 GAMES
I am almost sure that I have read articles where the Selection Committee specifically said that this was a consideration. However, the Dance Card Authors apparently couldn’t find any evidence to support this claim. To double-check the Dance Card formula, I scanned down through the NCAA Bubble Teams (RPI #36 through last team in) to see what I could find.

Finishing Strong
To supplement the Dance Card’s conclusions, if I could find “major� schools with strong finishes, “reasonable� RPI rankings, and in the NIT….then it would pretty well prove that finishing strong is not necessarily enough to impress the Selection Committee. Here are some examples of teams that fell into the NIT:

					W/L over
Year	RPI	SOS	Team	Conf	Last 10
2004	49	26	Notre Dame	BE	7-3
2003	49	44	BC	BE	7-3
2000	42	32	Villanova	BE	6-4
1999	40	23	Oregon	Pac 10	6-4

It looks like we might as well limit discussions of strong finishes to those occasions when certain State fans begin to dream up new definitions for standard terms.

Stumbling Down The Stretch
While there is considerable reason to doubt that strong finishes impress the Selection Committee, poor finishes are another matter. When looking over the Dance Card analysis and at-large bids from 2001-2005, there are two things worthy of note:

– Only four Bubble teams (ie RPI 36+) with losing records over the last 10 games received an at-large bid to the NCAA tourney.
– Three of the teams “selectedâ€? by the Dance Card formula, but rejected by the Selection Committee, had losing records over the last 10 games.

A losing record over the last 10 games does not necessarily mean that the bubble will burst, but the odds are not good. My conclusion is that it is easier to lose your way into the NIT over the last three weeks of the season than it is to win an at-large bid.

CONFERENCE TOURNAMENT RESULTS

Let’s look at some of the Dance Card misses (predicted in, but fell to NIT) from BCS conferences and NC State’s Bubble Years of 2003 and 2005.

				Conf	Conf	W/L vs	W/L
Year	RPI	Team	SOS	W/L	Rank	Top 50	Last 10
2005	65*	NC State	42	7-9	1	4-8	6-4
2004	49	Notre Dame	26	9-7	3	4-8	7-3
2003	53	NC State	45	9-7	3	2-8	5-5
2002	43	Villanova	14	7-9	6	3-10	5-5

* New RPI formula

It would be easy to say (and rightly so) that you can’t compare bubble teams from different seasons. However, the BE teams were skipped over for a number of teams with lower RPI rankings…just like State was picked over a good number of teams with higher RPI rankings. What I am looking for is something to explain what the Selection Committee saw that is not included in the Dance Card analysis. I think that it is pretty clear that the columns in the above table do not explain State’s acceptance and the rejection of the BE teams.

The only thing that stands out to me that would elevate State over other teams is the performance in the ACCT. Here are the conference tournament performances for these four cases:

Year	Team	Conference
		Tournament Games
2005	NC State	W (135) Florida St.
		W (7) Wake Forest
		L (4) Duke
2004	Notre Dame	W (97) West Virginia
		L (5) Connecticut
2003	NC State	W (76) Georgia Tech
		W (7) Wake Forest
		L (12) Duke
2002	Villanova	W (66) Syracuse
		L (8) Connecticut

There may well be other significant differences that I didn’t find. However, right now it looks like the Selection Committee was really impressed with State’s ACCT wins….and not impressed with close losses to top-10 teams.

THE MORE THINGS CHANGE…
There was an article last spring about the Selection Committee always having one final meeting after Selection Sunday. The purpose of the meeting was to discuss the overall process of selecting and seeding the field of 65. However, the consistency and reliability of the Dance Card suggests that the Committee is not actually changing very much about the Selection Process at all. I’m just not sure if the Selection Committee being so predictable is “good� or “bad�.

FINAL THOUGHTS
Whether you are on the Selection Committee or a message board community, it is a difficult job to sort through the various bubble teams and pick who “deserves� to be in the NCAA tournament. These bubble teams all have some high points to brag about and some warts that they would like to hide. I find it amazing that anyone could go through all of the available data and come up with a system that can sort through the “rough� and come up with some real “diamonds�.

The relative simplicity and outstanding accuracy of the Dance Card’s predictions just amaze me. Their work suggests that a number of major talking points on NC State (and presumably other team’s) message boards have been misguided since the actual data doesn’t support most of the discussion points I have seen in the past.

I suppose it is too much to hope for to expect talking heads and bracket creators to actually attempt to provide some basis for their predictions. My thanks to the Dance Card authors for bucking the trend and providing some meaningful analysis to all of the bubble talk.