2012 NCAA Tournament

I've gone through the process I described above for the #1 seed "pool." For that pool, I use the teams ranked #1 through #6 in the ARPI.

For the "record against teams already selected for the bracket" criterion, I considered results against all teams in the bracket. For that criterion, I ranked the teams as follows:

1 Stanford
2 Florida State
3 BYU
4 Penn State
5 UCLA
6 San Diego State

After my first run through using all the criteria, I had the teams as follows:

Stanford ahead of 5 teams
Florida State 1.5
BYU 1.5
UCLA 1
San Diego State 1
Penn State 0

I then selected Stanford, Florida State, and BYU for #1 seeds and limited the system to comparing the remaining three teams. After my second run, I had those three teams as follows:

UCLA 1
Penn State 0
San Diego State 0

The question is, why did the Committee give Penn State the fourth #1 seed? It could have been simply based on the "record against teams already selected for the bracket" criterion. Based on what I observed last year in the Committee's giving Tennessee a #4 seed though ranked #25, however, I have another possible reason. One of the things the Committee receives, as part of the Team Sheets, is each team's conference's average ARPI and the rank of that ARPI among conferences. My guess is that, when it comes to seeding, the Committee takes the conferences' ratings and ranks into consideration.

Here are the average ARPI ratings for the top 8 conferences this year:

1. ACC .6047
2. Pac 12 .5863
3. Big Ten .5772
4. SEC .5726
5. Big 12 .5699
6. West Coast .5697
7. Big East .5490
8. Mountain West .5337

Combined with Penn State's record against teams already selected, the Committee may have felt it was more appropriate to give a #1 seed to Penn State, as a representative of the third-ranked Big Ten, than to give two of the #1 seeds to the second-ranked Pac 12 or a seed to the eighth-ranked Mountain West. Also, if you look closely at the average ratings of the SEC, Big 12, and West Coast conferences, they are very close so BYU could be seen as a representative of that group.

I haven't done the remaining seed "pods" yet, and won't until later, so I can't say why those seeds worked out as they did. Under my system, with San Diego State and UCLA not getting #1 seeds they would go into the #2 seed pool and they may fare differently there than they did in the #1 pool due to matchups, common opponents, and so on.

I think the above rationale, however, could explain the #1 seeds.

 

Thanks for the link.

Part of what we're seeing is how strong the western region is in relation to the rest of the USA.

 

In response to discussion on a couple of other forums about what happens to the Sunday game if BYU reaches the College Cup, the following is from the Pre-Championship Manual:

"(Note: If an institution that has a policy against competing on Sunday or another religious day advances to the semifinals, pending the outcome of the semifinal games, the championship game may be contested on Saturday, December 1. The second and third rounds may also be contested on Thursday and Saturday if institutional policies prohibit Sunday plan.)"

 

The NCAA's "SelectionTeam Sheets" and "Selection Nitty Gritty" reports both are now available in the NCAA RPI Archive here: https://rpiarchive.ncaa.org/default.aspx. These are the data and RPI resource reports the Women's Soccer Committee used in the NCAA Tournament at large selection and seeding process. When you get to the linked page, set the Document Query table to Division I, Soccer, Women, Selection, All Reports, 2012 and you will see links to the two reports.

This is the first time the NCAA has published these reports for women's soccer in advance of the Tournament. In prior years, the reports generally were not available until mid-December. This is part of the NCAA's making more information available during the course of the season and is a very welcome addition to that information.

Thank you NCAA!

 

In a previous post, I discussed the #1 seeds and indicated that if I use the RPI only to define the pool of potential candidates (top 6 RPI teams), then apply the system I use for at large selections (which I discussed in detail in an even earlier post) to the remaining criteria and games data, and treat "results against teams already selected" as the tiebreaker if needed, then my system produces the same #1 seeds that the Committee selected.

I now have gone through the same process to analyze the #2 seeds. Essentially, the system looks at head to head results, results against common opponents, and results against teams in the bracket, assigning each one an equal weight. In case of the need for a tiebreaker, I'd use results against teams in the bracket. (There are lots of details about how I apply each of these criteria to the data.) Based on the past five years' seeding, I treat the candidate pool as the top 13 teams less those that already have received #1 seeds, which gives a group of 9 teams.

I ranked the teams under the "results against teams in the bracket" as follows:

1 Virginia
2 North Carolina
3 Florida
4 San Diego State
5 Santa Clara

(This one was tough -- their best win was only against #27 California. They had two home ties, against San Diego State and Portland. Although these are of some value, my system considers home ties as equal to neutral or away losses due to the effects of home field advantage, so in my comparison of teams' results, these both don't count for much. They also had an away tie, however, against BYU, which counts for a lot. Others could rate them higher.)​

6 UCLA
7 Ohio State
8 Baylor
9 Marquette

Using these rankings, and running them through my system together with head to head results and results against common opponents, I came out with the following rankings of the 9 competing teams:

Virginia came out ahead of 6 teams
San Diego State 5
Florida 5
North Carolina 5
UCLA 4
Santa Clara 3
Baylor 2
Ohio State 1
Marquette 1

Thus the system gives the four #2 seeds to Virginia, San Diego State, Florida, and North Carolina. This matches what the Committee did.

Next, I get to do the #3 seeds.

 

On to the #3 seeds, using the system I've already described. For records against other teams in the bracket, I ranked the teams as follows:

1 Wake Forest'
2 Duke
3 UCLA
4 Ohio State
5 Baylor
6 Tennessee
7 Santa Clara

Anyone who compares will see that my ranking Santa Clara below UCLA, Ohio State, and Baylor is a change from what I did for the #2 seeds. I did this after further thinking, believing that the Committee did not assign much value to ties. I'm not sure I agree with this, but it is an approach that is consistent with the Committee's decisions. I think Santa Clara's away tie against BYU should count for quite a bit, but having started down the road of discounting ties, I applied the discount there too.​

8 Marquette
9 Texas A&M
10 UCF

With these rankings for the records against teams in the bracket criterion, and adding in head to head results and results against common opponents, I came up with the following ranking of the teams:

UCLA ahead of 8 teams
Duke 7
Wake Forest 7
Baylor 5
Tennessee 5
Ohio State 4
Santa Clara 3
Marquette 2
Texas A&M 2
UCF 1

This gave #3 seeds to UCLA, Duke, and Wake Forest, leaving a tie between Baylor and Tennessee. The tiebreaker is record against teams in the bracket, and Baylor ranked ahead of Tennessee on this criterion, so the system gave the last #3 seed to Baylor. These #3 seeds match the Committee's seeds.
Part of what this shows is the vital importance of seed candidates' records against teams selected for the bracket. That is not the only factor that counts; head to head results and results against common opponents count too. Those other factors, however, don't always have data to allow a comparison of one team to each other team, so this tends to increase the effective weight of the records against teams selected for the bracket factor.

Now, on to the #4 seeds.