Updated January 2017
On the "NCAA Tournament: Predicting the Bracket, At Large Selections" page, I described how I have used the NCAA's criteria for at large selections to the NCAA Tournament, coupled with data related to those criteria, to identify a series of "patterns" that are consistent with all of the Women's Soccer Committee's at large selections over the ten years from 2007 through 2016 and all of the Committee's "non-selections" over that period. Some of the patterns involve a single criterion and the data related to it and some involve paired criteria and the data related to them. I then identified a series of "yes" the team gets an at large selection and "no" the team is denied an at large selection patterns. One can use the patterns, once the regular season is complete, to predict which teams will and won't get at large selections and which teams might be in play for any remaining at large spots; and as a basis for seeing whether the Committee's current decisions are consistent with the Committee's patterns over the last ten years.
I have gone through a similar process with seeds, within each seed "pod": the four #1 seeds, the four #2 seeds, the four #3 seeds, and the four #4 seeds. Although for seeding, the Committee is not bound to follow the criteria that apply to at large selections, the Committee's seeding decisions over the last ten years all have been consistent with the patterns I've identified.
Here are the patterns for #1 seeds. For an explanation of the columns in the table, there's an explanation on the "NCAA Tournament: Predicting the Bracket, At Large Selections" page. These patterns identify 92.5% of the #1 seeds over the last 10 years. For the years in which the patterns don't identify all four #1 seeds, they identify 3. In those years, for the fourth #1 seed, the patterns identify 2 candidate teams from which to choose the fourth #1 seed.
Here are the patterns for #2 seeds. These patterns identify 82.5% of the #2 seeds over the last 10 years. For the years in which the patterns don't identify all four #2 seeds, they identify either 3 or 2 of the seeds. In those years, the patterns identify 1 more team than the number needed to fill out the #2 seeds -- thus where one additional #2 seed is needed, the patterns identify 2 candidates with no "yes" and no "no" patterns and where two are needed, the patterns identify 3 candidates.
#3 AND #4 SEEDS
Here are the patterns for #3 and #4 seeds. The Committee's patterns are much less clear here, which is not surprising as the correct seeding decisions are much less clear. For each seed group, the patterns identify 57.5% of the seeds over the last 10 years. In the years in which the patterns don't identify all 4 of the seeds, which is all but one year, the patterns identify anywhere from 3 to 9 candidate teams to fill the remaining seed positions. Much of the Committee's difficulty with the #3 and #4 seeds appears to be in deciding which should be #3s and which #4s, rather than in deciding which teams should receive any seeds at all. This is reflected by the fact that the patterns, on average and without regard to the specific seed assigned, identify 83.8% of all seeds over the 10 year period. For the #3 and #4 seeds, however, the patterns leave a lot of flexibility as to where the Committee places teams.