College Hockey:

D-I Women’s Bracketology:
The 2005-06 Introductory Column

The Jan. 25 unveiling of USCHO.com’s 2005-06 Division I Women Pairwise Rankings (PWR) likely raises more questions than it answers. This column answers the most pressing questions.

What is the best available bracket projection right now?

Harvard at No. 1 Wisconsin
Minnesota at St. Lawrence
Mercyhurst at Minnesota-Duluth
Princeton at No. 2 New Hampshire

Where is Clarkson in that bracket?

Clarkson is ineligible because the program is in its second year of reclassifying from D-III to D-I. NCAA rules require a two-year waiting period before Clarkson is eligible at the higher level. This rule applies for all schools reclassifying in any sport, except for D-I basketball. Rensselaer is ineligible for the same reason.

Did you project the bracket simply by pairing the top eight teams in the PWR appropriately?

No, it just happens to work out that way right now. The process is not that simple. You have to account for a few minor differences between the current PWR and the NCAA criteria, and you have to consider the NCAA’s priorities in pairing teams.

So what are these differences between the current PWR and the NCAA criteria?

First, note the five NCAA selection criteria–
Rating Percentage Index (RPI) [won-lost record (35 percent), opponents' winning percentage (50 percent) and opponents' opponents' winning percentage (15 percent)];
Head-to-head competition;
Results versus common opponents;
Results during the last 16 games; and
Results against teams at .500 or above in the RPI.

There are a few caveats you need to keep in mind when perusing the PWR, which attempt to model the NCAA’s criteria.

Automatic Bids

The ECACHL, Hockey East, and WCHA champions get automatic bids for the championship, and there are five at-large selections determined by the NCAA criteria mimicked by the PWR. The autobid winners will not be known until March, so for now the best bracket projection just takes the top eight teams. This makes some teams’ selection look more secure than it really is. In the current bracket, for example, Harvard and Princeton would be bumped out if Providence and Brown were to win autobids.

Note that if Clarkson wins the ECACHL, the league will not receive an automatic bid.

Last 16 Games

Unlike men’s hockey, women’s hockey still uses record in the last 16 games as an NCAA selection criterion. The current PWR are always calculated using the last 16 games available, rather than what will likely be the last 16 games at the end of the season. Many teams have only just begun to play those last 16 games.

In the above bracket projection, I looked at which comparisons depended on games that were unlikely to be included in the last 16 at the end of the season. I found this did not make any difference right now, but that will not necessarily hold true later in the season.

Record vs. Teams with RPI > .500

Just note that in a volatile season like this one, the subset of teams with RPI above .500 changes every week. This matters a lot for the PWR, especially for teams in conferences that play opponents four times.

RPI adjustments

Since the 2003-04 season, the RPI used by the NCAA for the selection purposes has differed slightly from the PWR RPI. The NCAA drops up to four regular season games and all postseason games that lower a team’s RPI when included in the calculation.

To see how this works, take Wisconsin for example. Wisconsin’s current RPI of .65 can be written as the average of the RPI value for each game on the Badgers’ schedule. For instance, the RPI values for a win over UMD, the loss to New Hampshire, and a win over Vermont, respectively, are approximately .83, .52, and .48 (yes, a loss to UNH is indeed better for RPI than a win over Vermont). The RPI values for Wisconsin’s two wins over Vermont and two wins over North Dakota are less than Wisconsin’s unadjusted RPI of .65, so they get dropped from the calculation, and Wisconsin’s new RPI becomes .68.

So what effect does this have on the current RPI? Not much among the top 10. Wisconsin jumps above UMD by a considerable margin, and Harvard jumps ahead of Princeton, but neither of these changes affect the current PWR. But these adjustments could matter down the stretch.

There is a simple heuristic that usually works for figuring out whether the adjusted RPI matters. First, note if two teams are close in the RPI, less than .01 apart from each other. Second, note if the trailing team has played a couple more games to date against any of the bottom four teams Union, Vermont, Robert Morris, and Sacred Heart. In most scenarios, these are the only teams who have struggled enough to make the adjustment relevant. Applying this heuristic to the current RPI, you can tell Wisconsin will pass UMD, because Wisconsin played Vermont twice, and UMD played none of the bottom four. You can tell Harvard will pass Princeton, because Harvard played Union twice and Robert Morris once, and Princeton has yet to play the bottom four.

What are the caveats for how the NCAA pairs teams?

Now that the NCAA has a one-year precedent in producing an eight-team bracket, the tournament committee’s actions are somewhat easier to predict.

There was speculation last year that the NCAA would be willing to sacrifice bracket integrity (i.e. paring teams 1v8, 2v7, 3v6, 4v5) for the sake of minimizing travel costs. This speculation was justified, since the tournament handbook explicitly said that minimizing travel costs is a top priority aside from protecting the top two seeds, and this practice was followed in most every NCAA sport that did not have a profitable championship. However, to the benefit of women’s hockey, the NCAA preserved bracket integrity even though it meant flying all four teams to their quarterfinal matchups.

However, the NCAA will unquestionably sacrifice bracket integrity for one reason avoiding intraconference matchups in the first round. Having interconference competition is fundamental to NCAA ideals in organizing national championships. It turned out last season that the bracket with perfect integrity also avoided intraconference matchups, and that holds true for the current bracket projection for this season, although it might not hold true down the stretch. For instance, if Minnesota maintained the No. 5 spot and No. 4 St. Lawrence and No. 3 Minnesota-Duluth swapped rankings, the Gophers would likely still be paired with St. Lawrence.

One last point — that the NCAA was willing to pay to preserve perfect bracket integrity last season does not necessarily mean it will happen every season. Last season, the NCAA could have saved money by sending Providence to host sites at Dartmouth or Harvard. But ranking-wise, No. 1 Minnesota was so clearly above the rest of the field, and Providence was so clearly below the rest of the field, that it was only fair to send Providence to Minnesota. This season, there are potential scenarios for which the right decision might not be so clear-cut. If perfect bracket integrity dictated that Mercyhurst visit UNH, it would be entirely reasonable for the NCAA to send Harvard or Princeton instead to prevent air travel, provided that the Lakers were close enough to either school in the rankings. How close is close enough is still an entirely ambiguous case-by-case process, however. Such scenarios do not appear to be probable at the moment, but they are worth considering.

Which games this weekend are most important to NCAA prospects?

Both of them are at New Hampshire Mercyhurst on Friday and Providence on Sunday. This is the toughest weekend left on UNH’s schedule, so it will go a long way in determining whether the Wildcats get a top seed and home ice. For the Lakers, Friday’s game is a must-win if they have ambitions of hosting a quarterfinal game, and a win would go a long w