Wednesday, December 2, 2015

Our Quirky Tournament

We've fielded more than one question in the last two weeks from Engineer fans about the Pairwise Rankings. The response was thusly: it's November, don't worry about the Pairwise. Win hockey games.

Here's the chart we used a few years back to describe the Pairwise Rankings across each month of the college hockey season:

October: Worthless.
November: Farcical.
December: Quirky.
January: Interesting.
February: Important.
March: Essential.

So we've moved from farcical to quirky just recently. October is worthless because the PWR is completely unstable pretty much all the time. November is farcical because there's far too much hockey left to be played, not to mention that the Ivy League - a full 10% of the teams in the PWR - have barely started to play. There's also a solid number of inter-conference games to be played, which can swing things significantly if trends change.

December's review of the Pairwise can at least show us what teams are early contenders for at-large bids, and can reveal what teams are already dependent upon the automatic bid as their route to the tournament. But at this point, tracking game-to-game changes in the Pairwise remains a fruitless and futile exercise - and breaking down tournament fields can be even more tedious. So, let's not.

January is when the PWR at least becomes interesting, because the movement of teams becomes a touch easier to predict and there's a far more clear perception of contenders vs. pretenders, but even then it's still too early to take a lot of it to heart, at least toward the beginning of the month. Later in January, the number of inter-conference games dwindles rapidly, and the PWR becomes far more stable. By February, it can be a valuable resource from game-to-game in order to discern what teams need to see happen elsewhere for them to make the tournament - and by March, of course, the decisions are about to be made.

And what decisions are those? Well, unlike in other NCAA sports where the tournament participants are often decided in those "smoke filled rooms," the PWR makes things pretty clear who is in the tournament, and who isn't. For the sake of the chart below, Atlantic Hockey is expected - as usual - to only be able to send its conference champion to the tournament, and a conference tournament upset consists of a team that would not be in the tournament without the automatic bid winning that bid.

1-10: Always in the tournament, no matter what. 1-4 are #1 seeds, 5-8 are #2 seeds, 9 and 10 are #3 seeds.
11: Almost certainly in the tournament as a #3 seed. There would have to be upset champions in every conference for this team to be out.
12: Also likely in the tournament as a #3 seed. Four upset champions put this team out.
13: The good side of the bubble. Good odds of being in the tournament, but three upset champions put them out.
14: Firmly on top of the bubble. Two upset champions put them out.
15: The bad side of the bubble. Just one upset champion would put this team out.
16: Out of the tournament, unless the Atlantic Hockey champion is in the Top 15 AND there are no upset champions in other conferences.
17-60: Always out of the tournament without an automatic bid.

It's possible that this chart may have to have the bubble shifted upwards, to 12-13-14, or even 11-12-13, if the Big Ten and the WCHA continues to have as miserable a season as they have collectively been having, as they too could find themselves only able to send a conference champion.

So anyway, that's the long-winded reason why we don't care about the PWR right now. Besides - caring about the PWR right now makes one look downright silly when February arrives and you're one of the also-rans hoping for an automatic bid to make the tournament. We warned more than a few Clarkson fans last year who couldn't help but squeeeeeee! over their PWR position in November about this. They did look silly by February.

But human nature being what it is, we've been asked about the Engineers so many times that we broke down and took a peek at where they are in this currently-quirky ranking. They're tied for 13th with Denver and sitting in 14th after losing the tiebreaker (the head-to-head comparison) with the Pioneers. They'd be on the tournament bubble if the season ended today. It doesn't, not even close. Don't get too excited.

According to PWR, the presumed field would currently consist of 6 ECAC teams (including the Engineers), 4 NCHC teams, 3 Hockey East teams, and 1 each from the Big Ten, WCHA, and Atlantic Hockey. In other words, don't hold your breath on any of this being remotely real just yet.

What is it telling us? Well, it's telling us the same thing KRACH is telling us, now that Quinnipiac has tied two games and Colorado College finally won one - the ECAC is certainly well positioned as one of the strongest leagues in the country this year. At least, they are right now. Whether that will continue to ring true depends largely on what happens in December and January.

Now, have no fear, Puckman Nation. Should Rensselaer (as we are really forced to call the school during sessions like this because of the prevalence in PWR of RPI the ranking, the Ratings Percentage Index) still be in the discussion when the discussion actually starts to become interesting after the New Year, we'll debut Engineer Bracketology at that point and continue it until such time as it becomes pointless, if it does. (The good news is that, if it does happen, it's easy to slot the Engineers - they will play in Albany as hosts, should they make the tournament.)

Until then, consider this a quick primer on exactly how the PWR operates. Things have changed since 2011 and 2013, the last times we felt the need to run Engineer Bracketology. The dreaded and annoying TUC cliff has gone, and every team regardless of their RPI is now ranked in the PWR. That means that every possible comparison between the 60 teams - that's 1,770 comparisons for you math majors - is factored into the PWR. Essentially, every team is now a Team Under Consideration.

The PWR now only consists of three factors: RPI, record against common opponents (COp), and head-to-head results (H2H). The team with the better RPI gets 1 point, the team with the better COp record gets 1 point, and both teams get a point for every win they have in H2H. It makes things much easier.

RPI is a percentage equivalent to computed by including a 25% weight to your own winning percentage, a 21% weight to your opponents' winning percentage, and a 54% weight to your opponents' opponents' winning percentage.

But there's a wrinkle to RPI. A team can receive RPI bonuses for "quality" wins against teams in the Top 20 of RPI - from .0500 for a win against the top team in RPI, decreasing in increments until you get to the #20 RPI team, worth .0025.

All games not played on neutral ice are weighted differently, too. When you win on the road or lose at home, that game is weighted by a factor of 1.2 over a game on neutral ice. When you lose on the road or win at home, that game is weighted by a factor of 0.8 over a game on neutral ice. And finally, in a situation where a win actually leads to a decrease in total RPI - which can happen when very good teams beat very bad teams - that game is not counted. A win will always leave a team at least with the same RPI they had coming in if not better.

And that's basically all for now. More to come later, if it becomes necessary. Here's where the Engineers currently stand in terms of "quality wins." It's pretty solid - only Quinnipiac and Boston University have accrued a higher quality win bonus total than Rensselaer.

Quality wins
Boston College (#6, at home)
St. Lawrence (#10, on the road)
Yale (#11, at home)

Missed opportunities
UMass-Lowell (#8, on the road)
Michigan (#12, at home)
Harvard (#2, neutral ice)

Upcoming opportunities in the next 10 games
Harvard (#2, at home) - Saturday
Miami (#20, on the road) - Jan 2/3
Cornell (#9, at home) - Jan 15
Quinnipiac (#1, on the road) - Jan 21

And finally... KRACH is, most of the season long, the best way to compare two teams. Theoretically, taking the two teams' seasons to date into consideration, comparing two teams using KRACH would suggest that the team ranked higher overall would be favored to win a theoretical matchup on neutral ice. It doesn't take into effect anything like home-ice advantage, injuries, momentum, or anything like that, intangibles that have real effects on how games play out, but it's a very decent comparison of teams against each other (now that every team is neither perfect nor perfectly futile).

A good way of gauging conference strength is to look at where conference members are in terms of overall schedule strength according to KRACH. 10 of 12 ECAC teams have schedule strengths that are in the Top 20 in the nation (all but Clarkson at 27th and Colgate at 47th). In fact, 13 of the top 15 are in the ECAC and the NCHC, speaking to the dominance thus far of those leagues in inter-conference play (.648 and .605 respectively).

Here's where the Engineers and their opponents this season currently sit in terms of KRACH. Overall records are included along with Rensselaer's record against them, where applicable.

1. Quinnipiac (13-0-2)
2. Harvard (6-1-1, 0-1-0)
4. Boston College (12-1-0, 1-0-0)
7. Cornell (6-1-2)
8. Yale (5-2-2, 1-0-0)
10. St. Lawrence (8-3-2, 1-0-0)
11. UMass-Lowell (10-1-4, 0-1-0)
13. Rensselaer (8-5-2)
14. Michigan (7-2-2, 0-1-0)
15. Miami (5-7-2)
18. Clarkson (7-4-2, 0-0-1)
21. Dartmouth (2-5-1)
25. Western Michigan (4-8-1, 1-0-0)
28. Union (4-6-3, 2-0-0)
29. New Hampshire (4-6-3, 1-0-0)
30. Alaska-Anchorage (6-6-2, 0-1-0)
32. Brown (2-4-3, 0-0-1)
43. Alaska (4-8-2, 0-1-0)
45. Bentley (6-6-1, 1-0-0)
47. Princeton (2-8-0)
49. Colgate (4-10-1)
52. Arizona State (3-8-0)

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.