Yale and the Pairwise

[BigDEli]

Re: Yale and the Pairwise

The final poll of the year came out today, which is completely and utterly meaningless. But I think it's cute that Quinnipiac was given three first place votes, and that Canisius makes the list after winning its crappy league tournament and ALMOST beating a disinterested QU team in the regional, meanwhile finishing at .500 exactly. Show just how useless polls (sorry; poles) are, even during the season. Pairwise is all that matters, period. Just ask Yale!

I wish the last poll wasn't anonymous, just for the insight into how different pollsters weigh the tournament. 3 voters put Q ahead of Yale, presumable because they won 3 of 4 in the series, placing more value on the regular season than the tournament. But SCSU jumps up from 9 to 4 for winning a regional. BC finishes 1 spot ahead of Union even though they got smoked head to head, but Canisius gets top 20 votes just for scaring Q in the 1st round.

And am I reading the poll wrong, or doesn't Yale getting 993 points mean that not only did those 3 voters not put Yale #1, they didn't vote them 2nd either? Yes, I'm a biased Yale fan, but that's pretty odd.

 

[goblue78]

Re: Yale and the Pairwise

If those 3 voters simply mirrored the RPI, then Yale would have had 991 votes (47x20 + 3x17). So it's consistent with two of them ranking Yale at fourth and the other voter making them 2nd. (47x20 + 2x17 + 19). If two voters had them 2nd, then the other guy had them 7th, which is their ranking by pure winning percentage, which is even more asinine.

 

[paulfb]

Re: Yale and the Pairwise

Congratulations to Yale and its fans from this longtime BC fan. I easily "adopted" your team after BC's exit.

I've rooted for BC since 1973 but I also lived in Hartford from 1977 to 1988 and always enjoyed visits to Ingalls. I still go up there a couple of times a year from Staten Island, NY, and my girlfriend likes the museums, too.

Tip of the hat to Quinnipiac, too, for an inspired run.

This is great not only for the ECAC and the New Haven region, but for college sports fans in Connecticut that are sick and tired of "all UConn, all the time."

 

[matt]

Re: Yale and the Pairwise

When even Barry Melrose knows your team barely made the tournament, you know that everybody knows it. But, fulfilling my role as a Yale fan and a guy who “actually understands the Pairwise,” as a Minnesotan I met in the Marriott bar two hours before the Championship game called me, I think it’s worth recounting how close Yale was to being out of the tournament, instead of winning it.

First is the one everyone knows: Notre Dame had to beat Michigan on the last day of the season. People focus on this one because it happened after Yale’s season was over, but the Pairwise doesn’t care about when. Obviously, any extra autobid would have kept Yale out. In particular, had BU beaten Lowell in the HE finals, Yale would have been out. I leave it to others to speculate whether it would have been better, knowing what we now know, for Lowell to have thrown that game. Just sayin’.

But autobid shenanigans are the least of the miracles that kept Yale in. On March 1, Yale beat Colgate with an overtime goal with 2 seconds left that bounced off a stick left on the ice behind the net and happened to bounce in front of the net for Josh Balch to bang home. The game can be seen at http://www.youtube.com/watch?v=pXEl_dCS9RY and the play occurs at around 2:35 on the video. Suppose that game ends in a tie. Yale is 16th in the Pairwise, and out.

Even odder, though, go back to October 13. Yale’s season hadn’t even started yet. St. Lawrence was in Kalamazoo to play the first of two against Western Michigan. The Saints won in OT. Had they tied that game, Yale is out again.

The Pairwise is, famously, highly unstable. And any rating system is going to be highly unstable for teams on the bubble. But the notion that the season came down to an otherwise unmemorable goal scored in OT in October between two teams who weren’t in the tournament is, I think, still striking, given that it changed the identity of the National Champion.

Let me be clear, these results, and dozens of others I could dredge up (fun with the Pairwise!) are clearly one sided – there were obviously other bounces that went against Yale that would have put them safely in. It’s still striking to consider just how random the last five teams out and the last five teams in really is. The bubble doesn’t really matter in basketball because, as a practical matter in a 64 team tournament, no bubble team has a realistic chance. But in a 16 team tournament with only one true autobid (all the other autobids would have qualified, or been replaced, like Union and Brown, with another autobid who would have qualified), as we now know, you can easily change the National Champion with your selection criteria. Not that I’m complaining.

Let me be clear on one more thing. I’m not saying Yale didn’t “belong” in the tournament. Frankly, I don’t even know what that means. As far as I’m concerned, they belong every year because I’m a Yale fan. The formula determines who gets in, and they got in under the formula. In a smoke-filled backroom, they might well have been replaced by Brown, who played much better down the stretch. And maybe Brown would have won the tournament if they’d been in. Who knows? But while I prefer formulas to back rooms, this does demonstrate that the formula can have a piece of determining who is Champ, which is not, I think what the NCAA really wants, although in a 16 team field in a world of relative parity, they may not be able to help it.

While the closeness of Pairwise is obviously true, it just as obviously cuts both ways. In 2012, for example, Yale lost some tough games, including one where they led eventual national champion BC with less than 2:30 to go, only to lose. And, obviously, if the ECAC had done something as simple as eliminate the silly consolation game, there really wouldn't even be a topic here (as some St. Lawrence fans about that one...).

Regarding Basketball, "The bubble doesn’t really matter in basketball because, as a practical matter in a 64 team tournament, no bubble team has a realistic chance," is simply untrue. The bubble in basketball is significantly larger than in hockey, and is entirely made up of at-large teams. VCU was a bubble team in 2011, and had to play in one of the "play-in" games that got the tournament down from 68 teams to 64 (two of the play-in games are for 16 seeds/autobids, and the other two are for at-large/bubble teams). They won the play-in game and eventually advanced to the final four.

In hockey, as in basketball, there might be other systems. But short of simply letting in more teams, there is always going to be contention for those final at-large spots. In basketball it's subjective, and in hockey there's a flawed but objective system that, yes, can be swayed by absurd goals off discarded sticks (Andrew Miller's, in the Colgate game, if it matters). I'm good with it. ;-)

 

[matt]

Re: Yale and the Pairwise

I think that my Colgate example above certainly fits into your view. Obviously, if you change a win for a tie then almost any system will make you do worse. But NFL and NHL systems are immune to performance by teams who don't make the playoffs, for the most part. They come down to your own performance, not the performance of others. (OK, maybe the WMU example isn't perfect either, since they make it by changing that October loss to a tie. But it isn't the only one.)

NHL and NFL rating systems don't care about strength of schedule, at least not explicitly. When a tiebreaker comes down to conference record, they don't care whether both teams played the same number of games against each team in the conference. When baseball picks a wildcard team, they completely ignore the fact that teams in different divisions played different teams.

Looking at strength of schedule is almost a necessity in hockey, because so few interconference games are played. But it leaves a team's standing dependent not just on their performance (although I agree that's by far the biggest effect) but on the performance of teams who they aren't even competing with directly. So my point is that how you calculate what is in, essence, SOS (RPI, and Common Opponent criteria) matters more than it matters in football, for example.

"NHL and NFL rating systems don't care about strength of schedule, at least not explicitly."

Who told you this??? Here's the NFL tie-breaker system: http://www.nfl.com/standings/tiebreakingprocedures
See tie-breakers #5 and #6 among two-team ties in a division or a conference, for example. And YES, those tie-breakers have been used.

The NHL does not consider schedule; rather, they go to the far worse method of goal-differential. But the NHL has so many possibilities, including how it handles shoot-out games, that it's very difficult to get to that last tie-breaker.

 

[Patman]

Re: Yale and the Pairwise

College basketball is also a tale of 6 games and not 4.

Thus two more chances to thin out a true bubble team.