Thursday, February 08, 2007

Attacking the TUC

I guess this is basically Brian's argument, but reformulated in a different way, and using different examples.

USCHO's Jayson Moy;s latest Bracketology article came out Tuesday. Noticably absent from his NCAA tournament field were the red-hot Michigan State Spartans.

I don't have a problem with that. The system is designed to look at every game, instead of just individual games, so stuff like that is bound to happen. But it's more the way that that happened that exposes a pretty big flaw in the system.

The RPI, the NCAA's preferred method of selecting teams for NCAA tournaments says Michigan State is a better team than North Dakota and Colorado College But the PWR says that North Dakota and Colorado College belong in the tournament while MSU does not.

Again, in and of itself, not a huge problem. The real problem lies with the Pairwise's "Team Under Consideration" category.

Michigan State only has a 5-6-0 record against TUCs, which causes them to lose their comparison with BC(7-7-0 vs. TUCs), Vermont(7-8-0 vs. TUCs), North Dakota(12-9-1 vs. TUCs), and Colorado College(11-8-1 vs. TUCs) despite having a higher RPI. Most people wouldn't have a huge problem with that. They'd say MSU should have won more games against quality teams.

But what exactly is a quality team? The PWR only looks at the top 25 teams in the RPI, which is an arbitrary number. Who is to say that team number 25 is a worthy opponent, while team number 26 is not.

A big reason that Michigan State has a poor record against TUCs is that Nebraska Omaha and Lake Superior have recently fallen out of the top 25 in the RPI. MSU is 1-0-1 against UNO and 2-0 vs. LSSU, meaning MSU lost 3 wins and 1 tie from their TUC category.

There's a perfect example of how significant that it is. The NCAA arbitrarily assigns a hidden RPI bonus for quality road wins. It's generally accepted that the bonus is somewhere around .003. Lake Superior, who is currently 27th in the RPI, is one of the teams to get that bonus. If the RPI bonus was changed from .003 to .0061, which is just enough to push Lake Superior into the 25th spot in the RPI, making them a TUC, Michigan State would go from 14th in the PWR and not in the NCAA tournament, to tied for 8th in the PWR.

The whole point of the PWR is to take the human element out of things, but look at how much movement there is just based on where a human decided to draw the line on what is a good team and how much a human decided a quality win was worth.

The TUC comparison in the Pairwise means that the NCAA tournament field is decided, in large part by the teams in the 20-30 spots in the RPI as much as the teams that are actually competing for an NCAA tournament bid. Winning would certainly help Colorado College this weekend. But just as important would be Minnesota State beating North Dakota, because a Minnesota State loss might push them into 26th in the RPI, meaning Colorado College would lose a valuable 2 wins in the TUC category.

The difference between #25 and #26 isn't the TUCs only problem. The other problem is that there is no difference between #25 and #1. There's little doubt that the WCHA has been the best conference in college hockey this year. But what has made their teams so good is not necessarily the top teams, but the ability of the bottom half of their conference to sneak into the TUC category.

For example, Minnesota State has played 22 games against TUCs, but 12 of those games have come against teams in the top 5 in the RPI. Colorado College, meanwhile, has played 20 games against TUCs so far, but 10 of them have come against teams 17th or lower in the RPI. The problem is that MSU's games against the top five teams count the same as CC's games against lower ranked TUCs. The TUC category is probably the only place in the world where someone will tell you it's just as easy to beat Minnesota as it is to beat Minnesota State.

The teams in the top 15 that are competing for NCAA tournament bids should generally have an advantage over the teams in the 20-25 range. So whichever conference gets the most teams in that range should have an advantage over the other conferences.

This is not an argument for a human-selected tournament field. That's definitely not a good idea. But I think it's also time to examine the formula for deciding the field. Ironically, the simplest solution to this problem that I've seen came from math-guru Alton Hollingsworth who proposed that every team is compared each other, with head-to-head being the sole factor, and the RPI being the tiebreaker in case the two teams haven't played. It's simple, and there's no arbitrary demarcation lines.

(NOTE: Some of this was made moot on Thursday night when UMass-Lowell beat Northeastern, hurting BC and helping St. Lawrence just enough to allow Michigan State to pass Boston College in the PWR.)


Packer487 said...

Has anyone ever run that formula (compare teams head to head w/RPI as the tiebreaker) to see how it would come out?

It's so simple it might actually make sense....

Donald said...

No doubt that people will poo-poo this because the team I support isn't in contention at all for the NCAA's however my opinion would be the same regardless.

Bracketology is just as STUPID as midseason polls.

Yes ... sometimes (or even if it is often) the predictions might turn out to be reasonably accurate but saying that "Joe Blow State University" (no relation to St. Cloud) is going to win this or that league tournament is a complete act of hubris.

I guess I just don't get the mentality of the thing. All the "bracketology" purveyors could jump off a cliff and I wouldn't miss them.

Here's hoping Merrimack, Bowling Green, Wayne St, AIC, RPI and Duluth all win their respective league championship games and secure autobids.

Chris said...

I don't think it's hubris. It's just that an assumption has to be made, so they make the most logical. Nobody is saying nobody else can get an autobid. It just keeps things simple.

Alton said...


The teams selected to the tournament field would have been exactly the same last year with that formula. The only seeding changes would have been (1) Miami would be the fourth #1 seed instead of Michigan State and (2) New Hampshire would have been a #3 and Maine a #4 seed, instead of the other way around.

New Hampshire, Maine and Nebraska-Omaha would have been the last teams in; Dartmouth, Denver and St Cloud would have been the best teams out.

I don't think the proposed formula would result in very many changes in the tournament field year-to-year; the main features ares that it would eliminate the arbitrary TUC cutoff, where (this season) the #25 team is the same as the #1 team and the #26 team is the same as the #58 team, and it would also eliminate the idea under "common opponents" that a team that is 7-1 against two common opponents is somehow worse than a team that is 2-0 against them.

Anonymous said...

Sour Grapes, This year the CCHA just doesn't measure up to Hockey East or the WCHA

gmf1a said...

Why doesn't anyone talk about KRACH anymore?

It is dangerous to tweak the system everytime the computer does not reflect public opinion.

Public opinion is what we should avoid. Teams with a small fan base are just as deserving as Mich State

Anonymous said...

The RPI ranking is perhaps the one that should be thrown out completely. Take a team like Notre Dame,who has a high RPI ranking and its based on a wins against the 39th ranked toughest schedule in the country. If Minnesot Duluth played that schedule ,along with all wcha teams ,their RPI's would be just as high. The RPI is a joke in its whole!!! DOESNT ANYONE ELSE SEE IT?

Chris said...

It's not a matter of the computer not reflecting public opinion. It's a matter of the computer being the one affected by opinion.

This was the most important part of what I wrote, which I think I did a poor job of highlighting:

The whole point of the PWR is to take the human element out of things, but look at how much movement there is just based on where a human decided to draw the line on what is a good team and how much a human decided a quality win was worth.

The TUC comparison in the Pairwise means that the NCAA tournament field is decided, in large part by the teams in the 20-30 spots in the RPI as much as the teams that are actually competing for an NCAA tournament bid."

I'm not saying Michigan State does or does not deserve to be in the tournament. I'm saying the difference between them hovering on the bubble and being a solid third seed in the tournament is because the current system sees a game against Minnesota State as being equal to a game against New Hampshire, and a game against Lake Superior as being equal to a game against American International, which I think we can all agree makes absolutely no sense.

Anonymous said...

59 team tournament. Lower seed plays @ the higher seed until you have 16 teams and then go ahead and send em off to the regionals.

Seriously though, they've gotta come up with something better. The RPI doesn't work, the PWR fails miserably, and the KRACH isn't perfect either. Time to make fundamental changes to the way the NCAA conducts the tournament.