Quest for #1 seed

[Steve M]

Good response John.  It's interesting how the average fan's projected brackets changed after a few of the Bracketology articles were published.  When the PWR tables were first posted on USCHO (nice work on that BTW), many people were assembling brackets that considered travel and attendance, and didn't blindly follow "natural brackets" according to the 1-16 rankings.  By now most people assume the 1-16 ranking matchups will be used even if it means over half the field has to be shipped across the country to accomodate them.   The apparent reason for this assumption is that that formula has been followed in Jayson's articles.  

As I've said earlier, the last 2 years the committee was able to set up brackets that followed the 1-16 ranking matchups as closely as possible while following the rules, without having to send any more teams out of region than was already necessary to follow the rules.  We will only know what takes precedence after they are forced to make a choice.

[Trotsky]

[Q]jtwcornell91 Wrote:
There is no sanctity of 1,8,9,16 etc.  The only place the handbook mentions 1-16 rankings is to divide the teams into four bands, and to rank the #1 seeds to place them into the nearest regional.  Trying to pair the lowest-ranked #1 seed with the highest-ranked #2 seed falls under the umbrella of "competitive equity" and "balancing the brackets", and they presumably will do it if their other considerations allow but it is not a hard and fast rule.  They do not think of a team as on overall #5 or #11 seed, but as a 2a or 3c seed.  This is a point they stressed to us when explaining how overall #1 Cornell got paired with overall #14 Mankato State when they could instead have avoided the intraconference matchups by pairing overall #3 Minnesota with overall #12 Harvard.  All the 1-seeds must play 4-seeds and the 2-seeds must play 3-seeds.  They'll even schedule intraconference matchups in the first found if they're forced to in order to maintain the 1-4 seeding in each regional.  That's what they mean by "integrity of the bracket".[/q]

I believe you are correct. However, the use of the term in their description is technically faulty.  Ensuring that a 1-band plays a 4-band and a 2-band plays a 3-band within each regional ought to be considered the integrity of the seedings.  Integrity of the brackets ought to mean ensuring that the winner of the #1 1-seed regional will play the winner of the #4 1-seed.  That, after all, is what "bracketing" means (as opposed to the reseeding used in the ECAC Tournament).

[billhoward]

Right, wrong, or variably correct. Well-written or cut-and-pasted boilerplate in the top half of each article (Jason, remember the inverted pyramid rule that puts the new and important stuff *on top*). The Bracketology column shines the light of publicity on the NCAA seeding committee and puts pressure on them to do a better job.

The more rigid the formula, the less likely some one team gets screwed by a dumb, irritated, or biased committee. And at the same time, it leads to lesser shaftings such as #1 Cornell playing #14 not #16 in the almost miracle year of 2003.

If the committee had a bit more leeway, you can see how it would be helpful if they could just, say, switch one of the region's #4 seeds to a #3 seed to avoid an undesirable matchup (like say two teams that just met in the Beanpot and played a three-overtime duel ought not to have to go at each other six weeks later). Or maybe it could avoid having even one intra-conference matchup in a tournament with five teams from the conference. But once you give the committee that power, what's to stop them from making a lot more swaps? And should they be able to make secondary considerations such as if a team gets an “unfair” first round pairing (eg #1 overall vs. #14 not #16 overall) to avoid some other miscarriage, do you try to even it out by, say, not putting Cornell in Amherst with its 200x95 near Olympic surface instead of 200x85 because big ice disadvantages Cornell's defensive style?

Meanwhile, what a rollercoaster for Cornell if you follow that long, thin strand of Bracketology type down the screen this week:
At the beginning of the column, Cornell starts out as a #1 seed (4th overall) and plays Colgate in Amherst. (Oops, that can’t happen: 2 teams, same conference.)
Cornell instead plays UAH, still in Amherst. Others in bracket: ND, BC.
Readjusting for bracket integrity (natural #1 vs. natural #16, etcetera):
Cornell plays Colgate again (theoretically).
Back to Cornell vs. UAH
Applying bonuses for good wins drops Cornell out of a #1 seed (also drops Colgate and Dartmouth out altogether) to #6
Cornell vs. BU, East Regional. Also Maine vs. Denver. Farther to travel, better ice size for Cornell.

And all this is predicated on the top teams winning  their tournaments.

Also, we don't know what the formula added to RPI is for good road, neutral, and home wins, although it might be possible to deconstruct the formula after the selections are made if some teams that would have been in with one RPI formula fall out with another.

Right now we should be rooting for Fair Harvard to knock off BC in the Beanpot consolation. Cornell's RPI won't get much help from quality wins down the stretch since the highest ranked of our last six opponents is #28 (out of 56 D1 teams) St. Lawrence.

[Robb]

[Q]billhoward Wrote:

Right now we should be rooting for Fair Harvard to knock off BC in the Beanpot consolation. Cornell's RPI won't get much help from quality wins down the stretch since the highest ranked of our last six opponents is #28 (out of 56 D1 teams) St. Lawrence.

[/q]

Cornell won't get ANY help from quality wins down the stretch.  Games against teams from your own conference don't count, even if it's a non-conference game.

[KeithK]

One other wrench in this.  The committee changes over time.  Different folks running things could decide to weight the various factors differently (attendance vs. natural brackets).  The only thing the handbook tells them is a requirement is the banding.

That said, I don't have any idea how frequently the committee turns over...

[billhoward]

So many rules for RPI and TUC and so forth. I have to write that on my shirt cuff: You can have an emotionally satisfying win against a quality team in your conference, but it's not an RPI-affecting quality win.

[KeithK]

That's what you get when you have a system that's fundamentally limited and constantly tweaked to try to address these limitations.

[jtwcornell91]

[Q]KeithK Wrote:

 One other wrench in this.  The committee changes over time.  Different folks running things could decide to weight the various factors differently (attendance vs. natural brackets).  The only thing the handbook tells them is a requirement is the banding.[/q]

That is certainly true.  The Marsh committees of the mid-to-late-nineties looked at individual comparisons much more than the committees of the past few years, who just looked at 1-16 in PWR.

[billhoward]

There is not, is there, consideration for performance late vs. early in the season? A young team that goes 9-6 first half and 11-4 second half is a 20-win team on a tear, but it's the same number of wins and same percentage as a team that starts off 12-3 and slumps with a 8-7 record second half.

The coach of that streaking team would want the upward slope to be recognized. But the coach of the cooled-off team beset by injuries is going to say, "Sure, we had a so-so second half, but we still won two-thirds of our games. We deserve the same shot as anyone else with a .667 record."

Last year, we wouldn't have wanted Vesce's injury and end-of-season non-availability to lead the seeding committee to say, "Cornell isn't going anywhere without Vesce, so let's give it to someone with a better chance."

[Ken \'70]

[Q]jtwcornell91 Wrote:

 [Q2]Ken '70 Wrote:

 You're mistaken if you think they casually disregard the sanctity of 1,8,9,16 etc.  

Do you think Jayson Moy of USCHO doesn't know what he's talking about, or have you observed the committee disregarding that seeding methodoloy in past years?[/Q]
Jayson Moy is a secondary source....Trying to pair the lowest-ranked #1 seed with the highest-ranked #2 seed falls under the umbrella of "competitive equity" and "balancing the brackets", and they presumably will do it if their other considerations allow but it is not a hard and fast rule.[/q]

You've lost track of the original focus of this particular string which is about how seedings are arranged relative to the regional finals. Go back and see calgARI's post about a tie at 4th and what followed.  

First of all, as USCHO notes at http://www.uscho.com/FAQs/?data=selection#b1 :

"Though not in its stated list of guidelines, the committee has made a concerted effort in the two years the 16-team field has existed to maintain a strict bracket. In other words, teams are given overall seed numbers, 1-16. The brackets are arranged 1-16-8-9, 2-15-7-10, 3-14-6-11, 4-13-5-12."

While the primary guiding principal in determining seedings and brackets is mentioned in the guidelines, that is:

"1. The top four teams, as ranked by the committee, are the four No. 1 regional seeds and will be placed in the bracket so that if all four teams advance to the Men’s Frozen Four, the No. 1 seed will play the No. 4 seed and the No. 2 seed will play the No. 3 seed in the semifinals."

the necessary corollary to this isn't, except by the phrase "competitive equity".

It would violate anyone's sense of competitive equity to allow bracket modifications such that the 1 seed had a more difficult trip to the semis than the 3 seed, for example.

In the first round, modifications are allowed to avoid intra-conference matchups unless they "corrupt bracket integrity".  There is no language that allows modifications such that potential regional final matchups are compromised in regard to seedings and competitive equity (e.g., allowing a 1 v 5 regional final instead of 1 v .

Modifications are limited as to cause and at what round they can occur, both in the handbook and as a matter of historical fact.  They do not occur under some broad brush and undefined "other considerations".  

Since you dismiss Moy and USCHO as "secondary sources", you must be a more primary one.  Please provide your bona fides.

[KeithK]

Moy is a secondary source because the only primary sources are the committee members themselves.  And they ain't talkin'.

There's pbviously a difference of opinion as to what "bracket integrity" means.  If it means preserving the natural 1-16 pairings, then #5 will certainly play #4 in the second round.  If it refers to the banding then #5 vs. #4 isn't guaranteed.  I tend to think it's the latter, based on the language of the guidebook and statements by the committee at the Townhalls in the past.

Anyway, we have two years of evidence for how the committee will seed the 16 team tournament.  I don't think we have enough "historical fact" to draw absolute conclusions.

[jtwcornell91]

[Q]KeithK Wrote:
Moy is a secondary source because the only primary sources are the committee members themselves.  And they ain't talkin'.[/q]

Yeah, I almost said "a secondary source, just like I am", but I didn't want to claim the same authority, since he's put more time into playing the what-if game in recent years than I have.  Back in the days of the 12-team tournament, however, I have played exactly the same weekly game.  Examples:

http://slack.net/~whelan/tbrw/tbrw.cgi?1998/pairwise.980105
http://slack.net/~whelan/tbrw/tbrw.cgi?2002/pairwise.020317

(Navigate around for links to other "If the Season Ended Today" columns, as well as the best explanations I was able to assemble between the selection handbook and questions to the NCAA back when the staff were more accomodating about educating coaches, journalists and fans.  Also, I think the old "You Are The Committee" scripts should still work and walk you through seeding a 12-team tournament.)

I haven't done it weekly since the advent of the 16-team bracket, but like many hockey-lers I've posted by prediction of the brackets on selection Sunday, and last year I managed to get it exactly right: http://lists.maine.edu/cgi/wa?A2=ind0403&L=hockey-l&D=0&F=P&P=32389&F=

The distinction between what the NCAA says and how USCHO paraphrases or interprets that is important.  USCHO is a great service, but it's not the gospel.

[jtwcornell91]

[Q]billhoward Wrote:

 There is not, is there, consideration for performance late vs. early in the season? A young team that goes 9-6 first half and 11-4 second half is a 20-win team on a tear, but it's the same number of wins and same percentage as a team that starts off 12-3 and slumps with a 8-7 record second half. [/q]

There was; record in the last 16 (originally 20) games was an additional criterion, but since there was no trace of schedule strength (as opposed to common opponents or record vs TUCs, which at least select the opponent pool), it was dropped a few years back.  (In the early days of the MAAC, teams like Quinnipiac ended up really high in the PWR by winning a lot of 2-1 comparisons on the strength of record in last 16 and vs TUCs.)

We did propose a way to incorporate strength of schedule into all the selection criteria, but it never got much traction:
http://slack.net/~whelan/tbrw/tbrw.cgi?kpairwise

[adamw]

[Q]jtwcornell91 Wrote:
The distinction between what the NCAA says and how USCHO paraphrases or interprets that is important.  USCHO is a great service, but it's not the gospel.[/q]

Actually, I think it is the gospel :-)  Or pretty close.

I think sometimes Jayson doesn't always carefully distinguish between the part that has leeway, and the parts that are absolute.  But he's more or less dead on.

Over the last two years, it has become absolutely obvious what the committee is doing, and Jayson is going by that -- along with a little insider knowledge that's hard to share.  Forget what the handbook says.  In practice, the committee has clearly attempted to keep 1-16, 2-15, etc... integrity across the board - except in only the most strict of circumstances - such as avoiding the first-round matchups against teams from the same conference.  That's just about the only reason that has changed under the 16-team field.

This caused confusion in year 1 - which if you go back to the ESPN selection show you will notice our confusion. We weren't sure, for example, why UNH and BU were paired in Round 2 - something the committee had always tried to avoid in the past.  Later, I realized what they did, and wrote a Next Day column that laid it out.

Last year, I think it was pegged almost exactly. Let's just say that last year, I also got a little more of a peak into the process.

There's a beautiful story to share about the NCAA office transcribing the wrong bracket on Sunday morning, after we had the whole thing figured out the night before.  After a half hour of panic and befuddlement, the NCAA called ESPN to say - wait, that was the wrong bracket - and proceeded to send the exact thing we figured out a night earlier.

[jy3]

[Q]jkahn Wrote:

 [Q2]jy3 wrote:
how do they figure out the winning percentage. i always thought that a tie counted as a win and a loss so 3-1-2 turns into 5-3 = .625.[/Q]
A tie is, as it should be, 1/2 a win and 1/2 a loss, so 3-1-2 equals 4-2-0 = .667.[/q]
ahhhh, that makes sense