(PWR) PairWise Rankings - 2009

[Hale4David]

Hopefully it's your computer! What happened to the win out and have these great odds at a #1 seed?!? I know we tied yesterday, but that shouldn't drop our odds terribly.

I know little to nothing of how the PWR works but I'm guessing that the recent rough stretch WI has been on recently hasn't helped those predictions either. I still have to believe that if we win one vs. WI and make it to the final five that we'd be in the NCAA's...but like I said at the start of this post.....

[siouxweet]

Hey, Jim how is it that our game RPI with a win over CC was less than our game RPI with wins over Mankato and SCSU?

[siouxweet]

Yep UND is up to #7 and if the season ended today I'd guess the brackets would go like this. Of course, luckily it doesn't and UND can continue to move up with victories.

Bridgeport

2 Notre Dame

15 Air Force

7 UND

9 Yale

Minneapolis

4 Denver

14 Ohio State

5 Northeastern

12 Princeton

Manchester

1 BU

16 Bemidji

8 New Hampshire

10 Cornell

Grand Rapids

3 Michigan

13 Minnesota Duluth

6 Vermont

11 Miami

If the gophers aren't in the big dance flip us and Northeastern and put us in the west.

[ndhunter]

Yep UND is up to #7 and if the season ended today I'd guess the brackets would go like this. Of course, luckily it doesn't and UND can continue to move up with victories.

Bridgeport

2 Notre Dame

15 Air Force

7 UND

9 Yale

Minneapolis

4 Denver

14 Ohio State

5 Northeastern

12 Princeton

Manchester

1 BU

16 Bemidji

8 New Hampshire

10 Cornell

Grand Rapids

3 Michigan

13 Minnesota Duluth

6 Vermont

11 Miami

I think Notre Dame would go to Grand Rapids instead of Bridgeport which will change the sioux to Grand Rapids. Most of your brackets matchups are corect.

[siouxweet]

Yep UND is up to #7 and if the season ended today I'd guess the brackets would go like this. Of course, luckily it doesn't and UND can continue to move up with victories.

edit to correct seeding:

Bridgeport

2 Notre Dame

15 Air Force

7 UND

9 Yale

Minneapolis

4 Denver

14 Ohio State

5 Northeastern

12 Princeton

Manchester

1 BU

16 Bemidji

8 New Hampshire

10 Cornell

Grand Rapids

3 Michigan

13 Minnesota Duluth

6 Vermont

11 Miami

Bridgeport

3 DU

14 Ohio state

5 Northeastern

9 Yale

Minneapolis

4 Michigan

13 Minn-Duluth

6 Vermont

11 Miami

Grand Rapids

2 Notre Dame

15 Air Force

7 North Dakota

12 Princeton

Manchester

1 BU

16 Bemidji State

8 New Hampshire

10 Cornell

[stickboy1956]

Bridgeport

4 DU

14 Ohio state

5 Northeastern

9 Yale

Minneapolis

3 Michigan

13 Minn-Duluth

6 Vermont

11 Miami

Grand Rapids

2 Notre Dame

15 Air Force

7 North Dakota

12 Princeton

Manchester

1 BU

16 Bemidji State

8 New Hampshire

10 Cornell

DU is seeded 3 not 4. Both CHN and USCHO break PWR ties using the head to head PWR comparison, not RPI.

[siouxweet]

DU is seeded 3 not 4. Both CHN and USCHO break PWR ties using the head to head PWR comparison, not RPI.

guess I should have looked myself instead of thinking someone else got it right anyway. DU wouldn't play Duluth in first round anyway so the seeds may be different but the brackets would stay the same as Michigan wouldn't play OSU. I did go back and change it.

[jimdahl]

Hey, Jim how is it that our game RPI with a win over CC was less than our game RPI with wins over Mankato and SCSU?

Re: UND's RPI breakdown

It's the much-derided 51% weighting on opponents' opponents' win percentage. CC has a better win% than SCSU or Mankato against roughly similar competition so should arguably contribute more to UND's strength-of-schedule than either of those other teams. Instead, RPI dramatically overemphasizes (relative to the past and to other sports) the win% of CC's opponents when considering how beating CC affects our strength-of-schedule.

Despite CC being an obviously better team than Mankato, RPI gives more credit for playing Mankato because Mankato has played a slightly tougher schedule.

[yzerman19]

I'm going with RPI as the tie breaker, that is what I have known to be true.

Manchester

1. BU

16. BSU

8. NH (host)

10. Cornell (swapped with 9 Yale as Yale is a host)

Bridgeport

2. NtD

15. AF

7. UND

9. Yale (swapped with 10 Cornell to be a host)

Grand Rapids

3. MI

13. UMD (swapped with OSU to avoid first round intraconference)

6. Vt

11. Miami

Minneapolis

4. DU

14. tOSU (swapped with UMD to avoid first round intraconference)

5. NE

12. Princeton

They might consider swapping NE and UMD for attendance in Mpls, but the integrity of the bracket tends to be the most important

[siouxweet]

I'm going with RPI as the tie breaker, that is what I have known to be true.

Manchester

1. BU

16. BSU

8. NH (host)

10. Cornell (swapped with 9 Yale as Yale is a host)

Bridgeport

2. NtD

15. AF

7. UND

9. Yale (swapped with 10 Cornell to be a host)

Grand Rapids

3. MI

13. UMD (swapped with OSU to avoid first round intraconference)

6. Vt

11. Miami

Minneapolis

4. DU

14. tOSU (swapped with UMD to avoid first round intraconference)

5. NE

12. Princeton

They might consider swapping NE and UMD for attendance in Mpls, but the integrity of the bracket tends to be the most important

If possible the west will have either the gophers, and if they don't make it either UMD or UND.

[YaneA]

I'm going to ask what is probably a dumb question. Unlike most of you posting in this thread, I can't do any parlor tricks like statistical analysis or stuff mathematical so here's my question: Isn't there any way to freeze the value of a win or the detriment from a loss in these formulas? I mean, if you beat a team when that team is riding high (think UND's sweep of the Gophers before their nosedive) shouldn't those wins be worth more "points" for lack of a better word than someone else beating that team when that team is in the dumper?

[sagard]

I'm going to ask what is probably a dumb question. Unlike most of you posting in this thread, I can't do any parlor tricks like statistical analysis or stuff mathematical so here's my question: Isn't there any way to freeze the value of a win or the detriment from a loss in these formulas? I mean, if you beat a team when that team is riding high (think UND's sweep of the Gophers before their nosedive) shouldn't those wins be worth more "points" for lack of a better word than someone else beating that team when that team is in the dumper?

That opens the door for subjectivity to play a role. So I wouldn't pursue it. Unless your considering a Saturday night win vs. UMD gets you an auto-entry.

[nodakvindy]

I'm surprised the computer says the chances are so low of moving with a sweep of the Badgers. A sweep will almost certainly flip the comparison against Denver, since the Pios only have one game and will fall behind in RPI. Also if Vermont and New Hampshire split or if UNH takes three points, I think that will allow the Sioux to pass UVM in RPI and flip that comparison, but prevent UNH from overtaking us in RPI. The ECAC and CCHA contenders are all off, so RPI changes should not be significant with them and at this point RPI moves are what will be flipping most comparisons. IF UND flips the Vermont and Denver comparisons, they would likely be at #4 and the final #1 seed.

Sioux fans should be cheering hard for Michigan State now. If they win this weekend, they would play either Michigan or Notre Dame next weekend. If they could ride Jeff Lerg to an upset series win over either of those teams, the Sioux could flip that comparison as well.

[jimdahl]

I'm going with RPI as the tie breaker, that is what I have known to be true.

All the Championship handbook says is:

When comparing two teams, the team earning the most points will be given consideration in the selection process.

If the point process provides a tie, the Rating Percentage Index may serve as the determining factor, regardless of the difference.

Though we always think of the rankings as deterministic, USCHO's FAQ says:

Teams are then ranked by PWR point total, with ties broken by looking at the Rating Percentage Index (RPI). Note: this tiebreaking procedure is used solely for convenience in displaying the PWR, and will not necessarily match the committee's process. This is especially true near the end of the top 16, where the committee looks more closely at head-to-head comparisons when selecting the last few teams.

However, the committee does indeed use RPI to break ties within those head-to-head matchups. For example, if that head-to-head comparison is tied, or if there is a transitive tie in a three-way comparison (A defeats B, B defeats C, C defeats A), then RPI is indeed used to break the deadlock.

Emphasis added.

To me that implies that in the case of a tied PWR ranking the team that wins the head to head comparison gets the nod (not to be confused with the h2h games criterion), with RPI used to resolve ties only when the comparison itself is tied.

[nodakvindy]

Adam Wodon has written a couple of articles in the past clarifying this. Pairwise ties are broken by RPI. It was demonstrated on a three way tie between Maine, UMass and St. Lawrence in 2007. The same held true in 2008 on a tie break between Denver and BC. USCHO's belief otherwise is part of why Jason Moy has not correctly picked the final bracket in recent years.

http://www.collegehockeynews.com/news/2007...8_final2007.php

http://www.collegehockeynews.com/news/2008/03/24_outof.php

[jimdahl]

Adam Wodon has written a couple of articles in the past clarifying this. Pairwise ties are broken by RPI. It was demonstrated on a three way tie between Maine, UMass and St. Lawrence in 2007. The same held true in 2008 on a tie break between Denver and BC. USCHO's belief otherwise is part of why Jason Moy has not correctly picked the final bracket in recent years.

Thanks -- great link. The PWR table here continues to rank teams with RPI as the tie-breaker, as it has since that language was added to the championship handbook and me not yet having seen convincing evidence otherwise (despite CHN and USCHO doing it differently on their tables). What I don't understand is why, if Wodon knows it wrong, CHN continues to publish an incorrect table some two years later?

[stickboy1956]

Thanks -- great link. The PWR table here continues to rank teams with RPI as the tie-breaker, as it has since that language was added to the championship handbook and me not yet having seen convincing evidence otherwise (despite CHN and USCHO doing it differently on their tables). What I don't understand is why, if Wodon knows it wrong, CHN continues to publish an incorrect table some two years later?

Sounds like we need to hold Wodon's feet to the fire when he posts his upcoming Bracket Analysis - let's check to see if he breaks PWR ties by RPI, even if the CHN rankings do not.

[siouxweet]

as hard as this is for me to say, BC sweeping NE this weekend would help us in the pairwise as it would kill NE's rpi.

[siouxweet]

I know little to nothing of how the PWR works but I'm guessing that the recent rough stretch WI has been on recently hasn't helped those predictions either. I still have to believe that if we win one vs. WI and make it to the final five that we'd be in the NCAA's...but like I said at the start of this post.....

I think you are right on with what we need to do. a split with wisc and sweep our first round series should get us in.

[The Whistler]

I think I'm figuring this out. Winning is good, right?

:(

[nodakvindy]

Thanks -- great link. The PWR table here continues to rank teams with RPI as the tie-breaker, as it has since that language was added to the championship handbook and me not yet having seen convincing evidence otherwise (despite CHN and USCHO doing it differently on their tables). What I don't understand is why, if Wodon knows it wrong, CHN continues to publish an incorrect table some two years later?

Jim, you are right and I should have pointed out the discrepancy, as I saw they are showing Denver ahead of Michigan right now. But, when Adam does his pre-announcement picks, he clearly knows it and that's why his predicted brackets have been right and USCHO's have not.

Also in Adam's defense, his analysis of the status of the rankings is more useful that a "snapshot" of what the bracket would be right now. There are generally a few teams that are almost locked into a PWR position and therefore safe, while there are others that have a seemingly good PWR position but are hanging on to several comparisons by a thread. I personally like to know which of those teams are the latter. I'm enough of a stats geek to look at it myself, but I think Adam does a great job of analyzing it for those that can't/don't want to.

[tho0505]

as hard as this is for me to say, BC sweeping NE this weekend would help us in the pairwise as it would kill NE's rpi.

I sure hope not, I don't want to see BC anywheres near the NCAA.........or DU........ or UMN.......

[stickboy1956]

Jim - Re the WCHA what if calculator - what is the tiebreaker formula if CC, Wisco, UMTC, UMD and SCSU all end up with 31 points?

When I use the calculator it ranks them in that order 3-7.

[nodakvindy]

Jim - Re the WCHA what if calculator - what is the tiebreaker formula if CC, Wisco, UMTC, UMD and SCSU all end up with 31 points?

When I use the calculator it ranks them in that order 3-7.

From the last WCHA weekly press release

In the event that ties are encountered in the determination of WCHA ranking or designation of home teams for playoff purposes, the following procedures will be used in the order given to break the ties:

a) If two or more teams are tied, head-to-head competition during the regular (conference) season will be used to break the tie.

b) If two or more teams are still tied after (a), the highest seed will go to the team with the most WCHA (conference) wins during the regular season.

c) If two or more teams played a four-game series during the regular season and the teams have the same win-loss records for those series and the same number of WCHA wins, the team having the least number of goals scored against it in the four-game series shall have the higher rank. If two or more teams played a two-game series during the regular season, proceed to tie-breaker d).

d) If two or more teams are still tied after applying the provisions of (a), (b) and ©, the team having the greatest ‘winning margin’ during the regular season will have the higher rank. Winning margin = WCHA goals for during the regular season minus WCHA goals against.

A lot would depend on how they decide to consider the first tie breaker. Would a team have to beat every team head to head to win this tiebreaker? Would it be best overall winning percentage among the tied teams? Would it be the team that had the best series record among the team?

Another factor is whether you start over after breaking the first tie.

I think the first tiebreaker would be too difficult to determine with 5 teams, so they would go to the second and in that scenario, St. Cloud would be #3 as they would have 15 conference wins. #4 would depend on how Wisconsin got their two points. If it was a split, they would be #4 and Minnesota would be #5, since Wisconsin would have 14 wins and Minnesota 13. If the Sioux and Badgers tie twice, then the #4 and #5 would go to the head to head matchup (UW 2-1-1). Finally UMD and CC would have 12 league wins and the #6 would go UM-D by virtue of their 1-0-1 mark against CC.

so I would predict

3. St. Cloud State

4. Wisconsin

5. Minnesota

6. Minnesota-Duluth

7. Colorado College

[yzerman19]

A Sioux win and a Denver loss tonight will put the Sioux in a 4 way tie for #4 in the PWR With Denver, Vermont and Northeastern.

Interesting stat that I hope is telling: UND has the second best TUC record of the top 25 after BU, and only Denver has played more TUCs.

Also only BU and (unfortunately) UMD have the H2H vs the Sioux.

What does this mean? Not much...its paper, but it gives me confidence come tournament time.