What's new
By:
USCHO Fan Forum

This is a sample guest message. Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!

  • The USCHO Fan Forum has migrated to a new plaform, xenForo. Most of the function of the forum should work in familiar ways. Please note that you can switch between light and dark modes by clicking on the gear icon in the upper right of the main menu bar. We are hoping that this new platform will prove to be faster and more reliable. Please feel free to explore its features.

2010-2011 Rutter DI Computer Rankings Thread

Re: 2010-2011 Rutter DI Computer Rankings Thread

LakersFan said:
Not sure how this relates to Dave's point above, as I think some (if not most) of the variation is due to position in the rankings. The closer you are to the top or bottom, the more variable your estimate because you are not playing teams above or below you. The correlation between absolute value of the rating and the standard deviation is .80.
Click to expand...
Yes, it does make it clear that Cornell's rating is more uncertain than Wisconsin's. Probably pre-Clarkson tie, the contrast in uncertainty was even larger.

The uncertainty measures for Cornell are still lower than what my gut tells me.

Good luck with the teaching/grading.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

For games played through February 6, 2011

Code: 
  	Team 			Rating 	
1 	Wisconsin 		1.8004 
2 	Cornell 		1.5201 	
3 	Minnesota 		1.1790 	
4 	Minnesota Duluth 	1.0516 	
5 	Mercyhurst 		1.0285 	
6 	Boston University 	1.0054 	
7 	North Dakota 		0.7356 	
8 	Boston College 		0.6586 
9 	Bemidji State 		0.3694 
10 	Harvard 		0.3077
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

We say too much negative about the rankings, but they do a good job most of the time. Apart from Cornell, I think everyone's ranking is right where they belong. Particularly, I agree with ratings 3 through 6, and that these teams are very close, yet the teams deserve to be ranked where they're at.

I also don't think it does teams any good when expectations are inflated, like calling BU the solid No. 3 team in the country. It makes it sound as if losing in the quarterfinals would be an underachieving season for them. But it should be exceeding expectations for them to make the Frozen Four this season.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

So, how solid a #3 are they after losing to BC tonight?

I can't lie.... the scenario of BU #4 and Hurst #5 means I save a ton of $$$$$$.

:D
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

BU is less solid a No. 3 in the PWR, but BC moves up to No. 4. Things are looking are up for you.

My guess is the teams are probably about 6 & 7 in the Rutter rankings now. For one conference, this could be the largest discrepancy ever between the PWR and the various methods with solid statistical foundations.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

The Harvard-Northeastern result last night prompts me to ask how it's accounted for in the various rankings algorithms. True to their respective HE and ECAC outlooks, the official Harvard athletics website reports it as just a tie, the NU website as a tie-and-shootout-loss. As a tie, the result lowered H's won-lost percentage, but what is its effect in the more sophisticated rankings algorithms?
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

Watson Rink said:
The Harvard-Northeastern result last night prompts me to ask how it's accounted for in the various rankings algorithms. True to their respective HE and ECAC outlooks, the official Harvard athletics website reports it as just a tie, the NU website as a tie-and-shootout-loss. As a tie, the result lowered H's won-lost percentage, but what is its effect in the more sophisticated rankings algorithms?
Click to expand...
The NCAA criteria (PWR), KRACH, and Rutter do not consider shootouts, whether they be WCHA regular season games or Beanpot.

WCHA teams usually list their overall record, but only list the shootout wins in conference record. Not sure what Northeastern's doing. Hockey East no longer has shootouts, and the game is nonconference, so I see no reason why Northeastern should reflect the result as anything but at a tie in their record.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

Here are my rankings along with the raw RPI (that is, not adjusted for games against teams that cause your RPI drop) after the Beanpot.

Code: 
	Team 			Rating 	RPI.Rank 	RPI
1 	Wisconsin 		1.7981 	1 		0.6356
2 	Cornell 		1.5197 	2 		0.6193
3 	Minnesota 		1.1759 	5 		0.5833
4 	Minnesota Duluth 	1.0502 	7 		0.5777
5 	Mercyhurst 		1.0291 	6 		0.5803
6 	Boston University 	0.9415 	3 		0.6021
7 	North Dakota 		0.7290 	8 		0.5613
8 	Boston College 		0.7176 	4 		0.5857
9 	Bemidji State 		0.3665 	14 		0.5113
10 	Harvard 		0.2953 	10 		0.5295

Why is BU lower in my rankings than in RPI/PWR. I think BU's two ties against Vermont really hurt them in my rankings (almost as bad as two losses, considering the difference in quality) compared to the RPI. Those ties also don't show up in the TUC calculation for PWR. Can't even begin to explain why BC is currently ranked four in the PWR.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

LakersFan said:
Can't even begin to explain why BC is currently ranked four in the PWR.
Click to expand...

Your rankings correctly determine that Hockey East is a weaker conference than the WCHA. BU and BC lost to some weak teams, who in turn lost to some weak teams. The RPI barely picks this up. Teams' RPI ratings are much more about within-conference dispersion than across-conference dispersion, so it's not surprising that co-leaders of Hockey East are ahead of 2nd-place Minnesota and UMD in the RPI and not in your rankings (where Minnesota & UMD get appropriate props for beating Wisconsin, and Mercyhurst gets props for beating Cornell). So everyone in Hockey East is overrated. BU has a fantastic win pct. so BC gets a lot of credit for going 2-2 against BU.

RPI creates an illusion of more across-conference parity then there really is. To illustrate, consider if you had conference A and conference B, and every team in conference A was better than every team in conference B. Each team in conference B places conference A once and loses, and then plays a large within-conference schedule. Your ranking would appropriately rank every team in conference A ahead of conference B. In the RPI, the nonconference results would become of diminishing importance as the share of the within-conference schedule becomes larger. If there was much more parity in conference A than in conference B, and the nonconference schedule was tiny, you'd see the leader in conference B ranked No. 1 in the RPI.

Despite this superior feature of your rankings, I would never see the NCAA adopting it, because it likes to encourage an illusion of cross-conference parity,and encourage participation across conferences in the NCAA.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

dave1381 said:
Despite this superior feature of your rankings, I would never see the NCAA adopting it, because it likes to encourage an illusion of cross-conference parity,and encourage participation across conferences in the NCAA.
Click to expand...

I would be happy to provide my services to the NCAA, free of charge, in case they are reading. KRACH has BC 8th as well, so the other rating system based on statistical theory has the same top 8.

Looking at the current Boston College/Minnesota Pairwise Comparison:

Boston College vs. Minnesota
.5896 1 RPI 0 .5853
6-3-1 .6500 1 TUC 0 .4167 4-6-2
1-0-0 1.000 0 COp 0 1.000 2-0-0
0-0-0 0 H2H 0 0-0-0

If you use the RPI weights that NCAA uses in Men's hockey, UM jumps BC in RPI, using an RPI that does not remove wins that result in a lower RPI. (Aside to Dave: does USCHO.com do this by hand or is there a great algorithm I have yet to run across). Minnesota also plays most of all of its TUC against top 8 teams, while BC split with BU and is 4-1-1 against teams 9-12 in TUC.

Unless BC collapses and UM wins out against TUC, BC has a lock on TUC. Dave, do you think UM remaining schedule (2 vs. Bemidji, 2 vs North Dakota, and the WCHA tourney) is strong enough to leap them over BC in RPI?
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

And WCHODR has BC 8 as well.

For Minnesota, what matters is jumping either BU or BC. One of these teams won't win Hockey East. Neither Boston team has particularly daunting schedules coming up, but they may be expected-RPI-optimal in the sense that these are the kind of opponents that maximize their expected RPI given their probability of winning given the gain in RPI.

To answer your question, I calculated an expected RPI for BC, BU, and Minnesota for all D-I opponents using win probabilities from your rankings (I ignored ties), and your RPI calculation. The ideal opponent is one whose strength is overrated by RPI's way of measuring strength. The best opponent for BC or BU is Princeton. The best possible opponent for Minnesota is BC (and BU's second-best possible opponent). The worst possible opponents are typically the best teams.

The expected RPI for BC's remaining opponents are:
Opponent (Games), Expected RPI, Rank of Expected RPI out of D-I opponents
Maine (2), .5756, 14
Harvard (1), .5734, 15
UNH (2), .5835, 4
HE Semi, .5786, 9 (I averaged 4 potential opponents to get this #)
HE final, .5567, 24 (just assumed BU)

Average expected RPI of remaining games: .5753

The expected RPI for Minnesota's remaining games
Bemidji (2), .5982, 16
NDakota (2), .5883, 24
Quarterfinals (2 or 3), .5850, 24 (I averaged OSU and MSU, though drawing Bemidji is is ideal for the Gophers)
Semifinal and Final (2) potential opponents
North Dakota .5883, 24
UMD .5905, 20
Wisconsin .5359, 32

The average RPI of these games is .5834

Of course I've made a bunch of simplifying assumptions along the way that become complicated by hand when we consider the tree of conference tournament possibilities.

So in expected terms with similar simplifying assumptions, they look similar in expected terms of their RPI. Probably Minnesota has roughly a 50-50 chance of passing BC, even if you're right that Minnesota is the far better team?
 
Last edited:
Re: 2010-2011 Rutter DI Computer Rankings Thread

Well, all that discussion about the probability of BC getting a home game in the tournament now appears to be for naught.

For games played through February 13, 2011

Code: 
  	Team 			Rating 	
1 	Wisconsin 		1.8695 
2 	Cornell 		1.5295 
3 	Minnesota 		1.2383 	
4 	Minnesota Duluth 	1.1132 	
5 	Mercyhurst 		1.0628 	
6 	Boston University 	0.9642 
7 	North Dakota 		0.6911 	
8 	Boston College 		0.5671 
9 	Bemidji State 		0.3158 
10 	Harvard 		0.3017
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

How does this currently compare to the Pairwise? (I can't seem to connect to the rankings portion of the page!) If someone could paste that in someplace, I'd appreciate it.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

Rank Team PWR W-L-T Win % Win % Rank RPI RPI Rank
1 Wisconsin 11 28-2-2 .9062 2 .6611* 1
2 Cornell 10 25-1-1 .9444 1 .6393* 2
3 Boston University 9 24-4-3 .8226 4 .6111* 3
4 Minnesota 8 22-7-2 .7419 5 .5904* 4
5 Mercyhurst 7 26-5-0 .8387 3 .5889* 5
6 Minnesota-Duluth 6 18-7-3 .6964 7 .5825* 6
7 Boston College 5 19-6-5 .7167 6 .5763* 7
8 North Dakota 4 17-10-3 .6167 11 .5523 8
9t Providence 2 19-11-1 .6290 10 .5411 9
9t Dartmouth 2 17-9-0 .6538 9 .5380 10
9t Harvard 2 14-8-4 .6154 12 .5362 11
12 Quinnipiac 0 20-10-2 .6562 8 .5327* 12

Read more: http://www.uscho.com/rankings/#ixzz1DxvlcdVd
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

Can anyone explain to me why the KRACH Strength of Schedule column works the way that it does? For example, consider St. Cloud State, whose schedule nets a rating of 183.2 that ranks 6th, while Minnesota-Duluth has a SOS of 284.6 that is #1 in SOS rank. As we head into the final weekend of the WCHA regular season, these two teams have played the exact same common opponents in conference. Their other conference games are two games versus each other. In nonconference action, UMD played UConn twice while St. Cloud had 3 with Mercyhurst and one each with Northeastern and Wayne State. So the net difference in the schedules beyond the 24 identical games for each is that UMD played UConn (x2) and SCSU (x2), while St. Cloud saw UMD (x2), Mercyhurst (x3), NU, and WSU. In what strange model are those first 4 games viewed as being considerably more difficult than the other 7?
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

From what I understand, the SOS is more of a back-of-the-envelope kind of thing. It's equal to the team rating in KRACH divided by the win ratio (the number of wins / number of losses). So when take winratio * SOS = KRACH rating. I'm not sure what the correct interpretation of comparison of SOS factors actually is.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

For games played through Feb. 20, 2011

Code: 
  	Team 	            Rating 	
1 	Wisconsin 	    1.8613 	
2 	Cornell 	    1.3700 	
3 	Minnesota 	    1.1848 	
4 	Minnesota Duluth    1.1022 	
5 	Mercyhurst 	    1.0598 	
6 	Boston University   0.8517 	
7 	North Dakota 	    0.6978 	
8 	Boston College 	    0.6107 
9 	Dartmouth 	    0.4038 		
10 	Providence  	    0.3635
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

LakersFan said:
For games played through Feb. 20, 2011

Code: 
  	Team 	            Rating 	
1 	Wisconsin 	    1.8613 	
2 	Cornell 	    1.3700 	
3 	Minnesota 	    1.1848 	
4 	Minnesota Duluth    1.1022 	
5 	Mercyhurst 	    1.0598 	
6 	Boston University   0.8517 	
7 	North Dakota 	    0.6978 	
8 	Boston College 	    0.6107 
9 	Dartmouth 	    0.4038 		
10 	Providence  	    0.3635
Click to expand...

I think this is a great recommendation for the NCAA selection committee. It avoids all intra conference 1st round matchups. Not the best travel schedule, but what I like best is the potential for an all WCHA Frozen Four. I am sure for that reason alone it won't meet a lot of the committee members criteria.
 
Re: 2010-2011 Rutter DI Computer Rankings Thread

Fro games played through Feb. 27, 2011

Code: 
  	Team 	            Rating 
1 	Wisconsin           1.8785 	
2 	Cornell             1.3976 	
3 	Minnesota           1.2349 
4 	Minnesota Duluth    1.1296 
5 	Mercyhurst          1.0600 	
6 	Boston University   0.8521 
7 	North Dakota        0.7125 	
8 	Boston College      0.6135 
9 	Dartmouth           0.4041 	
10 	Providence          0.3884
 
You must log in or register to reply here.
Back
Top