Polls 2013-14

[Trotsky]

Back to 9th in PWR after the Q win.

[ugarte]

Back to 9th in PWR after the Q win.

And beating Princeton drops us into a 3-way tie for 11th. Never go to New Jersey.

[Dafatone]

Back to 9th in PWR after the Q win.

And beating Princeton drops us into a 3-way tie for 11th. Never go to New Jersey.

I thought RPI didn't drop if you beat a bad team.  Did the other teams around us win, or am I misunderstanding that mechanic?

[nyc94]

Back to 9th in PWR after the Q win.

And beating Princeton drops us into a 3-way tie for 11th. Never go to New Jersey.

I thought RPI didn't drop if you beat a bad team.  Did the other teams around us win, or am I misunderstanding that mechanic?

Northeastern beat Maine and North Dakota beat Duluth to raise their RPI above ours.

[Jim Hyla]

So here they are, and since there are no games till next weekend, they shouldn't change.:-D

CHN's

Code Select Expand


Rk Team        PCWs RPI Rk preRPI QWB-† W-L-T Win % Wgtd Win % - ‡
1 Boston College 58 .6106 1 .6015 .0091 25-4-4 .8182 .8344
2 Minnesota 57 .5998 2 .5917 .0081 21-4-5 .7833 .7750
[b][u]3 Union        55 .5789 3 .5736 .0053 22-6-4 .7500 .7516[/u][/b]
4 Ferris State 55 .5697 4 .5689 .0008 23-7-3 .7424 .7610
5 St. Cloud State 54 .5637 5 .5595 .0042 18-7-5 .6833 .6774
6 Wisconsin 53 .5598 6 .5552 .0046 19-9-2 .6667 .6538
[b][u]7 Quinnipiac 52 .5532 8 .5477 .0055 21-8-5 .6912 .6818[/u][/b]
8 Mass.-Lowell 51 .5533 7 .5504 .0030 20-8-4 .6875 .6815
9 Northeastern 50 .5503 9 .5441 .0062 18-10-4 .6250 .6265
10 North Dakota 49 .5494 10 .5485 .0009 18-9-3 .6500 .6395
[b][u]11 Cornell 47 .5489 11 .5465 .0024 14-7-5 .6346 .6429[/u][/b]
12 Vermont 47 .5428 12 .5396 .0032 17-11-3 .5968 .5949
[b][u]13 Colgate 47 .5411 13 .5329 .0082 16-11-5 .5781 .5732[/u][/b]
14 Michigan 44 .5398 14 .5358 .0040 15-10-3 .5893 .5970
15 Notre Dame 44 .5393 15 .5354 .0039 19-12-2 .6061 .5817
16 Providence 42 .5389 16 .5348 .0042 17-9-6 .6250 .6132


and USCHO's.

Code Select Expand


Men's Division I PairWise Rankings

Rank Team       PWR W-L-T Win % Win %   RPI RPI Rank
                                                Rank
1 Boston College 58 25-4-4 .8182 1 .6106* 1
2 Minnesota 57 21-4-5 .7833 2 .5998 2
[b][u]3t Union        55 22-6-4 .7500 3 .5789* 3[/u][/b]
3t Ferris State 55 23-7-3 .7424 4 .5697* 4
5 St. Cloud State 54 18-7-5 .6833 7 .5637* 5
6 Wisconsin 53 19-9-2 .6667 8 .5598* 6
[b][u]7 Quinnipiac 52 21-8-5 .6912 5 .5532 8[/u][/b]
8 Mass-Lowell 51 20-8-4 .6875 6 .5533 7
9 Northeastern 50 18-10-4 .6250 11t .5503* 9
10 North Dakota 49 18-9-3 .6500 9 .5494 10
[b][u]11t Cornell        47 14-7-5 .6346 10 .5489 11[/u][/b]
11t Vermont        47 17-11-3 .5968 16 .5428 12
[b][u]11t Colgate        47 16-11-5 .5781 22 .5411 13[/u][/b]
14t Michigan 44 15-10-3 .5893 19t .5398 14
14t Notre Dame 44 19-12-2 .6061 15 .5393 15
16 Providence 42 17-9-6 .6250 11t .5389* 16

Read more: http://www.uscho.com/rankings/pairwise-rankings/d-i-men/#ixzz2u9Z8A7X5

Since they have all agreed on the system, and I like CHN's presentation better, I'll only post CHN's from now on.

Code Select Expand


Women's Division I PairWise Rankings

Rank Team        PWR W-L-T Win % Win %   RPI RPI Rank
                                                Rank
1 Minnesota 15 32-1-1 .9559 1 .6778* 1
2 Wisconsin 14 25-5-2 .8125 2 .6128* 2
[b][u]3 Clarkson 13 25-4-5 .8088 3 .6001* 3
4 Harvard        12 21-4-4 .7931 5 .5932* 4[/u][/b]
5 Boston College 11 24-5-3 .7969 4 .5885* 6
[b][u]6 Cornell        10 20-5-4 .7586 6 .5901* 5[/u][/b]
7 Robert Morris 9 24-7-3 .7500 7 .5647* 7
8 Mercyhurst 8 22-7-4 .7273 8 .5624* 8
[b][u]9 Quinnipiac 7 20-5-9 .7206 9 .5594* 9[/u][/b]
10 North Dakota 6 17-11-4 .5938 12 .5477 10
11 Boston Univ 5 21-11-1 .6515 11 .5398 11
12t Northeastern 2 17-13-2 .5625 14 .5179 12
[b][u]12t Princeton 2 14-11-4 .5517 16 .5177 13[/u][/b]
12t Minnesota-Duluth2 13-13-6 .5000 18 .5176 14
12t Syracuse 2 18-13-3 .5735 13 .5138 15
12t Ohio State 2 14-15-5 .4853 19 .5081 16

Read more: http://www.uscho.com/rankings/pairwise-rankings/d-i-women/#ixzz2u9fJVezf



[Jeff Hopkins '82]

Back to 9th in PWR after the Q win.

And beating Princeton drops us into a 3-way tie for 11th. Never go to New Jersey.

I thought RPI didn't drop if you beat a bad team.  Did the other teams around us win, or am I misunderstanding that mechanic?

Northeastern beat Maine and North Dakota beat Duluth to raise their RPI above ours.

And Q losing to Colgate dropped the bonus points we got for beating them.

[nyc94]

Back to 9th in PWR after the Q win.

And beating Princeton drops us into a 3-way tie for 11th. Never go to New Jersey.

I thought RPI didn't drop if you beat a bad team.  Did the other teams around us win, or am I misunderstanding that mechanic?

Northeastern beat Maine and North Dakota beat Duluth to raise their RPI above ours.

And Q losing to Colgate dropped the bonus points we got for beating them.

I went to CHN to see what the effect would be if Quinnipiac had won.  We would still be 11th and Quinnipiac would 6 instead of 7.  But our RPI actually goes down from .5489 to .5481.  If I am reading it correctly our unmodified RPI falls even more but our bonus is larger.

[Dafatone]

Back to 9th in PWR after the Q win.

And beating Princeton drops us into a 3-way tie for 11th. Never go to New Jersey.

I thought RPI didn't drop if you beat a bad team.  Did the other teams around us win, or am I misunderstanding that mechanic?

Northeastern beat Maine and North Dakota beat Duluth to raise their RPI above ours.

And Q losing to Colgate dropped the bonus points we got for beating them.

I went to CHN to see what the effect would be if Quinnipiac had won.  We would still be 11th and Quinnipiac would 6 instead of 7.  But our RPI actually goes down from .5489 to .5481.  If I am reading it correctly our unmodified RPI falls even more but our bonus is larger.

I wasn't sure what to make of the Q-Gate game RPIwise.  I mean, having played both teams twice, either one winning (or losing) would impact our RPI, right?

RPI is, I think (and a quick search backs this up) just a function of winning percentage + opponents' win percentage + opponents' opponents' win percentage, with each component having a different coefficient (and something in place to keep bad wins from hurting you).  Would it really matter at all which of Q and Gate won, or that we went 1-1 against Q and 0-2 against Gate, as far as RPI is concerned?

[Jim Hyla]

It's not a poll, but anyone want to see why shootout wins are stupid, just look at the NCHC standings for the top two teams.

Code Select Expand

                W L T SW Win % Pts. Overall: W L T Win %
North Dakota 13 7 0 0 .650 39 18 9 3 .650
St. Cloud State 12 5 3 0 .675 39 18 7 5 .683

Read more: http://www.uscho.com/conference/nchc/#ixzz2uF3MmdIP

SCSU gets second because they have 3 ties and ND doesn't. They have the "better record" but lose because they lost the shootouts. They also won twice at ND. Fortunately first place can be determined on the ice, when they play ND at home this weekend.

[Jeff Hopkins '82]

Back to 9th in PWR after the Q win.

And beating Princeton drops us into a 3-way tie for 11th. Never go to New Jersey.

I thought RPI didn't drop if you beat a bad team.  Did the other teams around us win, or am I misunderstanding that mechanic?

Northeastern beat Maine and North Dakota beat Duluth to raise their RPI above ours.

And Q losing to Colgate dropped the bonus points we got for beating them.

I went to CHN to see what the effect would be if Quinnipiac had won.  We would still be 11th and Quinnipiac would 6 instead of 7.  But our RPI actually goes down from .5489 to .5481.  If I am reading it correctly our unmodified RPI falls even more but our bonus is larger.

I wasn't sure what to make of the Q-Gate game RPIwise.  I mean, having played both teams twice, either one winning (or losing) would impact our RPI, right?

RPI is, I think (and a quick search backs this up) just a function of winning percentage + opponents' win percentage + opponents' opponents' win percentage, with each component having a different coefficient (and something in place to keep bad wins from hurting you).  Would it really matter at all which of Q and Gate won, or that we went 1-1 against Q and 0-2 against Gate, as far as RPI is concerned?

My impression (not backed by math) is that RPI is all about your record plus your in-conference opponents' out-of-conference play.  In conference, the opponents' record will tend toward 0.500.  OOC record will shift that above or below 0.500.  So if your conference is doing well against other conferences, their OOC won-lost record will be higher and therefore, the overall record will be higher.

Again, not backed by math.  Just my instinct.

[Trotsky]

I've always worried about the "can't be punished for a win" fudge creating different outcomes strictly based on game order.  Say for instance Yale and Harvard play the same schedule with exactly the same results (their h2hs are ties) except Yale plays UAH (1-31-1) the first game of the season and Harvard plays them the last. Does the fudge factor mean that Harvard will have a better Modified RPI because they aren't allowed to drop, while Yale doesn't get the benefit because, having played UAH first, their proportionately smaller bump is not counted as a "drop"?

(I hate the "can't be 'punished'" epicycle anyway, since it rests on an emotional misperception of "punishment".  I'd much prefer building the impossibility of losing value by winning or gaining value by losing into a continuous function rather than creating an artificial stepwise "floor." )

[Trotsky]

Cornell is 11 in the USCHO poll:

Code Select Expand


[u]US PW Team[/u]
03 03 Union
08 07 Quinnipiac
11 11 [color=#D61616]Cornell[/color]
15 18 Yale
16 13 Colgate
24 24 Clarkson
-- 25 RPI
-- 31 SLU
-- 32 Brown
-- 40 Harvard
-- 45 Dartmouth
-- 55 Princeton


[Jim Hyla]

I've always worried about the "can't be punished for a win" fudge creating different outcomes strictly based on game order.  Say for instance Yale and Harvard play the same schedule with exactly the same results (their h2hs are ties) except Yale plays UAH (1-31-1) the first game of the season and Harvard plays them the last. Does the fudge factor mean that Harvard will have a better Modified RPI because they aren't allowed to drop, while Yale doesn't get the benefit because, having played UAH first, their proportionately smaller bump is not counted as a "drop"?

(I hate the "can't be 'punished'" epicycle anyway, since it rests on an emotional misperception of "punishment".  I'd much prefer building the impossibility of losing value by winning or gaining value by losing into a continuous function rather than creating an artificial stepwise "floor." )

I would hope that the system would come out the same for both of them.  The only way for that to happen would be to figure out the RPI without UAH in the schedule for both teams, regardless of when they played them.

[ugarte]

I've always worried about the "can't be punished for a win" fudge creating different outcomes strictly based on game order.  Say for instance Yale and Harvard play the same schedule with exactly the same results (their h2hs are ties) except Yale plays UAH (1-31-1) the first game of the season and Harvard plays them the last. Does the fudge factor mean that Harvard will have a better Modified RPI because they aren't allowed to drop, while Yale doesn't get the benefit because, having played UAH first, their proportionately smaller bump is not counted as a "drop"?

(I hate the "can't be 'punished'" epicycle anyway, since it rests on an emotional misperception of "punishment".  I'd much prefer building the impossibility of losing value by winning or gaining value by losing into a continuous function rather than creating an artificial stepwise "floor." )

I would hope that the system would come out the same for both of them.  The only way for that to happen would be to figure out the RPI without UAH in the schedule for both teams, regardless of when they played them.

Pretty sure what it means is that midseason RPI is a work-in-progress. While the games might have a different affect on RPI at the time of the game, they will have the exact same weight in the final result. If you play a team that is 5-0 and they finish the season 5-25, your final RPI treats the game as one against a 5-25 team. Basically you aren't punished for having played UAH in your hypothetical, regardless of when the game was.

This has to be right, right?

[KeithK]

I've always worried about the "can't be punished for a win" fudge creating different outcomes strictly based on game order.  Say for instance Yale and Harvard play the same schedule with exactly the same results (their h2hs are ties) except Yale plays UAH (1-31-1) the first game of the season and Harvard plays them the last. Does the fudge factor mean that Harvard will have a better Modified RPI because they aren't allowed to drop, while Yale doesn't get the benefit because, having played UAH first, their proportionately smaller bump is not counted as a "drop"?

(I hate the "can't be 'punished'" epicycle anyway, since it rests on an emotional misperception of "punishment".  I'd much prefer building the impossibility of losing value by winning or gaining value by losing into a continuous function rather than creating an artificial stepwise "floor." )

I would hope that the system would come out the same for both of them.  The only way for that to happen would be to figure out the RPI without UAH in the schedule for both teams, regardless of when they played them.

Pretty sure what it means is that midseason RPI is a work-in-progress. While the games might have a different affect on RPI at the time of the game, they will have the exact same weight in the final result. If you play a team that is 5-0 and they finish the season 5-25, your final RPI treats the game as one against a 5-25 team. Basically you aren't punished for having played UAH in your hypothetical, regardless of when the game was.

This has to be right, right?

I'm pretty sure this is right because the selection committee isn't tracking RPI (or PWR) throughout the season the way we do.  They look at the results after the dust has settled on March 23. Hopefully the various sites that calculate the metrics during the season use the same approach: calculating ratings using the entire season to date to get today's rankng.

The fact that there's a percieved need for a fudge factor simply points to the limits of RPI as a ranking system.  But we've covered that ground many, many times through the years.