fairbanks hockey puck
Registered User
Re: Pairwise and Bracketology 2013 Edition
thanks!!! for the info...
thanks!!! for the info...
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.
- Status
Re: Pairwise and Bracketology 2013 Edition
Do you mean a full table of all combinations for our perusal, or a (win2, win1, win0 or win3) like you normally do with a few teams, except this time for everybody?
As always, I'll also post the full table of all remaining possibilities tomorrow morning; I don't think I'll do it tonight with a game at 7MT. Plus, by morning you guys will have covered enough of the range with the YATC calculators to give me some confidence in the results
Do you mean a full table of all combinations for our perusal, or a (win2, win1, win0 or win3) like you normally do with a few teams, except this time for everybody?
Re: Pairwise and Bracketology 2013 Edition
PL,
Would you mind telling me which 3 comparisons you can flip? Just looking at the current grid, I am only seeing the possibility of Brown is Brown loses today, and Saint Lawrence by a strange change in RPI of .0003 cumulatively for each team.
No time to look for me for awhile. I have a busy afternoon with work.
Also, Priceless, As you know, you don't need the YATC calculators. I think it's quicker to use DIY by Whelan, because you can change the results more quickly. The only caveat in that is that you have to remember that his DIY breaks ties by individual comparisons, rather than by RPI.
PL,
Would you mind telling me which 3 comparisons you can flip? Just looking at the current grid, I am only seeing the possibility of Brown is Brown loses today, and Saint Lawrence by a strange change in RPI of .0003 cumulatively for each team.
No time to look for me for awhile. I have a busy afternoon with work.
Also, Priceless, As you know, you don't need the YATC calculators. I think it's quicker to use DIY by Whelan, because you can change the results more quickly. The only caveat in that is that you have to remember that his DIY breaks ties by individual comparisons, rather than by RPI.
Last edited:
Re: Pairwise and Bracketology 2013 Edition
Well, at a minimum I can certainly give you the type of stuff I've done in past years, e.g. <a href="http://blog.siouxsports.com/2012/03/16/friday-night-final-pwr-outlook/">Outlook as of Friday Mar.16, 2012</a>. Of course, I'm always making changes and am open to suggestion. By full table of combinations, you mean all 1,000,000 possible outcomes in a giant table, or something else?
Well, at a minimum I can certainly give you the type of stuff I've done in past years, e.g. <a href="http://blog.siouxsports.com/2012/03/16/friday-night-final-pwr-outlook/">Outlook as of Friday Mar.16, 2012</a>. Of course, I'm always making changes and am open to suggestion. By full table of combinations, you mean all 1,000,000 possible outcomes in a giant table, or something else?
Patman
Rodent of Unusual Size
Re: Pairwise and Bracketology 2013 Edition
Sure, all 1 million outcomes
The last time I did the pairwise it took about 4 seconds on average to augment and compute. This time I did it correctly and more efficiently... thought I haven't done a timing yet (laptop is also much faster than before). I don't know what you're getting Jim, but the way I see it there's no hope in doing that many outcomes without the ability to compute in parallel. I can't do it, as I don't have a server and loads of processors to abuse.
Edit: I've started a timing run but I'm headed out to kickball and I'd really like to restart the lappy as there is a fair amount of lag
Sure, all 1 million outcomes
The last time I did the pairwise it took about 4 seconds on average to augment and compute. This time I did it correctly and more efficiently... thought I haven't done a timing yet (laptop is also much faster than before). I don't know what you're getting Jim, but the way I see it there's no hope in doing that many outcomes without the ability to compute in parallel. I can't do it, as I don't have a server and loads of processors to abuse.Edit: I've started a timing run but I'm headed out to kickball and I'd really like to restart the lappy as there is a fair amount of lag
Last edited:
Re: Pairwise and Bracketology 2013 Edition
I am not sure it's a million. I think this calculation was done earlier, and with only ECAC having a 3rd place game it comes out to 2^17 * 3, which I think was about 400,000.
But, still. Are you actually computing all of those? And, if so, are you using 50/50 odds, or KRACH odds?
And, finally, a simple table like the one you have, with the number of victories for the weekend, is great. In the case of the 4 ECAC teams, separated into w/l and l/w if it makes a difference as well would be nice. But, not required. Thanks.
I am not sure it's a million. I think this calculation was done earlier, and with only ECAC having a 3rd place game it comes out to 2^17 * 3, which I think was about 400,000.
But, still. Are you actually computing all of those? And, if so, are you using 50/50 odds, or KRACH odds?
And, finally, a simple table like the one you have, with the number of victories for the weekend, is great. In the case of the 4 ECAC teams, separated into w/l and l/w if it makes a difference as well would be nice. But, not required. Thanks.
P
Priceless
Guest
The NCAA has never clarified how they break ties, and I reposted one of last year's examples of how the committee used two different methods to break ties in 2005 vs 2007.
Alaska could flip St Lawrence which it currently loses by only .0002 RPI. I'd have to check what losses would do to their RPI but the Brown and Cornell comparisons could flip with losses. However Alaska could also lose the comp wins they have with Lowell and Denver. When I'm on my PC I can be more precise.
Patman
Rodent of Unusual Size
If Jim is using C or java he might have a shot at that figure. Bottom line is there isn't a fast in Monte Carlo when you have to repeatedly manipulate data objects and pair wise isn't necessarily quick.
On the other hand, I'd love to have a lot of the stuff I've written as specialized C function calls from R. I'm rather limited in my C programming abilities, but I've seen what it can do when called upon,
Jim, what programming environ are you using?
Re: Pairwise and Bracketology 2013 Edition
You're right, I was just remembering last year's number when there was an additional 3rd place game.
But, yes, I'll do it exhaustively after tonight. I also usually switch from KRACH-weighted outcomes to pure share of remaining scenarios (i.e. 50/50) after this weekend. To answer Pat's question, the engine I use for these simulations is in Java and it is parallel.
You're right, I was just remembering last year's number when there was an additional 3rd place game.
But, yes, I'll do it exhaustively after tonight. I also usually switch from KRACH-weighted outcomes to pure share of remaining scenarios (i.e. 50/50) after this weekend. To answer Pat's question, the engine I use for these simulations is in Java and it is parallel.
Last edited:
no offense, just competitive
P
Priceless
Guest
An awful lot of teams are New Hampshire fans today. They can help teams on the bubble by eliminating Providence and potentially Lowell/BC if they win Hockey East and want to stay in Manchester. As a 4-seed UNH could send them west. The reward for winning HEA could be a trip to Grand Rapids...quite a prize.
SCSU red army
New member
Re: Pairwise and Bracketology 2013 Edition
could someone tell me what teams winning today could help improve st.clouds position in the pairwise. I don't understand how the system works really
could someone tell me what teams winning today could help improve st.clouds position in the pairwise. I don't understand how the system works really
My look at it (from a Yale perspective) has me a Friar fan? Am I looking at it wrong? I have a UNH win dropping us to 8th.
P
Priceless
Guest
You typically want teams that you played to win, as well as teams near you in the pairwise to lose. In your case, you want UNO, UNH and CC to win. You really want CC to win because that guarantees they will be a TUC.
P
Priceless
Guest
Yes and no. From a short-term perspective you'd want Providence to win. However a UNH win eliminates PC and the comparisons that are flipped for now can be flipped back easily (especially Denver). Besides, if you want to stay east it would be beneficial to be 8 overall and be in Quinnipiac's region (Providence) than where you are now - heading to Toledo.
Re: Pairwise and Bracketology 2013 Edition
Maybe. I grant that the comparisons we lose can be flipped back easily. What I'm most interested in doing is bulletproofing us against two losses in AC. (How I hate that consolation game.) A 2nd loss to QPac would be the best thing, so I'm rooting for them tonight. My feeling is not how do I get to 5 vs.8; I agree with you that anything from 5-12 is fine. I just want to avoid 14-17. On the other end, I guess one extremely outside chance is 4... which I think would require some extraordinary things to happen to BC. We can't flip our BC comparison, but we could potentially flip the NoDak comparison (but only if NoDak wins tonight) and then hope BC loses a comparison.
Maybe. I grant that the comparisons we lose can be flipped back easily. What I'm most interested in doing is bulletproofing us against two losses in AC. (How I hate that consolation game.) A 2nd loss to QPac would be the best thing, so I'm rooting for them tonight. My feeling is not how do I get to 5 vs.8; I agree with you that anything from 5-12 is fine. I just want to avoid 14-17. On the other end, I guess one extremely outside chance is 4... which I think would require some extraordinary things to happen to BC. We can't flip our BC comparison, but we could potentially flip the NoDak comparison (but only if NoDak wins tonight) and then hope BC loses a comparison.
UnionHockeyManiac
Union Hockey fan for life
Re: Pairwise and Bracketology 2013 Edition
Can Union get an at large bid if they don't win a game in Atlantic City? Does Union need to win at least one game?
Can Union get an at large bid if they don't win a game in Atlantic City? Does Union need to win at least one game?
P
Priceless
Guest
I can tell you tonight. On the train with limited access.
Just looking through current pairwise is a mess. Minnesota in Manchester, BC, Yale and Lowell all going west...ugly.
Why is Minny in Manchester when the only thing I find is UNH at 9th? I am on a mobile.
Last edited:
- Status