Using the good wins bonus script on USCHO and assigning .0040, .0020, and .0010 for road, neutral, and home wins, Cornell winds up #1 overall. :-D
Do we know that it is going to be that simple? Maybe there is a sliding scale, so that a win over #1 is worth more than a win over #15. (I certainly think that the wins over BU at Lynah are more impressive than the win in Columbus.)
Also, since the bonus points are based on RPI, can a team become a TUC because of bonus points? Hypothetically, not factually based on this year's border teams. Only UAH has any quality OOC wins among teams within .02 of a .5000 RPI - a neutral site win over Ferris State(!), so I don't think that the bonus will be large enough to help them.
> a win over #1 is worth more than a win over #15
That would be defining criterion B in terms of criterion A, which makes little methodological sense (if criterion A is so great, just use it). Not saying they won't do it, just that it would be questionable. Of course, the whole thing is questionable, anyway.
QuoteGreg wrote:
Using the good wins bonus script on USCHO and assigning .0040, .0020, and .0010 for road, neutral, and home wins, Cornell winds up #1 overall. :-D
0.0025, 0.0020, and 0.0015 will put Cornell #1 as well.
0.004, 0.003, & 0.001 will keep the top 4 as they are.
0.005, 0.003, & 0.001 will put us ahead of CC.
0.01, 0.0075, 0.005 will play havoc with the top four seeds ::nut::
What's also interesting is every bonus point scenario I tried, puts St. Cloud back in the tournament, drops Harvard a position, and knocks Michigan State out. ::screwy::
QuoteGreg wrote:
> a win over #1 is worth more than a win over #15
That would be defining criterion B in terms of criterion A
I don't take your point. The RPI is set before they hand out bonus points. Once they have chosen to give out bonus points, there are at least four broad ways of awarding points: (1) all "15 or better OOC" wins are equal; (2) all "15 or better OOC" wins are given bonus, with a sliding scale for H/N/A; (3) all "15 or better OOC" get bonus points, with a sliding scale for 1-15; (4) all "15 or better OOC" get bonus points with sliding scales for H/N/A
and 1-15. They use method 3 for the BCS.
(Sure, I hear the objection: "The BCS is not an altogether enviable model." But it might be if it were used to select a playoff field with some depth, not just to create a national championship game. Someone else can judge whether the various computer rankings the BCS uses are valid; I think we all agree that the polls the BCS uses are worthless.)
I am not a fan of the bonus points, but I don't see why "all wins are treated equally" is better than the sliding scale, when the premise of a top 15 cutoff already implies that all wins are not equal. (Of course, bonus points appear to work in Cornell's favor, Greg, and I thought that was your criteria for evaluating the merit of the system. ;-) )
Has any one come up with a scenario of weights, where we do bad? It seems like this is going to help us no matter what...
> Of course, bonus points appear to work in Cornell's favor, Greg, and I thought that was your criteria for evaluating the merit of the system. ;-)
Certainly! :-D
My point is that if you apply bonus points to the ordering of teams based on RPI and differentially weight the bonus points based on the RPI ordering itself, then you're saying contradictory things: RPI captures what you want well enough to be used as an ordering criterion, and no it doesn't.
Say by simple RPI that BC is 7th and Ferris is 8th. Say that after bonus modification the teams swap. In that case, you're saying that Ferris is actually better than BC, which means bonus points awarded for wins against Ferris should in retrospect have been worth more and not less than those for wins against BC, with similar "should haves" for other swapped teams. Ugly. Uglier if an iterative correction method where you readjust only swapped pairs swapped them *back* creating an infinite loop.
The better way would be to (cough get rid of the good wins idea cough) have the RPI ordering altered by bonus points derived from an independent source. Since they're restricting the good wins to NC games anyway, you could do an RPI-like computation based on the subset of all NC games, for instance (although the sample for a given team would be so small it might be silly). (Or, since this system is assymetric since there are no "bad losses", to base it on an ordering system like Wayne Smith's that totally ignores losses.)
QuoteMatt Scheffer '97 wrote:
Has any one come up with a scenario of weights, where we do bad? It seems like this is going to help us no matter what...
http://uscho.com/polls/pwrbonus.php?roadbon=-.004&neutbon=-.002&homebon=-.001
:-}
Nice try, but it looks like we still end up in second.
If OSU loses and drops out of the top 15 in RPI then we might get hurt by it.
Quotebig red apple wrote:
Nice try, but it looks like we still end up in second.
But our RPI drops below UHN's:
http://uscho.com/polls/rpibonus.php?roadbon=-.004&neutbon=-.002&homebon=-.001
QuoteJohn T. Whelan '91 wrote:
QuoteMatt Scheffer '97 wrote:
Has any one come up with a scenario of weights, where we do bad? It seems like this is going to help us no matter what...
http://uscho.com/polls/pwrbonus.php?roadbon=-.004&neutbon=-.002&homebon=-.001
:-}
Ooooh. Negative numbers. I never thought of trying that! ::nut::
JTW has produced for USCHO his famous "You Are The Committee" script.
http://www.uscho.com/rankings/yatc.php
One question, however - doesn't the CCHA reseed after tonight's quarterfinal games? This doesn't seem to be reflected in the possible pairings.
Unless I'm mistaken, you can't just hit "back" on your browser to change a result and ask it to recalculate PWR. You really do need to start over. I backed up to see what would happen if I changed us beating Harvard to losing to Harvard and they vaulted to #3 in PWR. And CCs RPI was like .6550
I can't wait to hear the committee explaining the use of negative numbers. ::nut::
QuoteMatt Carberry wrote:
One question, however - doesn't the CCHA reseed after tonight's quarterfinal games? This doesn't seem to be reflected in the possible pairings.
They do, and it isn't reflected. If Notre Dame advances, they'll play Ferris State, not Michigan.
QuoteBill Fenwick wrote:
QuoteMatt Carberry wrote:
One question, however - doesn't the CCHA reseed after tonight's quarterfinal games? This doesn't seem to be reflected in the possible pairings.
They do, and it isn't reflected. If Notre Dame advances, they'll play Ferris State, not Michigan.
And fortunately, it doesn't matter. Saved by the non-upset. (I have a vague memory I had to work around that last year, but had forgotten.)
That undesirable feature is no longer there; repeated "calculate PWCs" clicks no longer lead to multiple counting of games. Also, there is now a button that lets you go back even if your browser doesn't let you re-submit a post query.