Official PWR

[DeltaOne81]

Not that it would be any different from JTW's (well, only a tiny bit), but USCHO finally officially released their PWR: http://uscho.com/polls/?data=pwr1m, they have a 3 way tie for 2nd with us leading UND and CC on the RPI tie-break.

Edit: How's this for big news - http://uscho.com/polls/?data=krach

[jtwcornell91]

Since the NCAA doesn't distribute them themselves, nobody's pairwise comparisons are more or less official than anybody else's.  (Does anybody say "hey, the official ECAC standings are out"?)

On the other hand, since everyone gets their game results from USCHO, and the USCHO PWR is calculated from the latest game results in their database, the one posted there will at times be more up-to-date than others.  (E.g., after the first game result of the day/night lands in the database and before my overnight cron job runs.)

[DeltaOne81]

Get technical with me and I'll get technical back :-P

While you're right that it isn't official via the NCAA, but at the same time USCHO did invent the statistical measure used to mimic the selection committee and named it PWR. So it kinda is "theirs."

While with standings, the exactly math is well defined, your PWR scheme is slightly different from USCHO. Yours breaks a tie on a comparision by comparision basis, breaking with with RPI. USCHO will give that comparison to nobody, and break the overall ties with RPI. I could certainly forsee a situation in which that could make a difference in the overall standings, so I was curious if USCHO's method would come out the same.

Of course I guess the next question is who's is closer to the committee's method. Though yours certainly makes the prettier graph :-).

And nothing takes away from your real victory of today, you being a major reason why KRACH is officially listed on USCHO - and getting a link to your site from their KRACH page.

-Fred

[jtwcornell91]

DeltaOne81 '03 wrote:

While you're right that it isn't official via the NCAA, but at the same time USCHO did invent the statistical measure used to mimic the selection committee and named it PWR. So it kinda is "theirs."

Actually, Keith Instone invented the concept of totalling up the number of comparisons; USCHO (in the person of Time Brule) came up with the name PWR:
http://lists.maine.edu/cgi/wa?A2=ind9803&L=hockey-l&D=0&F=P&P=20043&F=

While with standings, the exactly math is well defined, your PWR scheme is slightly different from USCHO. Yours breaks a tie on a comparision by comparision basis, breaking with with RPI. USCHO will give that comparison to nobody, and break the overall ties with RPI. I could certainly forsee a situation in which that could make a difference in the overall standings, so I was curious if USCHO's method would come out the same.

Well, my ECAC standings page breaks ties by the ECAC tiebreakers rather than alphabetically, but it's still the ECAC standings either way.

[Greg Berge]

What are the ECAC tiebreakers now?  We used to have privileged categories like "record against playoff teams," but now that's meaningless.

Oh, and is infinite regress still possible?  ;-)

[jtwcornell91]

I can explain the ECAC tiebreakers when I get back to New Orleans tonight.  The way out of the infinite loop was explained a couple of years ago.  The only change for this year is that "top5" and "top10" have gone back to "top4" and "top8".