PairWise Rankings

[ninian '72]

PairWise Rankings are up at uscho, along with KRACH AND KASA alternatives:

http://www.uscho.com/

Of immediate interest PWR has Cornell tied for 13th with Mass-Lowell and Vermont, Harvard tied for 8th with NoDak and OSU.  KRACH and KASA both have the Red in 13th, Harvard in 10th.  No other ECACHL team is ranked above Cornell in any of these systems.  

Saturday's game looms ever larger.

[KeithK]

[q]Of immediate interest PWR has Cornell tied for 13th with Mass-Lowell[/q]So we're tied with Lowell.  Has anyone noticed Lowell's record?  Dead last in HE at 0-5-2.  But still in the PWR rankings (for the moment) on the basis of a perfect 9-0-0 non-conference record, including 6-0 against ECAC teams.  You know, this is the kind of thing that hurts the ECAC.  Not one of these six ECAC teams (Union, RPI, Clarkson, SLU, Colgate, Dartmouth) could manage to beat the HE doormat?  Granted four were one goal games (five if you ignore ENG) so some of this can be attributed to randomness, but please!  Hopefully Brown can break the string against them on the 18th.

BTW - I guarantee that UML will tank in the PWR long before March if they don't put on a conference run.

[KenP]

 ::help::  Can someone explain to me how KASA differs from KRACH?  I understand how KRACH works, but I don't understand how KASA incorporates and uses the home-vs-away differences.

[jtwcornell91]

[Q]KenP Wrote:
Can someone explain to me how KASA differs from KRACH?  I understand how KRACH works, but I don't understand how KASA incorporates and uses the home-vs-away differences.
[/q]

Each team effectively has three ratings: home, away, and neutral ice.  The factor by which the home ratings exceed the neutral ice ratings, and the neutral ice ratings exceed the away ratings, is assumed to be the same for each team, and this one factor is treated as an additional unknown parameter.  Instead of just fitting the ratings of each team by requiring that each team have the same expected and actual winning percentage, you fit all those along with the home ice factor by requiring expected=actual winning percentage for each team, and also that the overall winning percentage for home teams equals the value you'd predict from the ratings.

[KenP]

What conclusions can you draw comparing the two?  For example, Wisconsin's KRACH is 467.3 and their KASA is 381.0.  Does that mean KASA is attributing more of their success to home ice and less to the team?  Similarly, does Cornell's (slightly) higher KASA rating "credit" them for playing good teams like MSU and Vermont on away ice?

[jtwcornell91]

[Q]KenP Wrote:

 What conclusions can you draw comparing the two?  For example, Wisconsin's KRACH is 467.3 and their KASA is 381.0.  Does that mean KASA is attributing more of their success to home ice and less to the team?  Similarly, does Cornell's (slightly) higher KASA rating "credit" them for playing good teams like MSU and Vermont on away ice?
[/q]

Because the normalization is arbitrary, you can't really compare the ratings themselves; instead you should compare the RRWPs for an apples-to-apples view of the situation.

A team with a higher KASA RRWP probably played more important games (i.e., games against comparably-ranked opponents) on the road, so their schedule is tougher according to KASA than to KRACH.