KRACH Computer Rankings - Updated 03/10 FINAL

The Latest 400 or so Topics

Moderators: Mitch Hawker, east hockey, karl(east)

SuperStar
Posts: 1285
Joined: Thu Oct 04, 2007 10:26 am

Post by SuperStar » Fri Dec 14, 2012 6:39 pm

Most teams have played 5/6 games. The season is only 23 games total. what do you mean there is "so little data" in 2/3 weeks the season is half over for most teams..

So when where the big "enough data" to make an intelligent decision? Thats what I don"t like about computer generated stats with such a short season.

If teams played 50, 60 games I could see the relevance.

MNHockeyFan
Posts: 6958
Joined: Tue Nov 07, 2006 10:28 pm

Post by MNHockeyFan » Fri Dec 14, 2012 7:58 pm

SuperStar wrote:Most teams have played 5/6 games. The season is only 23 games total. what do you mean there is "so little data" in 2/3 weeks the season is half over for most teams..

So when where the big "enough data" to make an intelligent decision? Thats what I don"t like about computer generated stats with such a short season.

If teams played 50, 60 games I could see the relevance.
Although more games is better, you don't need 50 or 60 of them before they become predictive. I would say early January, after all the Holiday tournaments, is when they'll really start to become meaningful.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Post by ghshockeyfan » Fri Dec 14, 2012 8:05 pm

I believe the season is 25 games plus sections and state. Maybe 30+ games for some teams that make it to state.

When are there enough data? Maybe never in the eyes of some. The disclaimer is there for now. I'll probably drop it in a week or so.

I appreciate that some don't like computer rankings for a variety of reasons. I like to see how they compare to the human LPH rankings and others out there.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Updated 12/15 AM 5537

Post by ghshockeyfan » Sat Dec 15, 2012 1:51 pm

Updated 12/15 AM
Last edited by ghshockeyfan on Sun Dec 16, 2012 10:52 am, edited 1 time in total.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Updated 12/16 5765

Post by ghshockeyfan » Sun Dec 16, 2012 10:52 am

Updated 12/16
Last edited by ghshockeyfan on Tue Dec 18, 2012 10:59 am, edited 1 time in total.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Re: Updated 12/18 AM 6282

Post by ghshockeyfan » Tue Dec 18, 2012 10:58 am

Updated 12/18 AM
Last edited by ghshockeyfan on Wed Dec 19, 2012 7:24 pm, edited 1 time in total.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Updated 12/19 AM 6745

Post by ghshockeyfan » Wed Dec 19, 2012 7:24 pm

Updated 12/19 AM
Last edited by ghshockeyfan on Fri Dec 21, 2012 9:07 am, edited 1 time in total.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Updated 12/21 AM 7061

Post by ghshockeyfan » Fri Dec 21, 2012 9:06 am

Updated 12/21 AM
Last edited by ghshockeyfan on Sat Dec 22, 2012 10:04 am, edited 1 time in total.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Updated 12/22 AM 7393

Post by ghshockeyfan » Sat Dec 22, 2012 10:03 am

Updated 12/22 AM
Last edited by ghshockeyfan on Sun Dec 23, 2012 3:07 pm, edited 1 time in total.

MNHockeyFan
Posts: 6958
Joined: Tue Nov 07, 2006 10:28 pm

Post by MNHockeyFan » Sat Dec 22, 2012 11:12 am

Still some very good teams on the combined list that are way down, Eden Prairie and Andover to name two. Hard to believe there are 50 teams in the state that are better, but based on the (still limited) data that's what the computer's saying!

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Updated 12/23 7793

Post by ghshockeyfan » Sun Dec 23, 2012 3:06 pm

Updated 12/23
Last edited by ghshockeyfan on Sat Dec 29, 2012 4:32 am, edited 1 time in total.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Updated 12/29 AM 8411

Post by ghshockeyfan » Sat Dec 29, 2012 4:32 am

Updated 12/29 AM
Last edited by ghshockeyfan on Sat Dec 29, 2012 10:00 am, edited 1 time in total.

Goalie-Dad
Posts: 105
Joined: Thu Oct 13, 2011 3:53 pm

Re: Updated 12/29 AM 8411

Post by Goalie-Dad » Sat Dec 29, 2012 7:11 am

ghshockeyfan wrote:Updated 12/29 AM

Most Recent Ranking Links:
==========================================
OVERALL:
http://www.bgoski.com/b/KRACH_OA.htm
CLASS AA:
http://www.bgoski.com/b/KRACH_AA.htm
CLASS A:
http://www.bgoski.com/b/KRACH_A.htm
SECTION:
http://www.bgoski.com/b/KRACH_SEC.htm
SOS (Strength-Of-Schedule) is at:
http://www.bgoski.com/b/KRACH_SOS.htm
==========================================

Past rankings are at:
http://www.ushsho.com/forums/viewtopic. ... sc&start=0
Please check your data. Mounds View is listed in KRACH with a 5-5 record. The accurate record is 7-1-2 (as per MN Hockey Hub).

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Re: Updated 12/29 AM 8411

Post by ghshockeyfan » Sat Dec 29, 2012 9:06 am

Goalie-Dad wrote:Please check your data. Mounds View is listed in KRACH with a 5-5 record. The accurate record is 7-1-2 (as per MN Hockey Hub).
Good catch! - Mound-Westonka (5-5) and Mounds View (7-1-2) records were inverted - refresh coming soon...

KRACH records will sometimes differ by a game or two if Hub was not up-to-date when data pulled from web and/or due to non-MSHSL opponents not counting for KRACH. In this case though, there was a legit issue with the records being inverted.

Also - the "refresh" will have rankings now for "Monticello" (aka Monticello/Annandale/Maple Lake) and "LSH/SP/TCU" (aka Le Sueur-Henderson/St. Peter/Tri-City United) as these were unranked before due to naming not aligning with Hub results.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Updated 12/29 10AM 8678

Post by ghshockeyfan » Sat Dec 29, 2012 9:59 am

Updated 12/29 10AM - for Mounds View/Mound-Westonka & Monticello/LSH/SP/TCU
Last edited by ghshockeyfan on Sun Dec 30, 2012 12:34 pm, edited 1 time in total.

HShockeywatcher
Posts: 6848
Joined: Thu Aug 31, 2006 10:21 pm

Post by HShockeywatcher » Sat Dec 29, 2012 10:51 am

Does playing fewer games have a negative impact on teams?
EGF is 10-0-1 right now, but have played [I'm assuming] 5 games against ND teams so they are 5-0-1 for the computer. Would they have a higher rating if those 5 wins were able to be factored in?

MNHockeyFan
Posts: 6958
Joined: Tue Nov 07, 2006 10:28 pm

Post by MNHockeyFan » Sat Dec 29, 2012 11:04 am

HShockeywatcher wrote:Would they have a higher rating if those 5 wins were able to be factored in?
ghshockeyfan will probably weigh in on this as well, but I would venture to say that, no, they would not have a higher rating because the computer model does not recognize teams from North Dakota, i.e. it cannot determine the relative strength of those teams as compared to Minnesota teams so it cannot factor in those results.

gardetto
Posts: 33
Joined: Thu Feb 16, 2012 6:09 pm

Post by gardetto » Sat Dec 29, 2012 11:04 am

HShockeywatcher wrote:Does playing fewer games have a negative impact on teams?
EGF is 10-0-1 right now, but have played [I'm assuming] 5 games against ND teams so they are 5-0-1 for the computer. Would they have a higher rating if those 5 wins were able to be factored in?
They only count MSHSL games. The ND games do not count

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Post by ghshockeyfan » Sat Dec 29, 2012 11:06 am

HShockeywatcher wrote:Does playing fewer games have a negative impact on teams?
EGF is 10-0-1 right now, but have played [I'm assuming] 5 games against ND teams so they are 5-0-1 for the computer. Would they have a higher rating if those 5 wins were able to be factored in?
Not unless the team plays significantly better in the games out-of-state. Meaning, are they getting similar/expected results both in the MN games and also the ND (non-MN) ones? If the team is playing consistently (beating the opponents they should), then there shouldn't be much impact.

PS - you are right on EGF. I ran a manual sanity check of ALL teams and found exactly what you described. EGF has about 5 games against non-MN teams. There are a select few others like this on the state borders - but most teams have 1-2 games at most, if any, against non-MN MSHSL teams.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Post by ghshockeyfan » Sat Dec 29, 2012 11:07 am

gardetto wrote:
HShockeywatcher wrote:Does playing fewer games have a negative impact on teams?
EGF is 10-0-1 right now, but have played [I'm assuming] 5 games against ND teams so they are 5-0-1 for the computer. Would they have a higher rating if those 5 wins were able to be factored in?
They only count MSHSL games. The ND games do not count
Correct!

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Post by ghshockeyfan » Sat Dec 29, 2012 11:13 am

MNHockeyFan wrote:
HShockeywatcher wrote:Would they have a higher rating if those 5 wins were able to be factored in?
ghshockeyfan will probably weigh in on this as well, but I would venture to say that, no, they would not have a higher rating because the computer model does not recognize teams from North Dakota, i.e. it cannot determine the relative strength of those teams as compared to Minnesota teams so it cannot factor in those results.
This is correct - note that I believe Mitch's rankings, and some others, may take into account non-MN/MSHSL games but my KRACH rankings do not.

I'll be interested to see how the KRACH compares to Mitch's rankings and others as the year goes on... I'm not saying KRACH is superior by any means to these other rankings - it's more that I'm just curious to see how they differ based on what they use as inputs, algorithms, etc.

I'm already pretty well convinced that Mitch's rankings will do a better job of predicting future results than KRACH. I do believe KRACH is a pretty accurate representation of past performance as discussed in earlier posts.

So - not sure that I'd go to Vegas and bet KRACH, but I would use KRACH to seed a tourney, for example, or for comparison to LPH polls.

HShockeywatcher
Posts: 6848
Joined: Thu Aug 31, 2006 10:21 pm

Post by HShockeywatcher » Sat Dec 29, 2012 11:15 am

gardetto wrote:
HShockeywatcher wrote:Does playing fewer games have a negative impact on teams?
EGF is 10-0-1 right now, but have played [I'm assuming] 5 games against ND teams so they are 5-0-1 for the computer. Would they have a higher rating if those 5 wins were able to be factored in?
They only count MSHSL games. The ND games do not count
I know. That's why I asked the question...considering that 8 (32%) of their games are against ND teams, it could be a huge factor.

Red River, who they play tonight and two more times this season, is obviously halfway decent. Not sure how they'd be ranked were they a MSHSL team, but 3 wins against a good team could be a huge factor in almost any ranking system.

Not a huge deal, and will only significantly affect 1 team in the state.

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Post by ghshockeyfan » Sat Dec 29, 2012 11:19 am

HShockeywatcher wrote:
gardetto wrote:
HShockeywatcher wrote:Does playing fewer games have a negative impact on teams?
EGF is 10-0-1 right now, but have played [I'm assuming] 5 games against ND teams so they are 5-0-1 for the computer. Would they have a higher rating if those 5 wins were able to be factored in?
They only count MSHSL games. The ND games do not count
I know. That's why I asked the question...considering that 8 (32%) of their games are against ND teams, it could be a huge factor.

Red River, who they play tonight and two more times this season, is obviously halfway decent. Not sure how they'd be ranked were they a MSHSL team, but 3 wins against a good team could be a huge factor in almost any ranking system.

Not a huge deal, and will only significantly affect 1 team in the state.
I believe some of the other ranking systems could do a ranking of Red River with MN teams for comparison. Some rankings may actually use the non-MN game results (like Red River, ND/WI teams, etc.) for additional data but just don't publish the non-MN teams in MN rankings.

alcloseshaver
Posts: 1473
Joined: Mon Dec 17, 2012 9:29 am
Location: Met Center Press Box

Post by alcloseshaver » Sat Dec 29, 2012 6:03 pm

I need someone to explain these rankings to me? How can teams like EGF, LF, Tech, and even Princeton be rated ahead of bigger AA schools from the metro? Teams like Rapids and Elk River are 24th and 37th ???
All these issues will play out on the ice as the season goes on but it seems the LPH and Karl's rankings are more accurate. What kind of formula is this?? What rankings formula is used to help the different sections do their seeding? Based on these rankings does it mean if Princeton and ER were in the same section princeton would be a higher seed?

ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Post by ghshockeyfan » Sat Dec 29, 2012 6:27 pm

alcloseshaver wrote:I need someone to explain these rankings to me? How can teams like EGF, LF, Tech, and even Princeton be rated ahead of bigger AA schools from the metro? Teams like Rapids and Elk River are 24th and 37th ???
All these issues will play out on the ice as the season goes on but it seems the LPH and Karl's rankings are more accurate. What kind of formula is this?? What rankings formula is used to help the different sections do their seeding? Based on these rankings does it mean if Princeton and ER were in the same section princeton would be a higher seed?
KRACH isn't used for any section or state tourney seedings, so no worries there. I'm not sure that there is any official formula or computer ranking used for section seedings. I think it is still the old coaches' vote approach. KRACH is explained online and to some degree earlier in this thread. The bottom line is winning and the strength of schedule against which those wins were achieved. Human based rankings have other factors considered, for better or for worse. This is what makes interesting the comparison of any computer ranking to that of the human nature. What is it that those humans "know" that wins and losses alone can't tell us? Is there justifiable bias in the human rankings? Food for thought, I guess. And, lastly, I agree 100% that this will all be settled on the ice as it should. The game is not played on paper, by computers, on forums, but instead on the ice. Thank goodness for that!

Post Reply