2013-14 KRACH RANKINGS - NOW ON MGHCA WEBSITE

[OsMetroDad]

D16Dad wrote

Very Metrocentric... Petition to be in Section 7 or Section 8, schedule a home and away game against 5 of the teams in your new section (Tuesdays and Thursdays are fun) and take advantage of your free pass. Seems like a fair deal...

D16 - Although I live in the Metro area, I agree 100% with you. Some of the hockey families around the Metro don't appreciate the travel that our Greater-MN hockey families endure so that their kids can play hockey. My kid's HS team just recently travel to TRFs for their post-Xmas tournament, it was a 5 hour drive plus, but worth every minuete in the car to get to the best hockey venue in the state at the Little Ralph! As far as the MN HS State tournament, the point of which is to have a variety of participation, not just the "BEST 8" per class. For the teams that make it, its a great experience not unlike the trips that many elite Spring/Summer teams make out to Boston, NAHA or even Toronto. I also think that by adding a third class, it could be $$$ worth while for the Mn HS League and at the same time promote the growth of girls hockey in the state from th etop down.

[36Guy]

I have a tear in my eye. I feel so bad that only one team from section 6AA will make it to state. bla-bla-bla......... What about all the teams in all the other sports that dont make it and way worse teams go to state? Were not going to change it ever boys! Deal with the way it is cause they are not changing it for girls hockey and then doing nothing in all the other HS school sports. You guys that dont like the rules can always move and take your chances at a differnt school.

I never said anything about 6AA...bla-bla-bla! In addition, I never asked for a change and I never talked about boys hockey. In my mind rankings are to try and figure out who is the best of the best and who we would like to see square off at the end out of the year whether you are from 6AA of Roseau.

As far as moving, I like my chances right where I am at!

[SECoach]

I have a tear in my eye. I feel so bad that only one team from section 6AA will make it to state. bla-bla-bla......... What about all the teams in all the other sports that dont make it and way worse teams go to state? Were not going to change it ever boys! Deal with the way it is cause they are not changing it for girls hockey and then doing nothing in all the other HS school sports. You guys that dont like the rules can always move and take your chances at a differnt school.

I never said anything about 6AA...bla-bla-bla! In addition, I never asked for a change and I never talked about boys hockey. In my mind rankings are to try and figure out who is the best of the best and who we would like to see square off at the end out of the year whether you are from 6AA of Roseau.

As far as moving, I like MY chances right where I am at!

Pretty much says it all....How did YOU like YOUR chances before YOU moved?

[sinbin]

Football already changed and basketball is discussing changes. I would agree, though, that hockey is behind those sports in terms of equity and we'll all just need to live with the status quo for quite some time, if not always. Until that day of equity arrives, the reality is that many lesser teams will continue to get free passes to the state tournament. Life isn't fair; we may as well teach that reality to our children when they're young . . . and keep hitting them over the head with it year after year after year until it sinks in.

Very Metrocentric... Petition to be in Section 7 or Section 8, schedule a home and away game against 5 of the teams in your new section (Tuesdays and Thursdays are fun) and take advantage of your free pass. Seems like a fair deal...

Great suggestion, D16D. It does indeed seem like a fair deal . . . and a lot of fun, to boot! What a great opportunity to play in all of Minnesota's storied northern rinks. Think of the Minnesota hockey history that our girls could experience firsthand. Think of the bonding and extra study time that could take place on the long bus rides. The only problem I foresee is losing out on playing tougher competition during the regular season, so we wouldn't be as battle-tested once we play in the state tournament each year.

[D16Dad]

Football already changed and basketball is discussing changes. I would agree, though, that hockey is behind those sports in terms of equity and we'll all just need to live with the status quo for quite some time, if not always. Until that day of equity arrives, the reality is that many lesser teams will continue to get free passes to the state tournament. Life isn't fair; we may as well teach that reality to our children when they're young . . . and keep hitting them over the head with it year after year after year until it sinks in.

Very Metrocentric... Petition to be in Section 7 or Section 8, schedule a home and away game against 5 of the teams in your new section (Tuesdays and Thursdays are fun) and take advantage of your free pass. Seems like a fair deal...

Great suggestion, D16D. It does indeed seem like a fair deal . . . and a lot of fun, to boot! What a great opportunity to play in all of Minnesota's storied northern rinks. Think of the Minnesota hockey history that our girls could experience firsthand. Think of the bonding and extra study time that could take place on the long bus rides. The only problem I foresee is losing out on playing tougher competition during the regular season, so we wouldn't be as battle-tested once we play in the state tournament each year.

That is the trade off, coast in or earn it...

[pucktech]

Pretty sure..!!?!..No Positive...!!!..this thread is called "KRACH Rankings"...

My opinion but if you want to present other rankings, opinions, etc...start a new thread...

I like not only this model but also the frequency. The person that started this put time/resources into it for quite a few years now. Our team isn't in the top 10 so I'm not saying I like it for placement of our team...but for another look at the overall competition...and if others have other math, etc...it might be interesting to look at that sometime too. Again, this thread is "KRACH Rankings"...

Thanks !

[sinbin]

Exactly, D16Dad, thanks for understanding!

[OsMetroDad]

Below is a tabulated comparison of ranking numbers vs. KRACH

Also posted a couple of graphs showing the relationships between KRACH & LSQRANK up on LSQRANK.WEEBLY.COM

Code: Select all

Comparison of Ranking Systems as of 05-JAN-2014
-----------------------------------------------

LSQRANK       Team  SEC CLS LSQGD KRACH   QRF     LPHCP_A   LPHCP_AA
--------------------------------------------------------------------- 
  1       HillMurray  4 AA  0.0    274     97.0               6
  2              BSM  6 AA  0.8    686    119.9               4
  3          Hopkins  6 AA  0.9    629    117.3               3
  4            Edina  2 AA  1.0    226     96.0               5
  5       Minnetonka  6 AA  1.1   1184    126.6               1
  6          Wayzata  6 AA  1.1    880    113.0               2
  7       LakevilleN  1 AA  1.2    221     96.9               7
  8      EdenPrairie  2 AA  1.2    190    100.1               9
  9  ThiefRiverFalls  8 A   1.3    816     91.5       1
 10   EastGrandForks  8 A   1.3    272     92.4       3
 11         Irondale  5 AA  1.9    113     93.4              10
 12       LakevilleS  1 AA  2.0    133     93.8              11
 13      DodgeCounty  1 AA  2.1     96     83.1              12
 14         Achiever  4 A   2.1    109     87.6       7
 15           Blaine  5 AA  2.1    144     97.9               8
 16       MoundsView  5 AA  2.5     68     86.4              15
 17       Stillwater  4 AA  2.6     67     81.1              14
 18       MapleGrove  6 AA  2.7     52     74.7              --
 19       Burnsville  3 AA  2.9     78     76.3              18
 20          Warroad  8 A   2.9     33     64.4       6
 21          Andover  7 AA  3.0     44     73.2              19
 22         ElkRiver  7 AA  3.1     39     71.9              --
 23            Eagan  3 AA  3.2     43     68.8              --
 24        Roseville  4 AA  3.2     46     83.7              13
 25         Eastview  3 AA  3.2     68     80.7              20
 26            Blake  5 A   3.2     91     84.9       2
 27        BloomJeff  2 AA  3.3     75     76.4              17
 28  ArmstrongCooper  6 AA  3.3     46     70.8              --
 29    RochesterCENT  1 AA  3.5     70     83.7              --
 30           Roseau  8 AA  3.6     20     53.3              --
 31          Buffalo  8 AA  3.6     21     63.5              16
 32              CDH  4 AA  3.6     56     84.9
 33      RochesterJM  1 AA  3.7     37     81.1
 34           Chaska  2 AA  3.8     63     82.1
 35       Centennial  5 AA  3.9     48     72.7
 36            Orono  5 A   3.9     29     65.9       8
 37        EastRidge  3 AA  4.0     20     70.7
 38      SouthStPaul  4 A   4.0     61     68.1       5
 39            Anoka  5 AA  4.1     27     60.9
 41      AppleValley  3 AA  4.4     10     44.3
 43          RedWing  1 A   4.4     30     67.5       4
 44       HolyFamily  2 AA  4.5     32     60.8
 45     ChamplinPark  5 AA  4.6     18     61.1
 46         Hastings  3 AA  4.7     21     64.7
 47    WhiteBearLake  4 AA  4.7     17     61.6
 48      GrandRapids  7 AA  4.7     23     57.1
 49          Bemidji  8 AA  4.8     15     49.2
 50            Breck  5 A   4.9     16     48.2      13
 51          StCloud  8 AA  5.0     42     75.8
 52        PriorLake  2 AA  5.2     10     42.3
 53       NorthWRCTY  8 AA  5.2     13     55.7
 54           Duluth  7 AA  5.3      9     48.3
 55       ForestLake  7 AA  5.3     18     59.2
 56    MoundWestonka  5 A   5.3     30     66.3      11
 57           SPAVIS  4 A   5.4     13     48.4      16
 58          Cloquet  7 AA  5.5     10     44.3
 59      ProctorHerm  7 A   5.5     15     59.4      14
 60   SpringLakePark  5 AA  5.6     22     61.1
 61           ParkCG  3 AA  5.6      5     39.5
 62       CoonRapids  5 AA  5.7      5     36.2
 63     ChisagoLakes  4 A   5.9      7     45.3      18
 64           Tartan  4 AA  5.9     11     57.5
 65          Hibbing  7 A   6.0     12     49.6      10
 66      TotinoGrace  5 A   6.1     31     59.0       9
 67         Shakopee  2 AA  6.2     13     42.4
 68        NewPrague  1 A   6.2      8     49.0      17 
 69       RiverLakes  8 AA  6.2     13     54.2
 71        Mahtomedi  4 A   6.4      9     44.6      --
 72       Farmington  1 AA  6.5      5     44.5
 73         Owatonna  1 AA  6.5      4     50.2
 74           Rogers  5 AA  6.5     10     57.2
 75      NorthStPaul  4 AA  6.6      1     35.2
 77         Woodbury  3 AA  6.6      5     39.3
 78              SSR  8 AA  6.9     20     62.0
 79       Alexandria  6 A   7.0      9     49.1      15
 81        Princeton  5 A   7.1      6     47.4      --
 83           NewUlm  3 A   7.3     12     54.3      20
 84           IFALLS  7 A   7.3      3     37.0      --
 85        StFrancis  7 AA  7.3      5     40.4
 86        MooseLake  7 A   7.5      5     41.4      --
 87        MplsNovas  5 A   7.6      4     35.7      --
 88      MankatoEast  2 A   7.7     12     61.6      12
 89         Moorhead  8 AA  7.8      3     31.6
 92           Simley  4 A   8.1      2     35.4      --
 94      FergusFalls  6 A   8.3      2     34.9      --
 95        Cambridge  7 AA  8.3      0.2   27.0
 96       HolyAngels  5 A   8.6      2     30.1      --
 97     StPaulBlades  4 A   8.7      2     28.2      --
 98           Sibley  4 A   8.7      0.6   25.8      --
100        Crookston  8 A   8.8      2     30.7      --
101         Rosemont  3 AA  8.9      1     27.4
102       Northfield  1 A   9.0      2     35.9      19
103     DetroitLakes  6 A   9.0      4     30.8
105         Brainard  8 AA  9.1      0.6   21.3
108              LDC  2 A   9.2      3     36.4
109      StLouisPark  6 AA  9.4      0.7   22.3
110       NorthMetro  6 AA  9.4      1.7   29.0
111      PequotLakes  6 A   9.9      2     32.2
112           Winona  1 A  10.2      0.9   30.8
113          Willmar  2 A  10.2      1     29.7
114       Hutchinson  2 A  10.4      2     31.0
115      MankatoWest  2 A  10.4      0.9   36.6
117        AlbertLea  1 A  10.6      2     39.8
118        Faribault  1 A  10.7      0.5   36.3
120          Luverne  3 A  10.7      2     46.7
121    RochesterMayo  1 AA 10.8      0.4   33.3
123       ParkRapids  8 A  11.0      2     32.9
125          Eveleth  7 A  11.5      2     29.7
126           Austin  1 A  11.5      0.7   31.9
127        SilverBay  7 A  11.8      0.6   17.0
128             LPGE  6 A  12.1      0.7   29.7
130       MHHAunited  4 A  12.4      0.3   20.9
131           Waseca  1 A  13.5      0.1   17.1
132              LOW  8 A  14.2      0.1    8.2
133          StPeter  2 A  14.6      0.2   23.8
134         Marshall  3 A  14.9      0.1   24.0
135         Fairmont  3 A  15.4      0.1   26.7
137           Morris  6 A  17.3      0.1   15.7
138           Windom  3 A  20.6      0.0   12.3
139      Worthington  3 A  28.9      0.0    9.9
==========================================================
 40          Hanover  1 NH  4.2 
 42          Bismark  1 ND  4.4 
 70        WestFargo  1 ND  6.3 
 76       CentralWIS  1 WI  6.6 
 80             USMW  1 WI  7.0 
 82      FortFrances  1 ON  7.2  
 85        StFrancis  7 AA  7.3 
 90          Hayward  1 WI  7.8 
 91         Onalaska  1 WI  8.0 
 93          StCroix  1 WI  8.2 
 99      NewRichmond  1 WI  8.8 
104           Hudson  1 WI  9.1 
106           Dryden  1 ON  9.1 
107            WSFLG  1 WI  9.2 
116        SunPraire  1 WI 10.5 
119       FargoNorth  1 ND 10.7 
122       FargoSouth  1 ND 10.8 
124       GrandForks  1 ND 11.4 
129         Superior  1 WI 12.3 
136        Brookings  1 SD 15.9 

[MNHockeyFan]

Below is a tabulated comparison of ranking numbers vs. KRACH

OsMetroDad, thanks for doing these! I do wonder, though, if this an "apples-to-apples" comparison because the KRACH rankings have not been updated since January 1st, whereas the LSQGD's take the games since then into account. There were a number of important games played from the 1st to the 4th, and ideally any comparison would take into account the same set of games in tabulating the results.

[OsMetroDad]

I think the KRACH pages were already updated from the past weekend as I was extracting the data. The biggest thing that is nagging at me is not so much the apples to apples thing as it is a description or plus/minus value of an indivdual ranking value itself. In the LSQRANK method, one takes the square root of the corresponding diagonal element of the inverted design matrix to retrieve the estimated standard error of that ranking value. In the explanation of the KRACH method, he talks about performing an iterative solution, but does not get into specifics of the numerics. I know from the LSQRANK adjusted data that the plus/minus values range from +/- 0.5 to 1.0 goals for the GD values for each team. Krach & QRS must have a similar accuraccy value. Knowing this value essentially let one determine whether or not there is a statistically significant quality difference between any two teams. Where this is potentially important is when using these computer solutions to rank teams into seeding positions for the post-season section play-offs.

[MNHockeyFan]

I think the KRACH pages were already updated from the past weekend as I was extracting the data.

Not sure about this, as your table had the exact same KRACH ranking points as what ghshockeyfan determined and posted on January 1. So I don't believe we're comparing "apples vs. apples".

In any case, I would like to understand the important differences between the two ranking models, simplified so that a college grad who didn't major in math could understand the key differences!

P.S. And maybe for the balance of the season, conduct an experiment, pitting the two ranking systems against one another, making it easy with wins vs. losses only with scores not figuring into it? Now that we have a good tracking record to work with, that would be interesting! And what the heck, why not throw QRF into the mix?

[ghshockeyfan]

KRACH hasn't been updated for over a week now it seems as I've been busy. Any comparison is a poor one for this reason. Theoretically, we should compare the three rankings using the same data sets as inputs. So, this comparison will never be of the purest form as I can see already that LQS, KRACH, AND QRF all have different sets of inputs (scores). Additionally, KRACH is NOT something I invented. It's well documented on the web and USCHO.COM has talked about it many times over the last decade plus. Many feel KRACH is superior to other seeding or ranking approaches used by the NCAA. I won't get into that here, it can be found on the web with a simple Google search. Bottom line, we have another really cool ranking to use as we build a consensus ranking of all those available. That's an awesome thing for those that care about this sort of thing. Very cool.

One last observation, LQS has hill murray high and that to me feels like a representation more of where they've been trending over time as opposed to a rankIing reflecting what they have actually accomplished with their entire body of work this season. Meaning, is LQS telling us the hottest teams as of late and projecting that to an ultimate ranking?

Obviously if hill continues their upward trend, then they'll be at or near the top by season's end. I get the feeling that is what LQS is telling me.

Comparing this to all other rankings, they feel more about what you've accomplished as opposed to what you're trending towards.

So, for me, I always liked KRACH as it seemed to tell me where a team deserved to be ranked based on their body of work against opponents faced to-date as opposed to trend analysis or future prediction proposes. Interestingly, I'd say KRACH correlates pretty well with LPH up until the end of season state tourney adjustment for the teams that participated.

[sinbin]

Very good explanation, GHSHFan. This would seem to align with most of our "gut feels" of trends, too, based on some of the most recent results. I suppose that the LPH rankings, although we know they're potentially fraught with issues, are theoretically a hybrid of the scientific tehniques. E.g., "I'd like to give H-M the #1 ranking since I think they're currently the best team, but I have to rank Tonka above them, since Tonka beat H-M earlier in the season." I realize this doesn't work when you have "A defeated B, B defeated C, but C defeated A" scenarios and that every ranker will have his/her own personal ideas and possible agendas. When in doubt, take the average or weighted average of multiple models and that often turns out to be a decent consensus estimator.

[OsMetroDad]

Yes, Hill Murray has been pushed to the top by the Least Squares method and mainly because of the game observation where HM beat EP by a significant MOV earlry on in the season, and EP continues to rate high, so in a way HM and EP are "linked". The algorthim that computes the weighting of the game MOV observations does not differentiate between early-mid-late season games, so right now it computing a geometric average over the entire season to date, i.e., 1-1-1 would represent an equal weighting scheme. I would think that it maybe would be better to weight late season games more than early season games, just need some guideance as to what would be ok to put in as pre-defined or user-defined options in the GUI. It is also possible to input a user-defined weight per game observation, but I only do this for the small group of games that that get out of hand into the double-digit range of MOV's. These are games where a much better team beats a lesser team by 10 or more goals plus a large shot differential. You have also sparked an idea that I have put on the back burner, but maybe need to be investigated now, and that is the analysis the cross-correlation between teams in the system of equations. Basiscally these values are the off-diagional values of the inverted N matrix. Values near 1.0 would represent a strong correlation and values near 0.0 would be a weak correllation. In layman's terms, it sort of like summing up the 1st and 2nd connection that one has in LinkedIN professional network. I would need to think about how this info could be incorporated into the computation of the weights.

[ghshockeyfan]

Updated 1/7 5PM 14640

[ghshockeyfan]

Updated 1/11 2AM 15385

[ghshockeyfan]

Updated 1/12 6PM 15769

[ghshockeyfan]

Updated 1/15

[ghshockeyfan]

Updated 1/17

[ghshockeyfan]

Updated 1/19 1AM 16628

[OsMetroDad]

ghshockeyfan - I updated the LSQRANK.WEEBLY.COM web site for the MN HS Girls A & AA classes. A little movement among the teams, but after both Hopkins lost to EP and LVN lost both to Achiever & Burnsville, Hill Murray shot back up to Number 1 in the least squares solution. Looking at the data, even though Hill Murray has 3 loses, only one of those loses has a negative composite MOV which explains why HM moves to the top in the solution. The other factor is that HM had a significant +MOV against EP, which is now also ranked high.

Did you ever get around to figuring out how to output a posteriori sigma value for the KRACH ranking values? Would like to compare what LSQRANK is telling me to what we know with KRACH. LAQRANK outputs a GD (Goal Differential) value which is used to sort the rankings. It has approximately a +- of about 1.0 goal, but I do report of 0.1 goal resolution to spread the field. What it tells me is that statistically there is little difference between the teams at the top, are you seeing the same with KRACH?

One last thing, several coaches wnat to use the various computerized rankings as input in their section seeding meetings coing up about 1 week or so, they are wondering what percentage or coverage of the game are included in your results.

[36Guy]

ghshockeyfan - I updated the LSQRANK.WEEBLY.COM web site for the MN HS Girls A & AA classes. A little movement among the teams, but after both Hopkins lost to EP and LVN lost both to Achiever & Burnsville, Hill Murray shot back up to Number 1 in the least squares solution. Looking at the data, even though Hill Murray has 3 loses, only one of those loses has a negative composite MOV which explains why HM moves to the top in the solution. The other factor is that HM had a significant +MOV against EP, which is now also ranked high.

Did you ever get around to figuring out how to output a posteriori sigma value for the KRACH ranking values? Would like to compare what LSQRANK is telling me to what we know with KRACH. LAQRANK outputs a GD (Goal Differential) value which is used to sort the rankings. It has approximately a +- of about 1.0 goal, but I do report of 0.1 goal resolution to spread the field. What it tells me is that statistically there is little difference between the teams at the top, are you seeing the same with KRACH?

One last thing, several coaches wnat to use the various computerized rankings as input in their section seeding meetings coing up about 1 week or so, they are wondering what percentage or coverage of the game are included in your results.

STOP!!!!!

[ghshockeyfan]

Did you ever get around to figuring out how to output a posteriori sigma value for the KRACH ranking values? Would like to compare what LSQRANK is telling me to what we know with KRACH. LAQRANK outputs a GD (Goal Differential) value which is used to sort the rankings. It has approximately a +- of about 1.0 goal, but I do report of 0.1 goal resolution to spread the field. What it tells me is that statistically there is little difference between the teams at the top, are you seeing the same with KRACH?

One last thing, several coaches wnat to use the various computerized rankings as input in their section seeding meetings coing up about 1 week or so, they are wondering what percentage or coverage of the game are included in your results.

OS - PM me on here and we can discuss the stats/math detail.

For those of you interested in KRACH and the math, a simple google search or first post in this thread should give some links.

I ran a sanity check a few weeks back and found that the composite ranking of KRACH, LSQ, and QRF was best estimated by KRACH from a least squares fit perspective - almost 2-to-1 for QRF and nearly 3-to-1 for LSQ. Anyone interested can PM me for details - in other words, KRACH best reflects the average of the three (or the "composite") as of that time.

Lastly - for anyone still reading - every game posted to the Hub is included in KRACH *IF* it is against a MSHSL opponent.

And, many coaches and sections have used computer rankings (including KRACH) in their discussions for over a decade.

It's important to note that I'd still rank a section by the following:
1) head-to-head results
2) common opponent results - especially within section
3) overall record and strength of opponents faced
4) other stuff like computer rankings, etc.

[ghshockeyfan]

Updated 1/22 10PM 17896

[MNHockeyFan]

SOS (Strength-Of-Schedule) is at:
http://www.bgoski.com/KRACH_SOS.htm

Interesting that the Top 5 are all of the Lake Conference teams, interrupted by No. 6 BSM, and then followed by four of the South Suburban Conference teams at Nos. 7-10.