2006-7 KRACH "Power Rankings"

[MNHockeyFan]

"...based on its ranking algorithm."

And isn't it humans who determine the ranking algorithm and the relative importance of SOS vs. record? And how does one determine what the correct balance between the two should be? Right now the AP (human) Class A poll has Blake ranked No. 1 (despite them losing 3 games) whereas the KRACH (computer) poll has them No. 6 amongst Class A schools. So in this case the AP (human) voters are placing more emphasis on SOS than the ranking algorithm that's been programmed into the computer. Hard to say which is more "accurate" but one has to wonder if Breck or Alexandria, to name two, would have done any better if they had included CDH, Edina, Blaine, BSM, Henry Sibley and Mayo - all good teams that they went 3-3 against.

[ghshockeyfan]

"...based on its ranking algorithm."

And isn't it humans who determine the ranking algorithm and the relative importance of SOS vs. record? And how does one determine what the correct balance between the two should be? Right now the AP (human) Class A poll has Blake ranked No. 1 (despite them losing 3 games) whereas the KRACH (computer) poll has them No. 6 amongst Class A schools. So in this case the AP (human) voters are placing more emphasis on SOS than the ranking algorithm that's been programmed into the computer. Hard to say which is more "accurate" but one has to wonder if Breck or Alexandria, to name two, would have done any better if they had included CDH, Edina, Blaine, BSM, Henry Sibley and Mayo - all good teams that they went 3-3 against.

I wish I had the time to explain this. But I don't. Bottom line. It will work itself out as more data becomes available. Trust me on this. What I said above is proof from what transpired last season.

Also - a point for clairification.

AP=LPH=7 voters. I have a good idea of who they are, but that's immaterial. I think they do a heck of a job to be honest!

KRACH=Computer Algorithm which takes into acocunt actual game results but has no preconcieved bias besides W's & L's.

And - no - I don't sit back and come up with a weighting of SOS and use that based on how I see fit. It's built into the math of how KRACH works, and actually SOS is CALCULATED based on RATING. Meaning, SOS is simply a value that is an average of all the rank values of all the opponents that a team faces. SOS evaluation doesn't come before Rating calculation...

Again, wish I had more time, but I'll try to find the links at least to the math behind this for more detail...

[ghshockeyfan]

Updated 12/15 - 5:30P

http://www.bgoski.com/rank/Rankings.htm

MISSING: - NONE

[ghshockeyfan]

Updated 12/16- 10:30A

http://www.bgoski.com/rank/Rankings.htm

Missing: (note - these games don't impact rankings anyway as they're vs non-MN opponents - only change the records/stats for the MN teams...)

12/15/2006 East Range Lady Knights Fort Frances (CA)
12/15/2006 Winona Winhawks Viroqua (WI)

[ghshockeyfan]

Updated 12/17- 10:30A

http://www.bgoski.com/rank/Rankings.htm

Missing (Red impact computer rankings, black impact stats only):

12/15/2006 East Range Lady Knights Fort Frances (CA)
12/15/2006 Winona Winhawks Viroqua (WI)
12/16/2006 Austin Roch. Century
12/16/2006 Little Falls Detroit Lakes
12/16/2006 Park Rapids Lake of the Woods
12/16/2006 River Falls (WI) Dodge County
12/16/2006 Rosemount Lakev. South
12/16/2006 Souix Falls Blue (SD) Marshall

[ghshockeyfan]

I believe that after this coming week, & especially after the holiday tourneys heading into Jan, that the rankings will start to have some validity.

This being said, there is still one huge unknown that will only be addressed over time - and maybe not even until the very end of the season... That is teams that have a VERY weak SOS that continue to win... Until they play a top team, we won't know where they truly stand (by computer standards) and we still see some teams with weak SOS at the top of the rankings as they continue to win...

[cluelessinminnesota]

http://www.minnesota-scores.com/classqr ... 10&class=3

It seems to me this computer program has a better balance between the value of the coaches poll and the value of the game results.

And it seems to have all of the top ten teams scores faster.

[xk1]

I find that the validity of a poll is directly proportional to the ranking of your favorite team(s).

[cluelessinminnesota]

I agree ther can be a strong bias in that regard.

However, I also have a big problem with Breck being rated better than the best team in the state, EP.

[ghshockeyfan]

http://www.minnesota-scores.com/classqr ... 10&class=3

It seems to me this computer program has a better balance between the value of the coaches poll and the value of the game results.

And it seems to have all of the top ten teams scores faster.

I'd average the 3 at this point (LPH, KRACH, QRF) and that would be best at this point I assume. I actually think LPH is best now & likely long term with human vote input, KRACH will be best computer choice long term as it only looks at W/L, and QRF will forever be skewed as it looks at Goal Diff which we all know is not realistic when considering score control done by some coaches and not others, etc., etc.

When I chose KRACH, I chose it for the reason that it doesn't look at goal diff. Basing a ranking on a coach controlled value is pointless.

When I chose the MNGHSH site I chose it as it allows for complete boxscore capability - speed means nothing with incomplete data in my mind.

To date, I haven't found another ranking algorithm or stats site that accomplishes either of these goals as well as KRACH & MNGHSH site. When I do, I'll switch.

[ghshockeyfan]

I find that the validity of a poll is directly proportional to the ranking of your favorite team(s).

Amen.

[ghshockeyfan]

I agree ther can be a strong bias in that regard.

However, I also have a big problem with Breck being rated better than the best team in the state, EP.

I think that this was addressed in my above statement re KRACH:

I believe that after this coming week, & especially after the holiday tourneys heading into Jan, that the rankings will start to have some validity.

This being said, there is still one huge unknown that will only be addressed over time - and maybe not even until the very end of the season... That is teams that have a VERY weak SOS that continue to win... Until they play a top team, we won't know where they truly stand (by computer standards) and we still see some teams with weak SOS at the top of the rankings as they continue to win...

Lastly, quoting someone who contacted me recently about this, "KRACH makes you earn your rank by winning." I don't believe in giving teams "extra points" for goal diff considerations. Do they factor in EN in that? I doubt it, so... And, a W/L/T is a W/L/T is it not? But, I do agree, that teams that beat better teams should be ranked higher than teams that beat weaker teams... I can point out quite a few examples from KRACH, but, w/o any supporting info. from QRF it's hard to see where they come up with anything that they do statistically...

[ghshockeyfan]

http://www.minnesota-scores.com/qrf.php

QRF Explained


The Quality Results Formula (QRF) was created to try and rank teams from different parts of the state. I won't go into a lot of detail about the formula, but will give you an idea of how it works.

Each game is given an individual QRF (iQRF) based four things: 1) win or loss, 2) opponent wins, 3) Class differential between the teams, and 4) margin of victory (or defeat). Logically, the more you win by, the more wins an opponent has, and the higher your opponents Class is, the higher your iQRF is for that game.

When all of a team's iQRF's are added together and divided by the number of MN games played, you come up with a team's QRF.
*Only MN vs. MN games count.

This number is then compared to other teams in the state to come up with an overall ranking.

There are two issues that I have with the QRF:

1)

3) Class differential between the teams

To not look at games between teams equally no matter the class is pointless. A game, is a game, is a game. Class A is a lot weaker than AA as we all know, but to automatically skew things by assuming this? Wow. That is flawed if you ask me, unless with more detail someone could explain this to me better in a way that actually makes logical sense...

2)

4) margin of victory (or defeat).

What about teams that do vs. don't score control? What about EN? More info. may explain how these issues are addressed?

[xk1]

GHS has an overall ranking which every year makes some A teams out to be stronger than known better teams.
QRF A
1 Crookston (9-0-1) 62.1
2 Alexandria (10-0-0) 57.6
3 South St. Paul (8-2-0) 54.8
4 Blake (7-3-0) 52.8
5 Austin (8-1-0) 52.4
6 Breck (12-0-0) 51.1
7 Hibbing (5-6-1) 47.5
T-8 Simley (9-1-0) 46.5
T-8 Farmington (6-3-0) 46.5
QRF AA
1 Edina (9-0-2) 59.0
2 Cloquet/Esko/Carlton (8-1-1) 57.5
3 Holy Angels Academy (10-0-0) 56.9
4 Roseville (8-1-1) 56.6
5 Benilde-St. Margaret's (8-2-0) 56.2
T-6 Wayzata (6-1-3) 55.6
T-6 Eden Prairie (7-0-0) 55.6
8 Bemidji (9-1-1) 53.5
9 Forest Lake (9-2-1) 51.5
10 Moorhead (8-2-0) 45.8

By this Crooskston is #1 in the state an both Crookston and Alex are better than EP.

I think QRF is hurting from lack of data more right now because the teams are moving up and by 5 positions every Tues, Thurs and Saturday. After the holidays I think it will be OK to compare the two as there will be enough meaningful games in the data.

[cluelessinminnesota]

I agree that there should be no weighting difference between class A and class AA. That makes no sense...the games will play out.

The empty net goals are interesting because they create "easy points" in the differential equation and I don't know the equation completely. How much more important is a goal in a 2-1 game, 3-1 game and a 7-0 game?

Overall, I don't like the QRF explanation of how they determine their rankings, but personally I feel as though it seems to wade through the early season lack of data mess faster.

I also think I would like to see the Crookston and Bemidji teams soon.

[SEhockeyDAD]

As has been noted, its still early for things to shake out. Right now, KRACH has Austin ranked higher than Rochester Mayo despite the Mayo win over Austin. It should change after Austin plays some more significant opponents, I believe.

[ghshockeyfan]

As has been noted, its still early for things to shake out. Right now, KRACH has Austin ranked higher than Rochester Mayo despite the Mayo win over Austin. It should change after Austin plays some more significant opponents, I believe.

This is the case for many teams that have played head-to-head as one game does not a season make. I will agree though too that as the season goes on that the KRACH will become more accurate with more game data.

Krach has historically done a better job of picking winners than section seeding. Where it differes from section coach seeding it has usually been right.

What I should also say is that rankings mean nothing. Just fun to debate, and also interesting to me to see how accurately an algorithm can predict outcomes. Remember, KRACH also determines odds. If AHA is ranked 1000 and Simley 50, AHA is a 1000-to-50 or 20-to-1 favorite. Or - similarly - AHA has a (1000)/(1000+50)=.9523 * 100 = 95.23% liklihood of winning...

[ghshockeyfan]

I agree that there should be no weighting difference between class A and class AA. That makes no sense...the games will play out.

The empty net goals are interesting because they create "easy points" in the differential equation and I don't know the equation completely. How much more important is a goal in a 2-1 game, 3-1 game and a 7-0 game?

Overall, I don't like the QRF explanation of how they determine their rankings, but personally I feel as though it seems to wade through the early season lack of data mess faster.

I also think I would like to see the Crookston and Bemidji teams soon.

I shouldn't be so critical of QRF as it has some value, but I think shouldn't have the score diff & class considerations in the equation.

Interesting idea re: 7-0 goals vs. 1 or 2 goal game goals. My guess is that if one team decides to stop at 7 and another 12 that that can skew rating. Similarly, if one team wins 2-1 and the other 3-1 EN, is it really fair to use that 1-goal diff from EN? I don't know...

[ghshockeyfan]

GHS has an overall ranking which every year makes some A teams out to be stronger than known better teams.
QRF A
1 Crookston (9-0-1) 62.1
2 Alexandria (10-0-0) 57.6
3 South St. Paul (8-2-0) 54.8
4 Blake (7-3-0) 52.8
5 Austin (8-1-0) 52.4
6 Breck (12-0-0) 51.1
7 Hibbing (5-6-1) 47.5
T-8 Simley (9-1-0) 46.5
T-8 Farmington (6-3-0) 46.5
QRF AA
1 Edina (9-0-2) 59.0
2 Cloquet/Esko/Carlton (8-1-1) 57.5
3 Holy Angels Academy (10-0-0) 56.9
4 Roseville (8-1-1) 56.6
5 Benilde-St. Margaret's (8-2-0) 56.2
T-6 Wayzata (6-1-3) 55.6
T-6 Eden Prairie (7-0-0) 55.6
8 Bemidji (9-1-1) 53.5
9 Forest Lake (9-2-1) 51.5
10 Moorhead (8-2-0) 45.8

By this Crooskston is #1 in the state an both Crookston and Alex are better than EP.

I think QRF is hurting from lack of data more right now because the teams are moving up and by 5 positions every Tues, Thurs and Saturday. After the holidays I think it will be OK to compare the two as there will be enough meaningful games in the data.

I woudl agree with xk1, although I will say that by year's end I think the classes sort out just fine within the "overall" ranking...

[ghshockeyfan]

Updated 12/19- 11:30A

http://www.bgoski.com/rank/Rankings.htm

Missing (Red impact computer rankings, black impact stats only):
NOTE - I don't think either of these games may have been played or were played on other dates/etc...
12/16/2006 Souix Falls Blue (SD) Marshall
12/19/2006 Mounds View Irondale

[ghshockeyfan]

Updated 12/20- 10:45A

http://www.bgoski.com/rank/Rankings.htm

Missing (Red impact computer rankings, black impact stats only):

12/19/2006 Lake of the Woods Fort Frances (CA)

[ghshockeyfan]

Updated 12/21 - 12A

http://www.bgoski.com/rank/Rankings.htm

[rubberpickle]

As has been noted, its still early for things to shake out. Right now, KRACH has Austin ranked higher than Rochester Mayo despite the Mayo win over Austin. It should change after Austin plays some more significant opponents, I believe.

you can't base a whole season on one game... maybe Austin had an off game but saying a team is bad because they lose one game is poor judgement

mayo probably is the better team but they're also AA...

[ghshockeyfan]

Updated 12/22 - 11:30A

http://www.bgoski.com/rank/Rankings.htm

NOTE - Fairmont moved from 1AA to 3A as referenced in other threads... The MSHSL has conflicting section assignment for them...

ALSO - a ton of PPD's yesterday (I assume) see other threads for details...

[ghshockeyfan]

Updated 12/23 - 1:30A

http://www.bgoski.com/rank/Rankings.htm

Missing: 12/22 Eveleth vs. LOW