2006-7 KRACH "Power Rankings"

[SEMetro]

Because KRACH is only based on W and L's without influence of score differential or SOG's - it seems to have a hard time determining parity (assuming the RATING column is supposed to reflect overall strength in comparison to others). Most top 20 match ups of AA teams in sections and the state tournament were a coin flip - tons of upsets. I suspect the lack of cross-over play (6-8 nonconference games) also plays in.

[ghshockeyfan]

IF we had a seeded state tourney, that would get us closer - in my mind - to LPH/AP being more consistent with Tourney Finish = Ranking, however, we'd only really ever get to a Tourney Finish = Ranking in my mind with a Tier based tourney system so that all the best teams are given an equal shot of coming out of sections vs. putting 4 of the top 5-10 in one section.

What do you think of the seeding system to be used for the boys tourney this year? Do you think something similar might be used for the girls in the future?

I don't know if this was discussed at the Spring Meeting on Sat as I couldn't make it.

[ghshockeyfan]

Because KRACH is only based on W and L's without influence of score differential or SOG's - it seems to have a hard time determining parity (assuming the RATING column is supposed to reflect overall strength in comparison to others). Most top 20 match ups of AA teams in sections and the state tournament were a coin flip - tons of upsets. I suspect the lack of cross-over play (6-8 nonconference games) also plays in.

As I've pointed out before, Score Diff is a skewed stat, as is shot count, without pen count info.

That aside, I think that some don't like the fact that KRACH doesn't weight the more recent games more heavily. But, again, the point of KRACH is a ranking depending on ALL your results from the ENTIRE season, not just the last game played, a head-to-head contest, etc.

I think that more games would always result in better results/ranking from KRACH. What the "Rating" value says is odds of team a beating team b Value A/Value B - to -1

[hockeygod]

KRACH is the most accurate way that I have seen to measure how a season has gone or is going for teams as a whole. It's not perfect ( what is) but it is the one thing that measures teams without the influance and bias that humans put into any poll.

[SEMetro]

I'm just saying look at the state tourney - KRACH #9 goes 3-0, ##19, 3, and 23 go 2-1, ## 6, 27 go 1-2, and ## 10, 11 go 0-2.

Coin flips are just as accurate.

[ghshockeyfan]

agreed. When you have odds that are nearly 1-to-1 that's what happens. Should we expect anything different at state where the best teams are usually still playing? No. Also - I want to point out that there are "upsets" sometimes, and, in those cases, any human/computer/any model usually fails!

And, let's remember that the utility of KRACH was for SEEDING purpposes where a goup of say, 126 teams don't get to all play head-to-head...

[MNHockeyFan]

As I've pointed out before, Score Diff is a skewed stat, as is shot count, without pen count info.

I still think beating a team by 3-5 goals should count for more in a computerized ranking system like KRACH than beating the same team by just 1 goal. The effect creating by some teams running up the score (vs. other teams that do not because of coaching instructions, playing the 3rd line, etc.) cound be evened out by capping the score differential at some number. If this were done then beating a team by let's say 5 goals would receive a greater weight than if the margin were just one or two, but beating that same team by 9 goals would be the same as beating them by 5.

To put it a different way, if I had to pick a winner between two teams (Team A and Team B) that had identical records after having played the exact same schedule of opposing teams, and the average margin of victory for Team A was five vs. Team B's one, I would pick Team A to win the game. Right now KRACH would have them rated exactly even.

I do agree that shot count should not be factored into any ranking system.

[ghshockeyfan]

The score diff cap was proposed too by a friend of mine. I like it and I don't for a number of reasons. Bottom line, I just think that unlike pro/college athletics there are too many coach-controlled score diff factors in HS athletics - esp. G HS Hockey in MN.

Let me give but one reason why I don't think score diff is of value, nor shot count for that matter, w/o penalty count considerations:

If I tell you team A beats team B 4-3 in OT with a 29-29 shot count for example, what is your impression of this game? Pretty evenly matched? The first question I have is "what was the penalty, or more correctly, the PP count?"

http://www.esportsdesk.com/leagues/hock ... eID=333715

Next question someone will ask is about why does it matter if only 2 PP goals in game, but I think it should be obvious what toll the constant PK takes, all the shots, etc. It impacts the game beyond just the PP goals is the point.

[SEMetro]

GHS - I didn't the RATINGS section indicated that the odds, before tourney, were 1 to 1 - especially for the top two teams. In other words, didn't KRACH have big odds against the bottom teams #23 (NMS 2-1) and #27 (Mayo 1-2) having the same records as the two top teams #3 (BSM 2-1) and #6 (Rose 1-2). If KRACH had this as 1 to 1 somewhere (that would indicate parity in the rankings), I am missing something.

[SEMetro]

If I tell you team A beats team B 4-3 in OT with a 29-29 shot count for example, what is your impression of this game? Pretty evenly matched? The first question I have is "what was the penalty, or more correctly, the PP count?"

Your assumption is that penalties should not influence ratings. Wouldn't a team that takes a penalties (as determined by an objective ref) be a factor that would make them worse?

[ghshockeyfan]

GHS - I didn't the RATINGS section indicated that the odds, before tourney, were 1 to 1 - especially for the top two teams. In other words, didn't KRACH have big odds against the bottom teams #23 (NMS 2-1) and #27 (Mayo 1-2) having the same records as the two top teams #3 (BSM 2-1) and #6 (Rose 1-2). If KRACH had this as 1 to 1 somewhere (that would indicate parity in the rankings), I am missing something.

Read:

http://www.esportsdesk.com/leagues/news ... page=15865

KRACH INTERPRETATION:

An "Average" team is considered a rating of 10.000. Any team above this 10.000 value is better than "Average" in the state (for all teams in both classes altogether). Below is worse. But, KRACH also tells us by how much...

If you want to know Team A's odds of beating Team B, you simply take the number for A and divide it by Team B's. Also easy to see the probability of A beating B (A/A '+ B).

i.e. Using the 2003-4 data, the odds of SSP beating EP is 829.495/161.646 = 5.13 approx. 5 to 1 or the probability of EP beating SSP is (161.646/(161.646 '+ 829.495)) = approx. .163 or a 16.3% probability.

I think I understand your question. Post sections pre tourney rankings were - for example - BV vs. NM
BV = 99.798
NM = 81.694

=> (which implies that...)

BV was a 99.798/81.694 - to - 1 favorite or..
BV was a 1.22-to-1 favorite or...

BV's probability of beating NM was:
(99.978/(99.978 '+ 81.694)) = approx. .550 or a 55.0% probability

This game was decided in 2OT's and as you stated before, has about the same odds/probability as flipping a coin (50% or 1-to-1 ("even"))


RK TEAM RATING
1 Eden Prairie 1039.535
2 Edina 628.776
3 Benilde-St. Margarets 517.695
4 Academy of Holy Angels 334.995
5 Blake 275.478
6 Roseville 236.722
7 Wayzata 206.870
8 Centennial 204.585
9 Stillwater 190.320
10 Grand Rapids/Greenway 188.750
11 Bemidji 186.799
12 Crookston 162.208
13 Cloquet/Esko/Carlton 160.676
14 Coon Rapids 143.907
15 Alexandria 119.629
16 Warroad 109.280
17 Blaine 101.174
18 Irondale 101.003
19 Burnsville 99.798
20 Breck 99.254
21 Cretin-Derham Hall 94.075
22 Eagan 91.387
23 North Metro Stars 81.694
24 Hopkins 79.879
25 Forest Lake 66.637
26 Roch. Mayo 65.207

[ghshockeyfan]

If I tell you team A beats team B 4-3 in OT with a 29-29 shot count for example, what is your impression of this game? Pretty evenly matched? The first question I have is "what was the penalty, or more correctly, the PP count?"

Your assumption is that penalties should not influence ratings. Wouldn't a team that takes a penalties (as determined by an objective ref) be a factor that would make them worse?

No, my point is that shot count is impacted by penalties/power plays. If two teams have an equal shot count after a greatly unequal distribution of penalties you can NOT assume that the teams are equal based on shots alone.

Similarly, score diff. is misleading as well w/o more info.

I understand your point about penalties, but you're missing my reason for citing this!

[SEMetro]

NMS v. BMS 6+ to 1 (86+%)

Bemidji v. Mayo 3.86 to 1 (74+%)

B'ville v. GRG (about 2 to 1)

The huge difference between 1 and 26 in ratings column is where I think the math formula is weakest. The top teams had too many close games - no domination like prior years - and w's and l's alone wouldn't reflect that reality.

[ghshockeyfan]

NMS v. BMS 6+ to 1 (86+%)

Bemidji v. Mayo 3.86 to 1 (74+%)

B'ville v. GRG (about 2 to 1)

The huge difference between 1 and 26 in ratings column is where I think the math formula is weakest. The top teams had too many close games - no domination like prior years - and w's and l's alone wouldn't reflect that reality.

There were some "upsets" at state depending on how you look at things.

I disagree entirely about the difference between 1 and 26 rating values.

I'm not saying KRACH is perfect, but I believe roughly in the odds/probability it gives.

In years past, SSP was a HUGE favorite, as was EP, etc. This year, not so much. That difference alone points to the accuracy of the system, as many would say that #1 wasn't as dominant this year as in years past...

Last year EP was a 10-to-1 favorite over the #2 team.

In 2004-5 SSP was nearly a 5-to-1 favorite over the #2 team.

This year we see a 1.6-to-1 ratio between #1 & #2.

It would seem to me that this alone points to the pairity that you speak of being accurately reflected?

Maybe I'm missing the point, but I should probably look at the 1 to 25 spread each of the past few years too...

[MNHockeyFan]

The score diff cap was proposed too by a friend of mine. I like it and I don't for a number of reasons. Bottom line, I just think that unlike pro/college athletics there are too many coach-controlled score diff factors in HS athletics - esp. G HS Hockey in MN..

Until the game is pretty obviously in hand I don't think ANY coach will change strategy and purposely not try to score. Certainly not after just a one goal lead would any coach feel safe and sit the first line, tell his team to just dump the puck, not shoot, etc. You could set the cap at whatever number this might be when some coaches will tell his team to back off, however he or she might do it. In high school they use 6 in the third period before running time kicks in.

Let me give but one reason why I don't think score diff is of value, nor shot count for that matter, w/o penalty count considerations:

If I tell you team A beats team B 4-3 in OT with a 29-29 shot count for example, what is your impression of this game? Pretty evenly matched? The first question I have is "what was the penalty, or more correctly, the PP count?"

I think everyone agrees that shot count should have nothing to do with it. What matters is who wins, and arguably by how much, up to a point. Regarding penalties, like SEMetro says, they are part of the game, and you often pay a price. Just like the rankings become more accurate as the season progresses, so do team's average penalty counts begin to mean more, both in losses and goals allowed. If a team gives up an average of 3 goals a game because their players are in the box all the time, they are going to be a weaker team that one that only gives up an average of one shorthanded goal a game.

Next question someone will ask is about why does it matter if only 2 PP goals in game, but I think it should be obvious what toll the constant PK takes, all the shots, etc. It impacts the game beyond just the PP goals is the point.

Again we agree that shots shouldn't matter at all. But power play goals count just as much as 5 on 5 goals, both in whether you win or lose and in the final score. If Team A beats Team B by 3-1 largely because Team B took 10 penalties, KRACH will give Team A the points for the win. But if in their next game Team B again takes 10 penalties against Team C, and Team C wins 7-1, shouldn't Team C receive a few more points for their larger margin of victory?

Of course in the big picture this isn't a big issue, but for some of us at least it's an interesting topic to discuss.

[ghshockeyfan]

Agreed. Maybe what I really have a ? about isn't score diff, but how much it plays a role and at what point. W/o a better understanding of how it factors in and to what degree, I can't say that it's entirely bad for it to have some impact. I would argue that different coches start to control at different times, if at all. Some coaches may be confident with a 3+ goal lead and start to play deep into the bench to prevent a blowout/give other players valuable experience/save stars from injury in a pointless role in a blow-out, etc. Other coaches may not have that confidence in their goaltending to start to do something until 5, 6, 7+... Some don't control at all... Some may put in a back-0up G and then subsequently need to keep playing better players in case it gets to be a "first one to ten" sort of game or to minimize QSC's (or QSO's). For all these reasons, I still don't buy into score diff as being a real good guage...

Re: PP/PK, etc. - bottom line is that with lots of PP time comes more/quality shots and also more wear on top players usually for PK vs. PP team... I'm not disagreeing that penalties are a factor in the game and I do believe that a team that takes a lot regularly should be responsible for them and how they impact the team... But, I don't think that's my point however... Instead just trying to show that shot count alone, like score alone, has many different contributing factors that need to be looked at beyond those values alone...

In the end, this brings up the question of how to write a ranking algorithm that weighs all of these factors in a semi-logical and majority-accepted way. Problem is that you really can't use anything but W/L/T though (IMHO) as Krach does as you can't get shot counts for all games as you can at least get score - but again - I also believe that while you could use score diff it's meaningless w/o further details and also the coach controlled score issues discussed above...

[MNHockeyFan]

Agreed. Maybe what I really have a ? about isn't score diff, but how much it plays a role and at what point. W/o a better understanding of how it factors in and to what degree, I can't say that it's entirely bad for it to have some impact. I would argue that different coches start to control at different times, if at all. Some coaches may be confident with a 3+ goal lead and start to play deep into the bench to prevent a blowout/give other players valuable experience/save stars from injury in a pointless role in a blow-out, etc. Other coaches may not have that confidence in their goaltending to start to do something until 5, 6, 7+... Some don't control at all... Some may put in a back-0up G and then subsequently need to keep playing better players in case it gets to be a "first one to ten" sort of game or to minimize QSC's (or QSO's). For all these reasons, I still don't buy into score diff as being a real good guage...

Re: PP/PK, etc. - bottom line is that with lots of PP time comes more/quality shots and also more wear on top players usually for PK vs. PP team... I'm not disagreeing that penalties are a factor in the game and I do believe that a team that takes a lot regularly should be responsible for them and how they impact the team... But, I don't think that's my point however... Instead just trying to show that shot count alone, like score alone, has many different contributing factors that need to be looked at beyond those values alone...

In the end, this brings up the question of how to write a ranking algorithm that weighs all of these factors in a semi-logical and majority-accepted way. Problem is that you really can't use anything but W/L/T though (IMHO) as Krach does as you can't get shot counts for all games as you can at least get score - but again - I also believe that while you could use score diff it's meaningless w/o further details and also the coach controlled score issues discussed above...

OK, we've each made our case. But I couldn't help but notice you again brought shot count into your explanation!

We can agree to disagree about at what point (how much of a lead) different coaches will start to make adjustments because they feel the game is in hand and there is no way they're going to lose. You seem to think that with some coaches this may happen when they have only a 3 goal lead, but I've seen a lot of hockey games that are lost because the team (and coach?) get overconfident, too early. I think most coaches are aware of the danger in this, and in big games where the opponent is respected they DON'T back off until the game is definitely in hand. For lesser opponents that they KNOW they will beat, their strategy to "lay off" in some fashion will happen quicker, but ultimately by capping the score differential at a reasonable number, this will have little or no impact.

Just some additional thoughts to consider -