2006-7 KRACH "Power Rankings"

[ghshockeyfan]

Updated 01/28 - 11A

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

Missing (Red impact computer rankings, black impact stats only):
GAME DATE AWAY TEAM AWAY SCORE HOME TEAM HOME SCORE
1/27/2007 Lake of the Woods Park Rapids
1/27/2007 Owatonna Roch. John Marshall
1/27/2007 Roch. Century Faribault
1/27/2007 Silver Bay/Two Harbors Fort Frances, Ontario (CA)
1/27/2007 Windom Area Morris/Benson/Hancock

[MNHockeyFan]

Can't help but notice that as the season's progressed the KRACH rankings of the order of teams changes less and less each week. It now takes a big upset to have much impact, and for the past several weeks they've been very consistent, especially amongst the Top 5.

Even so, I still don't buy into the odds that the KRACH points system is supposed to give you, even this late in the season. For example right now EP has 2366 points and Edina 928, and there is no way I would consider EP as a 2 1/2-to-1 favorite to win if they were play each other. It wasn't that long ago when Edina had a big lead over EP, and since then neither team has lost a game, and Edina's had a tougher schedule.

If nothing else, it makes you wonder.

[ghshockeyfan]

Can't help but notice that as the season's progressed the KRACH rankings of the order of teams changes less and less each week. It now takes a big upset to have much impact, and for the past several weeks they've been very consistent, especially amongst the Top 5.

Even so, I still don't buy into the odds that the KRACH points system is supposed to give you, even this late in the season. For example right now EP has 2366 points and Edina 928, and there is no way I would consider EP as a 2 1/2-to-1 favorite to win if they were play each other. It wasn't that long ago when Edina had a big lead over EP, and since then neither team has lost a game, and Edina's had a tougher schedule.

If nothing else, it makes you wonder.

I'm biased, but it seems to me that a ranking should get more steady as time goes on unless there are a series of major "upsets" to change the order at the very top. It would seem that a ranking should be more volatile as you work away from the top vs. at the top? Any ranking that is extremely volatile at the very top would make me question the algorithm if anything with a huge data set to operate on as we have now and there haven't been a ton of huge upsets to justify huge movement at the top IMHO.

That aside, I guess I don't believe a ranking any less for at least claiming for its "rating value" to have some significance vs. none at all (QRF). I really think though that an average of the LPH poll, QRF, and KRACH likely will give the best indication of where a team should be ranked. Also, I've come to realize that the fall-off in Class A is amazing. Meaning, once you get away from the top few A teams, you only need to be an average team in the state (top half overall - to top 75) and you'll still get "ranked" in A. But, again, that's the nature of Class A vs. AA...

[ghshockeyfan]

Updated 01/31 - 7A

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

Missing (Red impact computer rankings, black impact stats only):
GAME DATE AWAY TEAM AWAY SCORE HOME TEAM HOME SCORE
1/30/2007 Forest Lake Mounds View

[Bensonmum]

GHS,
Somebody is looking at the KRACH ratings.....did you notice that the Section 2AA seeds followed your last posting exactly? Is it a coincidence?

[ghshockeyfan]

GHS,
Somebody is looking at the KRACH ratings.....did you notice that the Section 2AA seeds followed your last posting exactly? Is it a coincidence?

This may be true! I think that it more likely points to the validity of the rankings - sometimes... They will however not match some section seeding due to head-to-head considerations which seems to be the #1 consideration in seeding/ranking.

[ghshockeyfan]

Updated 02/02 - 8:30A

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

Missing (Red impact computer rankings, black impact stats only):
GAME DATE AWAY TEAM AWAY SCORE HOME TEAM HOME SCORE
2/1/2007 Orono Litchfield/Dassel-Cokato
2/1/2007 Red Wing Prior Lake

[ghshockeyfan]

Updated 02/04 - 11:30P

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

Missing (Red impact computer rankings, black impact stats only):
GAME DATE AWAY TEAM AWAY SCORE HOME TEAM HOME SCORE
NONE

[hotdog]

How are the KRACH rankings doing in sectional play? Anybody know?

[ghshockeyfan]

How are the KRACH rankings doing in sectional play? Anybody know?

In that KRACH nearly mirrored the seeding, my guess would be the same for both. The upsets always are hard on KRACH & seeding, but they're bound to happen when you play a game on ice and not on paper, through rankings, or via computer.

[ghshockeyfan]

Updated 02/13 - 1:30P

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

Missing (Red impact computer rankings, black impact stats only):
GAME DATE AWAY TEAM AWAY SCORE HOME TEAM HOME SCORE
NONE[/quote]

[ghshockeyfan]

How are the KRACH rankings doing in sectional play? Anybody know?

In that KRACH nearly mirrored the seeding, my guess would be the same for both. The upsets always are hard on KRACH & seeding, but they're bound to happen when you play a game on ice and not on paper, through rankings, or via computer.

I've been told that all but 2 games KRACH has been right so far in sectionals (BV/Eagan & PHM/FL were the KRACH "upsets" using the KRACH pre-Sections)... Of course, we'll see how it goes the next few days... and I've updated the KRACH now with the section games through this AM...

[MNHockeyFan]

I've been told that all but 2 games KRACH has been right so far in sectionals (BV/Eagan & PHM/FL were the KRACH "upsets" using the KRACH pre-Sections)... Of course, we'll see how it goes the next few days... and I've updated the KRACH now with the section games through this AM...

Some of us are hoping KRACH will be wrong two more times after the Section 6AA playoffs are over

[xk1]

A pair of Krach upsets

Mahtomedi beats SSP 4-3
Hastings 1-0 over HM

[ghshockeyfan]

A pair of Krach upsets

Mahtomedi beats SSP 4-3
Hastings 1-0 over HM

My guess, w/o looking, is that these were both 2v3 games?

[Bensonmum]

ghs,
I don't want to nitpick, but since this is probably the last time for the year we're talking about KRACH, just a couple things.

On the 5AA thread (Feb 4) I wrote:

Am I the only one who is a little suspicious of the high KRACH ranking of AHA? I'm looking at their schedule and I don't see how their SOS is so high... basically, since Thanksgiving they are 2-2 against top 20 Class AA teams. They've lost to a Class A team, and the bulk of their schedule is against Class A teams.... how do they get rewarded for it in the rankings... Burnsville has played only 3 games against Class A teams....All of this makes me doubt that AHA is getting out of this section alive.

Believe me I'm not trying to blow my own horn....I would have picked Eagan to win the section, and I picked Armstrong in 4AA. Both were bounced early. But the point about AHA is that they played an incredibly weak schedule for a AA team. I did some research and found that B'ville lost 8 games. 6 of those were to EP, Edina, and Eagan. Another was to CEC. The only 'questionable' loss was to Eastview. Knowing all this, I'd say B'ville should have been favored and ranked ahead of AHA. Something in KRACH might be askew in this particular case. Wins over weak opponents might get too much importance, as do losses vs very strong opponents. And AHA had 3 'big' wins the first week of the season--do these wins 'deteriorate' pointswise over time? In other words, does a win over the #5 ranked team last week count heavier in the most recent ranking than a win from 2 months ago?
Also: After watching the HM-Hastings game, there's no way I would conclude that HM is the better team, no matter their rankings. That was not an upset.

[SEMetro]

To be fair to AHA, maxpreps says AHA outshot BV 49 to 13. I don't think anyone has done that to BV before - not EP, not Wayzata, not Eagan.

[ghshockeyfan]

You bring up some good questions. Let me get off topic a bit to start...

Right now there have been "upsets" in KRACH, but many of those same games are considered upsets by the seeding as well.

That aside, back to the points...

1) AHA - KRACH looks at which teams you beat and how high they are ranked. Meaning, if a team is 19-1 and lost only to the #1 ranked team, the assumption is that the team should be ranked below #1. How far below? Well, what is the rank of the next team they played? If it is #3, I would guess KRACH ranks this team #2. If it is something greater than #3, then I would guess it depends on what other teams lost only to team #1 - and then the question becomes what is the next highest ranked team that they beat?

I don't know if this makes any sense, but KRACH does this for each of the 126 teams considering its opponents and "RATINGS" for each of those opponents relative to record, etc.

2) For clarification, losses in KRACH count for nothing really. KRACH really forces teams to win games to earn a ranking/rating. Many actually complain that KRACH punishes teams too much for not winning even when they play amazing SOS. But, a loss is a loss is a loss to KRACH, no matter if it's 100-1 in regulation or 1-0 in 100 OT's.

Where losses may be getting confused as having value is when a team loses to a good team, but beats an amazing team, and then gets ranked above the "good" team that it lost to. Why is that??? Many scream about this as they see the head-to-head loss and ask what happened??? Well, what happened is that the team is inconsistent. Meaning, that the team gets "credit" for a W vs. the "good" team, but KRACH essentially assigns a "loss" against the "amazing" team (in how the algorithm applies itself). The end result is that you get a "Rating" or Ranking above the good team (assumed W) and below the amazing team (assumed L). How much above the good and below the amazing team is determined by other schedule/results factors for the team in question as well as other teams.

3) Where W's vs. weak teams may be looking to be overvalued is somewhat similar to what is explained in #1 above. Essentially KRACH looks at who you've played, where those teams are ranked, and then how many "Wins" you have (keep in mind that a "tie" is .5 wins, so 2 ties = 1 W in KRACH's eyes). Where KRACH is hard for us to believe sometimes is in the extreme case where we think a team plays all but a select few games against "bad" opponents - or - when a team plays no top opponents ever and goes undefeated as a result (this isn't currently the case).

An example of this in general is when a team plays 25 games, 24 against the teams ranked 101-124 in the state. Obviously, beating all 24 teams ranked 101-124 means only that this team is to be ranked above 100 overall. The 1 other game then makes all the difference in how high this team should be ranked. The ranking of that 1 other team on the schedule, along with the result of that game, plays a HUGE role in how high this team is ranked. If the team loses that game, the team in question is then ranked somewhere between 100 and wherever the team it lost to is ranked. As to where exactly is again determined by other schedule/results factors for the team in question as well as other teams.

The case when a team is undefeated is a special case and is addressed by building in a fictitious tie for all teams as to not allow the undefeated team to be rated too high. This fictitious tie binds all the teams together as well in the eyes of the algorithm and doesn't count against nor help any team more than any other. It's simply a mechanism to make the math work.


4) As to if time helps/hurts wins/losses due to other teams changing... Only the current rank of opponents is used. In some ways, if a team was great early on and you beat them early, that makes no difference relative to now. Meaning, beat a good team early and they're now bad, you only are looked at as beating the bad team currently - not wrongly given too much credit for beating a bad team when they were considered good. Similarly though, the opposite is true. If you lost to a bad team early, and now that team is top ranked, you aren't punished for losing to a team that was very bad early but now is awesome.


5) "1 Game does not a season make." We have to remember this. KRACH is a guide determined over the course of a season. It doesn't allow teams to be considered great overnight, and makes them earn their ranking through their WINS and losses (and ties) over the course of an entire season (note the emphasis on WINS and ties). It may well be that we see Team A play Team B once and can swear that one is better than the other, but there are so many factors over time that go into looking at not just one game. Also, with all this being said, it should be obvious that KRACH rewards/punishes teams that are consistently good/bad, but not the opposite. A team that beats very good teams but loses to bad teams will only be seen by KRACH to be better than the bad teams that it lost to (as the good team wins count only as bad team wins in the end). This keeps inconsistent teams "honest" so to speak.


While we've looked at some of the "worst-case-scenarios" in the 5 points above, I still have yet to find a better algorithm than KRACH. Weighting & score differential considerations as done in QRF, etc. are flawed assumptions. KRACH only looks at who you play & your record. Win and win against the best teams and you will be ranked high by KRACH, don't, and especially against poor teams, and you won't be. The math is much more difficult however as it's based on a recursive definition of team rating value calculations.

[ghshockeyfan]

To be fair to AHA, maxpreps says AHA outshot BV 49 to 13. I don't think anyone has done that to BV before - not EP, not Wayzata, not Eagan.

this brings up another very interesting point that has been raised before to me...

Why not use shot count in a ranking algorithm?

Without penalty count, and also again without knowing how coaches score control, it's a somewhat deceiving statistic sometimes.

The best example I can come up with is from a first round section playoff game I saw. In this game Team A out-shot Team B 17-12 through 2 periods. In period 3, Team B out-shot Team A 18-3. So, the end shot count was 20-30 Team B. Now, without menion of the fact that Team A took 11.5 PIM to Team B's 2 in the 3rd, it would seem that Team B is better than Team A based on shot count considerations, but, my guess is that in reality the PIM had an impact on the final shot count more than anything else. AND, if PIM count was nearly equal through periods 1 & 2 in this case, then, I would actually assume that Team A is likely better than Team B since they led 17-12 shot count (and 2-1 on the scoreboard by the way).

As to the "Score Control" scenario, shot count may still give a better indicaiton of how much better opponents are realative to the same "common opponent" however if Team A decides to stop shooting after being up by 6 and Team B doesn't??? Then this, like final score differential, means nothing as far as comparision of strength of teams A & B. The data is skewed by the coaching decisions relative to score control is my point. Hence, you must throw out any algorithm that uses score (or shot) differential in G HS Hockey. Pros may be much different, as may be College, etc. where this scenario rarely enters the games.

[ghshockeyfan]

Updated 02/15 - 1A

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

Missing (Red impact computer rankings, black impact stats only):
GAME DATE AWAY TEAM AWAY SCORE HOME TEAM HOME SCORE
NONE

[ghshockeyfan]

Updated 02/18 - 10A

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

Missing (Red impact computer rankings, black impact stats only):
GAME DATE AWAY TEAM AWAY SCORE HOME TEAM HOME SCORE
NONE

[MNHockeyFan]

I know that KRACH gives each game played the same weighting - the wins at the end of the season and in sections do not count any more than the games won in early December. This is quite a bit different than LPH/AP which always seem to have the AA and A state champions on top in their final poll. It will be interesting to see what they do this year, with EP and Edina now out of it. Especially if there is a major upset at State, like maybe if North Metro or Burnsville were to take it all (which is entirely possible I might add!).

[ghshockeyfan]

I know that KRACH gives each game played the same weighting - the wins at the end of the season and in sections do not count any more than the games won in early December. This is quite a bit different than LPH/AP which always seem to have the AA and A state champions on top in their final poll. It will be interesting to see what they do this year, with EP and Edina now out of it. Especially if there is a major upset at State, like maybe if North Metro or Burnsville were to take it all (which is entirely possible I might add!).

I don't believe that the 8 teams that make it to state are always the top 8 teams by the end of the tourney. Section 6AA is an excellent example of this. Now, the team that wins it all? They should probably be #1, but from there, I don't think it should just go order of finish at State for LPH/AP ranking - but again, this is just my opinion...

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.

KRACH will likley never boost the tourney champion to #1 on its list just because it won the championship at State. KRACH will always look at the entire season's worth of results and come up with its list. Truly though, all that matters is how a team does those last 6 games of the season. Win them all and you're the state champion. Lose any one of the first 3, or the first two at state, and your season's over just like the other 110 teams that sit out the state each year...

[MNHockeyFan]

I don't believe that the 8 teams that make it to state are always the top 8 teams by the end of the tourney. Section 6AA is an excellent example of this. Now, the team that wins it all? They should probably be #1, but from there, I don't think it should just go order of finish at State for LPH/AP ranking - but again, this is just my opinion...

Sorry, I guess I wasn't clear - I only meant the state champ(s) as finishing No.1 in the final LPH/AP poll, not the other teams being ranked Nos. 2-8. Clearly EP and Edina deserve to be ranked in at least the top 5, no matter what happens at state.

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?

[ghshockeyfan]

2006-7 Final:
http://www.bgoski.com/rank/Rankings.htm