Page 1 of 1

Hal's First Attempt

Posted: Tue Dec 11, 2007 11:57 am
by Undercover Hockey Lover
:shock: Shocked...not really. If you consider whose done what to whom and by how much....it's pretty darn close. What your not seeing is the sentimental favorites at the top like all the people picked polls. You can tell it has no heart which is the best part about it. I'd give it an A maybe packerboy will give it a AA. :D

Posted: Tue Dec 11, 2007 12:04 pm
by Govs93
Eh - the machine needs a good swift kick if you ask me. Johnson & Como Park both got shafted.

Posted: Tue Dec 11, 2007 12:17 pm
by packerboy
The gross injustice to the East Sides' finest aside, I will give it a AA.

HAL does pretty well during the season. He historically struggles at the end when teams get beat in sections and they still are ranked #1 at the end of the year.

But considering who is programers are , he is doin OK.

He is completely operational and all of his circuits are functioning perfectly.

Something wonderful is going to happen.

Posted: Tue Dec 11, 2007 12:43 pm
by Sparlimb
To be honest, knowing HAL's bias towards northern teams, I was completely taken aback by Silver Bay's rating. They are a sleeper and usually HAL takes that into account...

Posted: Tue Dec 11, 2007 12:47 pm
by Sparlimb
Govs93 wrote:Eh - the machine needs a good swift kick if you ask me. Johnson & Como Park both got shafted.
The good news is Johnson has Holy Family Catholic firmly in the rear view mirror. Which is where they belong. Talk about a storied program fighting for electronic recognition. Now if Johnson could just concentrate on passing St. Cloud Tech...

Posted: Tue Dec 11, 2007 12:47 pm
by east hockey
Govs93 wrote:Eh - the machine needs a good swift kick if you ask me. Johnson & Como Park both got shafted.
Really? I have them listed in PS2? I thought I'd entered them as out-of-state opponents :P

Lee

Posted: Tue Dec 11, 2007 1:10 pm
by Neutron 14
How many bonus points do privates and Moorhead get? :lol:

We need a few more iterations I think. #4 AHA 0-1, gives up 8 goals in their loss to #9 Burnsville.

And apparently HAL got wind of Roseaus scrimmage loss to Blaine.

Posted: Tue Dec 11, 2007 1:16 pm
by elliott70
And why is Park Rapids so low???
:?


:D

Posted: Tue Dec 11, 2007 1:19 pm
by packerboy
Neutron 14 wrote:How many bonus points do privates and Moorhead get? :lol:

We need a few more iterations I think. #4 AHA 0-1, gives up 8 goals in their loss to #9 Burnsville.

And apparently HAL got wind of Roseaus scrimmage loss to Blaine.
HAL knows that AHA lost to Rosemount not Burnsville.

HAL gets wind of everything and can even read lips......wait there is a message coming in....."Leave here in 2 hours or your thread will be locked out".

Posted: Tue Dec 11, 2007 1:42 pm
by Neutron 14
packerboy wrote: HAL knows that AHA lost to Rosemount not Burnsville.
I stand corrected, Rosemount at #16!

And you know its a consipracy when Forest Lake (4-0) is the 6th rated team in 7AA.

Posted: Tue Dec 11, 2007 9:58 pm
by wishock
HAL not only reads lips. He knows the future. Moorhead 2. Roseau 7.

Posted: Fri Dec 14, 2007 4:52 pm
by east hockey
HAL2 (aka PS3) has been run. You won't see the results of that system on minnhock.com...only here.

To refresh memories of those who were here last season, PS3 doesn't consider scores, only wins & losses. There was some sentiment last season it was a more accurate system than PS2 by a number of people (including me).

Before I adopt it as minnhock.com's "official" rating system, it needs more research, research which I'm unable to find time to perform this time of year, and research which I'm unwilling to perform during the off-season. So, it's sort of a Catch-22 :)

With that boring preface stated, here are the current Top 15 in each class. The number in parenthesis after each school's name indicates their PS2 ranking in that class.

Class AA

1. Minnetonka (1)
2. White Bear Lake (8 )
3. Roseau (5)
4. Centennial (6)
5. Edina (3)
6. Cloquet (13)
7. Rosemount (11)
8. Holy Angels (4)
9. Moorhead (7)
10. Roseville (12)
11. Hill-Murray (2)
12. Grand Rapids (18 )
13. Blaine (9)
14. Benilde-St. Margarets (10)
15. Woodbury (16)

Class A

1. St. Thomas Academy (1)
2. East Grand Forks (18 )
3. Warroad (11)
4. Mound-Westonka (14)
5. South St. Paul (10)
6. Rochester Lourdes (9)
7. Mahtomedi (12)
8. Duluth Marshall (2)
9. Breck (4)
10. International Falls (3)
11. Red Wing (5)
12. Hibbing (8 )
13. St. Cloud Cathedral (13)
14. Blake (21)
15. Bemidji (24)

Lee

Posted: Fri Dec 14, 2007 5:47 pm
by Neutron 14
I liked it last year, and after seeing what PS3 did to Hill, I really, really, really like it! :lol:

Posted: Fri Dec 14, 2007 6:04 pm
by Sparlimb
I like systems that penalize a team for wining by more than 5 goals...


You're headed right down the lines of Mr. Sagarin. His system has both rankings and then averages the two to get the final ranking. You could do something similar. He explains it on his personal pages. Although if you did, I think it would make sense to open PS2 back up to full scores (not limit margin of victory like you do now) since the PS3 rating effectively does that...


I quote:

In ELO CHESS, only winning and losing matters; the score margin is of no consequence,
which makes it very "politically correct". However it is less accurate in its predictions for
upcoming games than is the PURE POINTS, in which the score margin is the only thing that matters.
PURE POINTS is also known as PREDICTOR, BALLANTINE, RHEINGOLD, WHITE OWL and is the best single PREDICTOR
of future games.

The overall RATING is a synthesis of the two diametrical opposites, ELO CHESS and PURE POINTS (PREDICTOR)


http://www.kiva.net/~jsagarin/sports/sports.html

Cool

Posted: Fri Dec 14, 2007 10:39 pm
by O-townClown
So what you are saying is, no matter how you run it, Duluth East isn't in the Top 15.

It sure looks like Hill-Murray, Roseau, and Edina are the strongest on paper. A few teams (like Minnetonka) may be right there too. Pretty competitive at the top.

Re: Cool

Posted: Fri Dec 14, 2007 11:06 pm
by east hockey
O-townClown wrote:So what you are saying is, no matter how you run it, Duluth East isn't in the Top 15.
Well, it's not necessarily what I'm saying, but the numbers crunched by two computerized rating systems are certainly saying it as of today.

Lee

Posted: Sun Dec 16, 2007 10:00 pm
by east hockey
Sparlimb wrote:I like systems that penalize a team for wining by more than 5 goals...


You're headed right down the lines of Mr. Sagarin. His system has both rankings and then averages the two to get the final ranking. You could do something similar. He explains it on his personal pages. Although if you did, I think it would make sense to open PS2 back up to full scores (not limit margin of victory like you do now) since the PS3 rating effectively does that...


I quote:

In ELO CHESS, only winning and losing matters; the score margin is of no consequence,
which makes it very "politically correct". However it is less accurate in its predictions for
upcoming games than is the PURE POINTS, in which the score margin is the only thing that matters.
PURE POINTS is also known as PREDICTOR, BALLANTINE, RHEINGOLD, WHITE OWL and is the best single PREDICTOR
of future games.

The overall RATING is a synthesis of the two diametrical opposites, ELO CHESS and PURE POINTS (PREDICTOR)


http://www.kiva.net/~jsagarin/sports/sports.html
Sparlimb, thanks for the link. I found that site to be quite interesting. One thing which jumped out at me was where he stated that ELO CHESS (similiar to PS3) was more accurate in predicting future winners. I don't have data to support PS2 versus PS3 in terms of predicting future winners, but using the two systems retroactively, teams with the higher PS3 rating won 87.4% of the time, compared to 85.1% using PS2. This doesn't necessarily translate to PS3 being a better system in future games, however.

Combining two systems is something I've thought about. I'm not sure exactly how the writer of that site does it with his, since it's not just a matter of combining the two ratings and then dividing by 2. One problem with combining PS2 and PS3 is that the standard deviation is much higher in PS2, which would lead to that system having a much heavier weight in a team's average rating. In other words, it's something I need to take a closer look at. If I can win the Powerball, I'll have all the time I need. Of course, if that happens, we'll be living on Maui and won't be paying much attention to anything other than the trade winds and ocean conditions, so.... Image

Lee

you mean that guy

Posted: Sun Dec 16, 2007 10:32 pm
by O-townClown
east hockey wrote: I'm not sure exactly how the writer of that site does it with his,
Jeff Sagarin, the godfather of published computer rankings? Cripes, he's been in USA Today for at least 20 years.

I know things have been around for almost 100 years, but your characterization of Sagarin is pretty funny. He may not have invented computer sports rankings, but he did raise it to another level.

Re: you mean that guy

Posted: Sun Dec 16, 2007 10:40 pm
by east hockey
O-townClown wrote:
east hockey wrote: I'm not sure exactly how the writer of that site does it with his,
Jeff Sagarin, the godfather of published computer rankings? Cripes, he's been in USA Today for at least 20 years.

I know things have been around for almost 100 years, but your characterization of Sagarin is pretty funny. He may not have invented computer sports rankings, but he did raise it to another level.
Gotta admit, I knew virtually nothing about Jeff until Sparlimb posted the link. His site looked familiar but I can't remember how or when I checked it out in the past. Since he doesn't cover high school or NCAA hockey, I didn't pay much attention to it. The first rating system which caught my eye was Tom Hawley's system devoted to Minnesota high school hockey back in the early 90's.

Lee

Sagarin

Posted: Sun Dec 16, 2007 10:56 pm
by O-townClown
Lee, gotcha. USA Today runs Jeff's rankings weekly for college football and basketall and has since at least the late 80s.

Golfweek has used him to rank professional, college, and junior golfers for at least five years.

His computer ranking is one of those used by the BCS for college football.

If you spend time on his site you'll see he handles Indiana HS basketball, NASCAR, and just about anything else where people compete.

What is interesting to me is that he seems to prefer using not only who wins but also the margin of victory. College teams got in the habit of scoring as much as they could, not to impress voters, but to impress computers. (Okay, actually to score more points with the formula.) The mumbo jumbo of this is ELO-Chess, etc...

When I read about your PS2 and PS3 it seems like you are attempting to do the same.

I guess the obvious question is, "if you didn't pattern your computer ranking after Jeff Sagarin, who did you look to?"

Re: Sagarin

Posted: Sun Dec 16, 2007 11:22 pm
by east hockey
O-townClown wrote:Lee, gotcha. USA Today runs Jeff's rankings weekly for college football and basketall and has since at least the late 80s.

Golfweek has used him to rank professional, college, and junior golfers for at least five years.

His computer ranking is one of those used by the BCS for college football.

If you spend time on his site you'll see he handles Indiana HS basketball, NASCAR, and just about anything else where people compete.

What is interesting to me is that he seems to prefer using not only who wins but also the margin of victory. College teams got in the habit of scoring as much as they could, not to impress voters, but to impress computers. (Okay, actually to score more points with the formula.) The mumbo jumbo of this is ELO-Chess, etc...

When I read about your PS2 and PS3 it seems like you are attempting to do the same.

I guess the obvious question is, "if you didn't pattern your computer ranking after Jeff Sagarin, who did you look to?"
Originally, Mitch and Tom Hawley provided some ideas. What it's morphed to (using a cross-system of goal differential and just plain wins/losses) is mainly the result of trying to find a "sweet spot", where a win beyond a certain margin had no further effect. So I've been playing around with that, trying to see what might work out best. The problem is with the definition of "best". The idea of me looking at rankings and saying "well, this looks right, this looks off" is the result of what I think, so it's hardly objective. The only objective measurement I've come up with is using those systems to predict future games. Specifically, section tournament games.

Tom did state that he has some "bells and whistles" built into the system. He didn't specify what those things are, and I didn't ask. Likewise, with PS2, there are a couple things (momentum factor, gradually phasing out last years results) which I also have but don't give out any specifics on how they work. In other words, I don't want to give everything away! Come to think of it, Tom used the past two seasons, but I'm not sure exactly how he utilized those stats. In fact, I don't know if he even does this anymore. I used to enjoy checking out his ratings in the Duluth News-Tribune back during the Spehar/Locker days.

So, I had some direction early on and then I just sort of went with it and made some minor revisions. Just enough to make my system unique.

Back to Jeff for a minute. I don't know how he came up with the idea of combining two systems which are pretty much polar opposites from the standpoint of asking the question "What's more important? Just winning or margin of victory?" I assume that his combining systems results in a more accurate system in the area of predicting future winners. It's interesting stuff he has, and I want to research what he's done some more.

Lee