Should head-to-head results be only factor in section seeding?

[mnmouth]

Loved it when 7 and 8 met in the Quarters. Guaranteed a Northern team in the Semis.

I see your point, an Andover - STMA quarterfinal would be fun.

It would be great to see Andover vs STMA. Hockey as we know it would end and the 218er’s would be demanding their own state tournament.

Nah. Way more fun to see a Northern team dump any private school or cake eater suburban team at state.

[The51]

Whatever the system is now, there's gotta be a better one.

After watching the 2AA semis last weekend, I have no idea how Shakopee was the 3 seed. Eden Prairie was way better and was the 5 seed.

Also, Centennial was definitely better than Champlin from what I saw yesterday. Both of these would have been caught by Pagestat.

Something like Pagestat (maybe Lee can help the MSHSL build their own version should be used)

[BlueLineSpecial]

Whatever the system is now, there's gotta be a better one.

After watching the 2AA semis last weekend, I have no idea how Shakopee was the 3 seed. Eden Prairie was way better and was the 5 seed.

Also, Centennial was definitely better than Champlin from what I saw yesterday. Both of these would have been caught by Pagestat.

Something like Pagestat (maybe Lee can help the MSHSL build their own version should be used)

Does Pagestat weigh section games differently than non-section games? I ask because the point of sections seedings is to rank the teams in the section. And an important factor is ranking a subset of teams is to look at head-to-head, and record against others in the section, common opponent W/L, etc. If the idea is simply that EP should have been ranked higher than Shakopee in 2AA because of a computer model of compiled stats, then why do we need sections at all? Just make 'sections' an NCAA basketball-style tourney where teams are ranked 1-64 and the last remaining 8 teams go to state. But if the idea of sections (beyond attempting to have statewide geographic representation) is to compare and rank teams of that subset into some sort of order, then there will be occasions where a 'better' team is ranked lower than another based on things like head-to-head, common opponents, W/L record in the section.

I think they mostly get it right, understanding that certain factors will weigh more heavily than others when you're trying to rank, say, an 8 team subset of the broader 64-ish teams.

I'm all for a Pagestat-type ranking. But again, at that point we might as well not have geographic/regional sections at all. Just put them 1-64 and have at it.

[east hockey]

Whatever the system is now, there's gotta be a better one.

After watching the 2AA semis last weekend, I have no idea how Shakopee was the 3 seed. Eden Prairie was way better and was the 5 seed.

Also, Centennial was definitely better than Champlin from what I saw yesterday. Both of these would have been caught by Pagestat.

Something like Pagestat (maybe Lee can help the MSHSL build their own version should be used)

Does Pagestat weigh section games differently than non-section games? I ask because the point of sections seedings is to rank the teams in the section. And an important factor is ranking a subset of teams is to look at head-to-head, and record against others in the section, common opponent W/L, etc. If the idea is simply that EP should have been ranked higher than Shakopee in 2AA because of a computer model of compiled stats, then why do we need sections at all? Just make 'sections' an NCAA basketball-style tourney where teams are ranked 1-64 and the last remaining 8 teams go to state. But if the idea of sections (beyond attempting to have statewide geographic representation) is to compare and rank teams of that subset into some sort of order, then there will be occasions where a 'better' team is ranked lower than another based on things like head-to-head, common opponents, W/L record in the section.

I think they mostly get it right, understanding that certain factors will weigh more heavily than others when you're trying to rank, say, an 8 team subset of the broader 64-ish teams.

I'm all for a Pagestat-type ranking. But again, at that point we might as well not have geographic/regional sections at all. Just put them 1-64 and have at it.

It weighs those the same as any other games. I don't know if changing the method would result in greater accuracy in accomplishing the stated goal of the system, which is to predict Section Tournament games. The only possibility I've come up with is to maybe do something along the lines of QRF (and I just threw up in my mouth even mentioning QRF) where if two teams rating are within a certain percentage then a head-to-head tiebreaker would be used. Would that make for a better predictive model? I wouldn't have any idea until I re-crunched the numbers going back to PageStat inception (1998). The idea intrigues me and sometimes intrigue is all it takes for me to set up another huge database, ignore my wife and our dogs, put on the headphones and spend a ton of time on a new project.

Lee

[lakescountrylife]

Whatever the system is now, there's gotta be a better one.

After watching the 2AA semis last weekend, I have no idea how Shakopee was the 3 seed. Eden Prairie was way better and was the 5 seed.

Also, Centennial was definitely better than Champlin from what I saw yesterday. Both of these would have been caught by Pagestat.

Something like Pagestat (maybe Lee can help the MSHSL build their own version should be used)

Shakopee beat Holy Family and Chaska during the regular season, teams which Eden Prairie lost to. If you check QRF, it agrees with Shakopee at 3 and EP at 5.

The best way to figure it out would be for EP & Shakopee to play each other

[BlueLineSpecial]

Whatever the system is now, there's gotta be a better one.

After watching the 2AA semis last weekend, I have no idea how Shakopee was the 3 seed. Eden Prairie was way better and was the 5 seed.

Also, Centennial was definitely better than Champlin from what I saw yesterday. Both of these would have been caught by Pagestat.

Something like Pagestat (maybe Lee can help the MSHSL build their own version should be used)

Does Pagestat weigh section games differently than non-section games? I ask because the point of sections seedings is to rank the teams in the section. And an important factor is ranking a subset of teams is to look at head-to-head, and record against others in the section, common opponent W/L, etc. If the idea is simply that EP should have been ranked higher than Shakopee in 2AA because of a computer model of compiled stats, then why do we need sections at all? Just make 'sections' an NCAA basketball-style tourney where teams are ranked 1-64 and the last remaining 8 teams go to state. But if the idea of sections (beyond attempting to have statewide geographic representation) is to compare and rank teams of that subset into some sort of order, then there will be occasions where a 'better' team is ranked lower than another based on things like head-to-head, common opponents, W/L record in the section.

I think they mostly get it right, understanding that certain factors will weigh more heavily than others when you're trying to rank, say, an 8 team subset of the broader 64-ish teams.

I'm all for a Pagestat-type ranking. But again, at that point we might as well not have geographic/regional sections at all. Just put them 1-64 and have at it.

It weighs those the same as any other games. I don't know if changing the method would result in greater accuracy in accomplishing the stated goal of the system, which is to predict Section Tournament games. The only possibility I've come up with is to maybe do something along the lines of QRF (and I just threw up in my mouth even mentioning QRF) where if two teams rating are within a certain percentage then a head-to-head tiebreaker would be used. Would that make for a better predictive model? I wouldn't have any idea until I re-crunched the numbers going back to PageStat inception (1998). The idea intrigues me and sometimes intrigue is all it takes for me to set up another huge database, ignore my wife and our dogs, put on the headphones and spend a ton of time on a new project.

Lee

I certainly don't want you to end up petless and divorced. But I'd be curious if adding slightly more weight to section W/L or common opponents changes things.

Again, I think things are working pretty well in section seedings. There will always be some outliers, and if I recall (Lee, you could confirm this pretty quickly I'm sure) Pagestat has consistently beaten the current method over the years, so maybe Pagestat or a similar modeling is the way to go. I continue to stand by the belief that (1) head-to-head (2) section W/L record and (3) W/L record of common opponents should weigh more than other stats like SOS or overall W/L record when seeding sections. But then again, I was called a bonehead...or something like that...on this board recently. So it's entirely possible I don't know what I'm talking about.

[BlueLineSpecial]

Blowhard! Not bonehead. I'm a blowhard!

[The51]

Whatever the system is now, there's gotta be a better one.

After watching the 2AA semis last weekend, I have no idea how Shakopee was the 3 seed. Eden Prairie was way better and was the 5 seed.

Also, Centennial was definitely better than Champlin from what I saw yesterday. Both of these would have been caught by Pagestat.

Something like Pagestat (maybe Lee can help the MSHSL build their own version should be used)

Does Pagestat weigh section games differently than non-section games? I ask because the point of sections seedings is to rank the teams in the section. And an important factor is ranking a subset of teams is to look at head-to-head, and record against others in the section, common opponent W/L, etc. If the idea is simply that EP should have been ranked higher than Shakopee in 2AA because of a computer model of compiled stats, then why do we need sections at all? Just make 'sections' an NCAA basketball-style tourney where teams are ranked 1-64 and the last remaining 8 teams go to state. But if the idea of sections (beyond attempting to have statewide geographic representation) is to compare and rank teams of that subset into some sort of order, then there will be occasions where a 'better' team is ranked lower than another based on things like head-to-head, common opponents, W/L record in the section.

I think they mostly get it right, understanding that certain factors will weigh more heavily than others when you're trying to rank, say, an 8 team subset of the broader 64-ish teams.

I'm all for a Pagestat-type ranking. But again, at that point we might as well not have geographic/regional sections at all. Just put them 1-64 and have at it.

The alleged purpose of sections is to get different geographical parts of the state represented. I for one would be in favor of doing away with them and having one large tournament. Maybe let the northern sections stay and have the other 6 all combine

[elliott70]

Whatever the system is now, there's gotta be a better one.

After watching the 2AA semis last weekend, I have no idea how Shakopee was the 3 seed. Eden Prairie was way better and was the 5 seed.

Also, Centennial was definitely better than Champlin from what I saw yesterday. Both of these would have been caught by Pagestat.

Something like Pagestat (maybe Lee can help the MSHSL build their own version should be used)

Does Pagestat weigh section games differently than non-section games? I ask because the point of sections seedings is to rank the teams in the section. And an important factor is ranking a subset of teams is to look at head-to-head, and record against others in the section, common opponent W/L, etc. If the idea is simply that EP should have been ranked higher than Shakopee in 2AA because of a computer model of compiled stats, then why do we need sections at all? Just make 'sections' an NCAA basketball-style tourney where teams are ranked 1-64 and the last remaining 8 teams go to state. But if the idea of sections (beyond attempting to have statewide geographic representation) is to compare and rank teams of that subset into some sort of order, then there will be occasions where a 'better' team is ranked lower than another based on things like head-to-head, common opponents, W/L record in the section.

I think they mostly get it right, understanding that certain factors will weigh more heavily than others when you're trying to rank, say, an 8 team subset of the broader 64-ish teams.

I'm all for a Pagestat-type ranking. But again, at that point we might as well not have geographic/regional sections at all. Just put them 1-64 and have at it.

It weighs those the same as any other games. I don't know if changing the method would result in greater accuracy in accomplishing the stated goal of the system, which is to predict Section Tournament games. The only possibility I've come up with is to maybe do something along the lines of QRF (and I just threw up in my mouth even mentioning QRF) where if two teams rating are within a certain percentage then a head-to-head tiebreaker would be used. Would that make for a better predictive model? I wouldn't have any idea until I re-crunched the numbers going back to PageStat inception (1998). The idea intrigues me and sometimes intrigue is all it takes for me to set up another huge database, ignore my wife and our dogs, put on the headphones and spend a ton of time on a new project.

Lee

We will throw in a free trip to Hawaii.

[east hockey]

Does Pagestat weigh section games differently than non-section games? I ask because the point of sections seedings is to rank the teams in the section. And an important factor is ranking a subset of teams is to look at head-to-head, and record against others in the section, common opponent W/L, etc. If the idea is simply that EP should have been ranked higher than Shakopee in 2AA because of a computer model of compiled stats, then why do we need sections at all? Just make 'sections' an NCAA basketball-style tourney where teams are ranked 1-64 and the last remaining 8 teams go to state. But if the idea of sections (beyond attempting to have statewide geographic representation) is to compare and rank teams of that subset into some sort of order, then there will be occasions where a 'better' team is ranked lower than another based on things like head-to-head, common opponents, W/L record in the section.

I think they mostly get it right, understanding that certain factors will weigh more heavily than others when you're trying to rank, say, an 8 team subset of the broader 64-ish teams.

I'm all for a Pagestat-type ranking. But again, at that point we might as well not have geographic/regional sections at all. Just put them 1-64 and have at it.

It weighs those the same as any other games. I don't know if changing the method would result in greater accuracy in accomplishing the stated goal of the system, which is to predict Section Tournament games. The only possibility I've come up with is to maybe do something along the lines of QRF (and I just threw up in my mouth even mentioning QRF) where if two teams rating are within a certain percentage then a head-to-head tiebreaker would be used. Would that make for a better predictive model? I wouldn't have any idea until I re-crunched the numbers going back to PageStat inception (1998). The idea intrigues me and sometimes intrigue is all it takes for me to set up another huge database, ignore my wife and our dogs, put on the headphones and spend a ton of time on a new project.

Lee

We will throw in a free trip to Hawaii.

Sounds like a Timeshare scam to me!

Lee

[ClassAGuy]

In Class A overall the Coaches seeded pretty well. All #1 seeds made it through to the Section Final.

There have only been 2 "upsets" so far in the entire 8 team sectionals of a lower seeded team winning
#7 Waconia over #2 Providence (Don't think incorrect seed more due to the Lions being so up and down and struggles keeping the puck out of the net)
#3 SCC over #2 Monticello (PageStat had it right however, they had Mankato East over Luverne wrong which the Coaches got right so it was a wash)

Seeding in 7A I think is where the biggest grips belong for Class A as Hibbing shoud've been 3 over Denfeld but again overall coaches were pretty close.

I agree with all on here the process and how seeding is done is hard to find the correct answer and finding a good solution is well even more complicated.

Overall, 15 of the 16 #1 seeds are in the Finals and mainly playing #2 seeds, 4AA could be the outlier, so overall coaches for this year at least did a pretty decent job!

[elliott70]

It weighs those the same as any other games. I don't know if changing the method would result in greater accuracy in accomplishing the stated goal of the system, which is to predict Section Tournament games. The only possibility I've come up with is to maybe do something along the lines of QRF (and I just threw up in my mouth even mentioning QRF) where if two teams rating are within a certain percentage then a head-to-head tiebreaker would be used. Would that make for a better predictive model? I wouldn't have any idea until I re-crunched the numbers going back to PageStat inception (1998). The idea intrigues me and sometimes intrigue is all it takes for me to set up another huge database, ignore my wife and our dogs, put on the headphones and spend a ton of time on a new project.

Lee

We will throw in a free trip to Hawaii.

Sounds like a Timeshare scam to me!

Lee

No scam, just give me some of that government pension and we have a deal.