The pairwise problem

[NV SIOUX]

We know not what the future holds, but we know who holds the future, GO SIOUX!!!

[siouxweet]

this weekend will be a very interesting weekend for us in the pwr and we don't even play. I know Jim had this in his blog but games of interest for us are:

du/umd- a denver sweep helps us in the pwr but but duluth wins will help us in the wcha race.

notre dame/ohio state- a notre dame sweep will flip our comparison with osu which we are now losing.

bc/umass-a umass win keeps this comparison in our win column, however a bc win flips it to bc

yale vs harvard/dartmouth-2 yale losses would flip this comparison

princeton vs colgate/cornell-as soon as we catch princeton in rpi in all liklihood we flip this comparison. that being said a cornell win helps us in two ways(rpi and cop).

minnesota/wisconsin-root for a split as this helps us in both the pwr and wcha race.

maine/nh-a maine sweep would probably flip the nh comparison.

[stickboy1956]

this weekend will be a very interesting weekend for us in the pwr and we don't even play. I know Jim had this in his blog but games of interest for us are:

du/umd- a denver sweep helps us in the pwr but but duluth wins will help us in the wcha race.

notre dame/ohio state- a notre dame sweep will flip our comparison with osu which we are now losing.

bc/umass-a umass win keeps this comparison in our win column, however a bc win flips it to bc

yale vs harvard/dartmouth-2 yale losses would flip this comparison

princeton vs colgate/cornell-as soon as we catch princeton in rpi in all liklihood we flip this comparison. that being said a cornell win helps us in two ways(rpi and cop).

minnesota/wisconsin-root for a split as this helps us in both the pwr and wcha race.

maine/nh-a maine sweep would probably flip the nh comparison.

Are you sure a bc win flips this comp?

We currently lead BC in RPI and TUC, but we are tied in COP. I know a BC win gives them the COP but their RPI would need to move past our's to flip this comparison. Would that happen with a BC win over Mass?

[siouxweet]

Are you sure a bc win flips this comp?

We currently lead BC in RPI and TUC, but we are tied in COP. I know a BC win gives them the COP but their RPI would need to move past our's to flip this comparison. Would that happen with a BC win over Mass?

we are only like .0016 ahead of them and a win over mass would move there rpi at least that much. plus ours would dip a little with umass losing.

[siouxguyinFF]

I noticed something while looking at the pairwise tonight. I wondered why our record against TUCs (teams under consideration) were different for different comparisons. I figured out that they throw out the games between the two teams in each individual comparison. I have not found that rule stated anywhere for sure?

I know beating the other team helps us in general (because we get head to head points in the comparison), but it seems like that hurts us in the TUC comparison. If you look at it right now, we would have one more pairwise point because we would win the comparison with the goophs if this rule was not true (we would have a better TUC record which would switch the comparison).

Anybody have any info on this part of the TUC part of the comparison?

[Sioux forever]

i don't understand the pairwise. is there a web site that explains how the system works. or would anyone like to type the novel for me so i can understand it.

[The Whistler]

Just pretend I was the guy that wrote this.

[stickboy1956]

I noticed something while looking at the pairwise tonight. I wondered why our record against TUCs (teams under consideration) were different for different comparisons. I figured out that they throw out the games between the two teams in each individual comparison. I have not found that rule stated anywhere for sure?

I know beating the other team helps us in general (because we get head to head points in the comparison), but it seems like that hurts us in the TUC comparison. If you look at it right now, we would have one more pairwise point because we would win the comparison with the goophs if this rule was not true (we would have a better TUC record which would switch the comparison).

Anybody have any info on this part of the TUC part of the comparison?

It would make sense not to count the head to head meetings twice in a comparison - that's why they throw our head to head meetings when deciding the TUC comparison.

[siouxguyinFF]

It would make sense not to count the head to head meetings twice in a comparison - that's why they throw our head to head meetings when deciding the TUC comparison.

I agree with that statement. I would think that makes sense as well. I was wondering if its stated in the rules anywhere that this is a fact? If not, I'm wondering why its not in the comparison.

[stickboy1956]

I agree with that statement. I would think that makes sense as well. I was wondering if its stated in the rules anywhere that this is a fact? If not, I'm wondering why its not in the comparison.

Here's a site that lists this exeption for determining the TUC comparison.

http://slack.net/~whelan/tbrw/tbrw.cgi?2009

Here's a link to this site's do it yourself pwr

http://slack.net/~whelan/tbrw/tbrw.cgi?2009

[yzerman19]

I do feel that the PWR must be revamped. As I look at Air Force and the EZAC teams with only a handful (or less) games against TUCs. I contend that your TUC record should carry more weight than say COP. I also contend that RPI is not the best measure, as any stat that must be adjjusted to reflect wins against bad teams is ridiculous. Do wins against those same bad teams count in COP? They sure do.

Now some of this is sour grapes from this Sioux fan, who if the season ended today would have to pick Vermont or some other team I really don't care about and invest myself in them for April.

No one wants to play UND right now, but because of three bad losses: Mass, Mich State, and Michigan Tech we are on the outside looking in.

I don't have the time to do this at work today, but I wonder what the PWR would look like if instead of winnning % vs TUCs, you counted wins vs TUCs or a derivative of total wins vs TUCs (say each TUC win is worth .25)...this would favor teams that play a lot of TUCs, and would weigh very heavily towards selection time.

Time to stop pontificating and go back to work.

[Goon]

I do feel that the PWR must be revamped. As I look at Air Force and the EZAC teams with only a handful (or less) games against TUCs. I contend that your TUC record should carry more weight than say COP. I also contend that RPI is not the best measure, as any stat that must be adjjusted to reflect wins against bad teams is ridiculous. Do wins against those same bad teams count in COP? They sure do.

Now some of this is sour grapes from this Sioux fan, who if the season ended today would have to pick Vermont or some other team I really don't care about and invest myself in them for April.

No one wants to play UND right now, but because of three bad losses: Mass, Mich State, and Michigan Tech we are on the outside looking in.

I don't have the time to do this at work today, but I wonder what the PWR would look like if instead of winnning % vs TUCs, you counted wins vs TUCs or a derivative of total wins vs TUCs (say each TUC win is worth .25)...this would favor teams that play a lot of TUCs, and would weigh very heavily towards selection time.

Time to stop pontificating and go back to work.

I think the big problem right now is that EZAC teams are over represented in the RPI/PWR this year.

However, I think a lot of this has to do with their top teams beating up on cup cakes and not playing many out of conference games. Yale, Cornell and Princeton are not as good as they are ranked right now. Think of this scenario, UND makes it to MPLS as a 3 - 4 seed and has to play Cornell. I can imagine we will hear a lot of complaining by the Cornell Coach Scaffer.

[stickboy1956]

I do feel that the PWR must be revamped. As I look at Air Force and the EZAC teams with only a handful (or less) games against TUCs. I contend that your TUC record should carry more weight than say COP. I also contend that RPI is not the best measure, as any stat that must be adjjusted to reflect wins against bad teams is ridiculous. Do wins against those same bad teams count in COP? They sure do.

Now some of this is sour grapes from this Sioux fan, who if the season ended today would have to pick Vermont or some other team I really don't care about and invest myself in them for April.

No one wants to play UND right now, but because of three bad losses: Mass, Mich State, and Michigan Tech we are on the outside looking in.

I don't have the time to do this at work today, but I wonder what the PWR would look like if instead of winnning % vs TUCs, you counted wins vs TUCs or a derivative of total wins vs TUCs (say each TUC win is worth .25)...this would favor teams that play a lot of TUCs, and would weigh very heavily towards selection time.

Time to stop pontificating and go back to work.

http://www.uscho.com/blogs/uscho/bracketology

Good analysis by UCHO regarding the RPI formula and how tweaking it effects this year's PWR.

The bottom line is that the PWR doesn't change much even if you tweak the RPI formula.

We are outside looking in right now because we lost 2 neutral site games to bad teams and lost to the 9th place team in our league at home.

Flip those games, and flip another bad loss (Wisco or Duluth) from earlier this year and we are 20-7-3 and playing for a number 1 seed, which the top team in the WCHA should be doing this time of year.

[jimdahl]

http://www.uscho.com/blogs/uscho/bracketology

Good analysis by UCHO regarding the RPI formula and how tweaking it effects this year's PWR.

The bottom line is that the PWR doesn't change much even if you tweak the RPI formula.

We are outside looking in right now because we lost 2 neutral site games to bad teams and lost to the 9th place team in our league at home.

Flip those games, and flip another bad loss (Wisco or Duluth) from earlier this year and we are 20-7-3 and playing for a number 1 seed, which the top team in the WCHA should be doing this time of year.

Though true, I think the bolded statement is a little misleading. Given that RPI is just one component of the PWR, in which most teams are compared on 2-4 criteria, reordering teams slightly in any single criterion is only going to make minor movements in the PWR. However, if we truly believed that those changes didn't matter, why bother making them?

The effect of the reweighting is very easy to describe -- the new formula diminishes the importance of strength-of-schedule in favor of win percentage (a more detailed defense of that statement is in A quick look at the effects of the new RPI formula, which I wrote in the change's first season, Jan '07).

That further encourages teams like UND to make sure their out-of-conference matchups are winnable instead of against good opponents (an attribute already present in COP and, to some degree, TUC). While UND doesn't seem to have begun optimizing its PWR chances by scheduling boring games, the incentive to do so increased by weakening the SOS component of RPI.

[Sioux_Hab-it]

I stated my belief a month ago that the bad weekend in Duluth was going to bite the Sioux in the PWR and maintain my belief that at the end of the season their only hope is going to be an autobid. I still see no reason to change my opinion.

[siouxweet]

I said this earlier in a different thread, but if we get two wins this weekend we flip a couple of comparisons(Wisc,Duluth) and with some help we could flip BC and Ohio State. There is a lot of time left so don't panic yet. Including this weekend we have 6 weekends of hockey left before selection sunday. The pwr is fun to look at and talk about but it was never meant to be looked at before the season was over. right now is only a snapshot of where teams are at. there are a lot of games countrywide left and a lot of things can happen. once again the bottom line is we need to win and this stuff will take care of itself. and it needs to start this weekend.

[sagard]

Though true, I think the bolded statement is a little misleading. Given that RPI is just one component of the PWR, in which most teams are compared on 2-4 criteria, reordering teams slightly in any single criterion is only going to make minor movements in the PWR. However, if we truly believed that those changes didn't matter, why bother making them?

The effect of the reweighting is very easy to describe -- the new formula diminishes the importance of strength-of-schedule in favor of win percentage (a more detailed defense of that statement is in A quick look at the effects of the new RPI formula, which I wrote in the change's first season, Jan '07).

That further encourages teams like UND to make sure their out-of-conference matchups are winnable instead of against good opponents (an attribute already present in COP and, to some degree, TUC). While UND doesn't seem to have begun optimizing its PWR chances by scheduling boring games, the incentive to do so increased by weakening the SOS component of RPI.

I couldn't believe it when the formula changed and we got barely a whimper from the WCHA coaches or league. Maybe they didn't complain because they had no choice. As soon as UND/UW/UM miss the tourney due to a bad record vs. tough schedule it will be cupcake city. Or should I say double/triple helpings of cupcakes.

The new formula narrows the range of RPI. A typical win near the end of the season will benefit a team 0.0020 to 0.0050. They use RPI as the top tie-breaker when differences between two teams in RPI are often in the range of one win or fractions of win. This is stupid. If they are going to make an asinine formula for RPI, the lease could do is change the tie-break formula in PWR to H2H or make a new stat with record vs. top 10 in TUC.

[Goon]

I couldn't believe it when the formula changed and we got barely a whimper from the WCHA coaches or league. Maybe they didn't complain because they had no choice. As soon as UND/UW/UM miss the tourney due to a bad record vs. tough schedule it will be cupcake city. Or should I say double/triple helpings of cupcakes.

I think it is time to load up half your schedule on cup cakes. Like the NCAA Division football does and pad stats.

[jimdahl]

I couldn't believe it when the formula changed and we got barely a whimper from the WCHA coaches or league. Maybe they didn't complain because they had no choice. As soon as UND/UW/UM miss the tourney due to a bad record vs. tough schedule it will be cupcake city. Or should I say double/triple helpings of cupcakes.

I think until they get bit by it, they take the attitude of "if we win, it will work itself out". Until you get screwed by a particular formula, you assume it's as good as any other.

Not only does a big name team on the bubble need to get excluded, but there needs to be an obviously comparable team who slipped in due to playing cupcakes. e.g. Minnesota and North Dakota have otherwise identical records, UM beat Holy Cross in the grudge match but North Dakota instead scheduled B.C. so of course lost, so UM gets in. Then people will start to think about it.

We actually saw something similar in D-II football a few years back when NDSU got excluded from the tournament in a season they beat I-AA Montana. The arbitrary SOS rules at the time gave less credit for beating a I-AA than losing to a top-tier D-II team. Those rules were change by the next Fall.

[Goon]

I think until they get bit by it, they take the attitude of "if we win, it will work itself out". Until you get screwed by a formula, you assume it's as good as any other.

Not only does a big name team on the bubble need to get excluded, but there needs to be an obviously comparable team who slipped in due to playing cupcakes. e.g. Minnesota and North Dakota have otherwise identical records, UM beat Holy Cross in the grudge match but North Dakota instead scheduled B.C. so of course lost, so UM gets in. Then people will start to think about it.

We actually saw something similar in D-II football a few years back when NDSU got excluded from the tournament in a season they beat I-AA Montana. The arbitrary SOS rules at the time gave less credit for beating a I-AA than losing to a top-tier D-II team. Those rules were change by the next Fall.

Isn't there a much beter way to do this? Say maybe the KRACH?

[siouxweet]

I think it is time to load up half your schedule on cup cakes. Like the NCAA Division football does and pad stats.

we're doing that next year with merrimack, but on the other hand we will go to cornell and have miami at home plus I'm assuming a home and home with bsu.

[Goon]

we're doing that next year with merrimack, but on the other hand we will go to cornell and have miami at home plus I'm assuming a home and home with bsu.

We have Merrimack (X2), OSU (1X), Miami (1X) (thanksgiving Tourney) at Cornell (x2) I would assume BSU (X2).

There has got to be another 2 games out of conference. There must be a trip to Boston or is Harvard coming to REA?

[siouxweet]

We have Merrimack (X2), OSU (1X), Miami (1X) (thanksgiving Tourney) at Cornell (x2) I would assume BSU (X2).

There has got to be another 2 games out of conference.

who else will be here for the thanksgiving weekend festivities?

[yzerman19]

I think BC is coming to the Ralph next year...

Do you think that the reasoning behind the adjustments was due to the dominance of a handful of teams residing in Hockey East, the WCHA and the CCHA? I could see some idiotic reasoning that it is good for the game to have more teams than the usual suspects having NCAA runs. Imagine UND finishing on a roll, losing the F5 championship game to UMD, being ranked say 7-9 and not making the NCAAs due to a loss the first weekend of October, and two 1 goal losses in a tournament? It would be a travesty.

[siouxweet]

I think BC is coming to the Ralph next year...

Do you think that the reasoning behind the adjustments was due to the dominance of a handful of teams residing in Hockey East, the WCHA and the CCHA? I could see some idiotic reasoning that it is good for the game to have more teams than the usual suspects having NCAA runs. Imagine UND finishing on a roll, losing the F5 championship game to UMD, being ranked say 7-9 and not making the NCAAs due to a loss the first weekend of October, and two 1 goal losses in a tournament? It would be a travesty.

If we don't lose another game until October, I think we'll be just fine.