Big Wins And Bad Losses: TUC

[4four4]

Big Wins and Bad Losses

This article was done a year ago but it still comes into play this year. So for decision sakes lets say a team like Minnesota drops out of the TUC. Does it hurt the top WCHA teams who beat them coming seeding time? Well it sure looks like it. If I am a fan of DU, Wisconsin, UND or any other WCHA elite team I am scratching my head.

 

[Federal League]

Re: Big Wins And Bad Losses: TUC

Big Wins and Bad Losses

This article was done a year ago but it still comes into play this year. So for decision sakes lets say a team like Minnesota drops out of the TUC. Does it hurt the top WCHA teams who beat them coming seeding time? Well it sure looks like it. If I am a fan of DU, Wisconsin, UND or any other WCHA elite team I am scratching my head.

One of the biggest reasons college hockey should switch from PWR to KRACH. Yep, I'm going there already in this thread.

 

[quiddlet]

Re: Big Wins And Bad Losses: TUC

So for decision sakes lets say a team like Minnesota drops out of the TUC. Does it hurt the top WCHA teams who beat them coming seeding time? Well it sure looks like it. If I am a fan of DU, Wisconsin, UND or any other WCHA elite team I am scratching my head.

If you're a fan of DU/Wisconsin/UND or whatever you can disagree, but you shouldn't be scratching your head. You may have a problem with the method but the philosophy is pretty clear and makes perfect sense.

TUC is a measurement of a teams wins against quality opponents. So yes, if Minnesota (for instance) drops out you lose the value of those games, but only because those wins clearly weren't as significant as they seemed before. The example teams would suffer similar fates under the RPI or KRACH systems... when Minnesota loses their RPI falls and so does the RPI benefit any team that beat them had because that win become less valuable.

Believe it or not at some point BU was 2-2 (no, really) this year and we were holding what looked like high quality wins against Michigan and UML... but it turns out Michigan and UML were not as good as we thought they were, so naturally those wins are not as high quality as they appeared, and that's the way it should be.

The philosophy behind TUC definitely makes sense to me.

 

[Bakunin]

Re: Big Wins And Bad Losses: TUC

You think the philosophy behind the TUC concept makes sense? Really?

Here is the problem with it as it is applied in the pairwise:

For a TUC comparison, one only looks at the top 25 teams on the basis of RPI. That means #25 Notre Dame counts right now while #26 NMU doesn't. The difference between the two of them is rather small - Notre Dame's RPI is .5135 while NMU's is .5088 - a gap of little more than four thousandths of a point. In fact, there's a pretty large group of teams (I count 11 of them) on either side of the cut line that are separated by a whopping two hundredths of a point in RPI (this group includes #23 MN all the way to #33 BU).

Logically speaking, does it make any sense to you that wins against a team should count more than wins against another just because it's in the top 25 of one statistical measure when a strong argument can be made that nearly a dozen teams near that cut-off point are essentially equal?

 

[Slap Shot]

Re: Big Wins And Bad Losses: TUC

Logically speaking, does it make any sense to you that wins against a team should count more than wins against another just because it's in the top 25 of one statistical measure when a strong argument can be made that nearly a dozen teams near that cut-off point are essentially equal?

I typically avoid all bracketology/PWR discussions so forgive me if this sounds like a stupid idea, but instead of the PWR completely discounting all wins against non-TUCs, why not instead assign a weighted multiplier for every team starting high to low and have that value added to whatever formula there is to calcute the PWR. A win against the #1 RPI counts as 1x, against the #2 counts as .98x, a win against the #3 counts as .96x, etc., etc., or something like that, on down the line.

 

[LynahFan]

Re: Big Wins And Bad Losses: TUC

I typically avoid all bracketology/PWR discussions so forgive me if this sounds like a stupid idea, but instead of the PWR completely discounting all wins against non-TUCs, why not instead assign a weighted multiplier for every team starting high to low and have that value added to whatever formula there is to calcute the PWR. A win against the #1 RPI counts as 1x, against the #2 counts as .98x, a win against the #3 counts as .96x, etc., etc., or something like that, on down the line.

Yes - that does sound like a good idea. If only there were already an elegant ratings system out there that rewarded you more for beating good teams on a sliding scale. What would I call such a system? Hmmm...KRUCH? KROCH? No, doesn't seem quite right. Give me a second, and I'll think of something...

/broken record

 

[Bakunin]

Re: Big Wins And Bad Losses: TUC

SS - KRACH already accomplishes that.

Ranking teams by a severely flawed approach and then multiplying wins against them by some number that tries to compensate for the flaw doesn't eliminate the flaw. In practice, it wouldn't be all that different from the NCAA's occasional formula tweaking and asterisking (which created other unintended consequences that didn't improve the situation much if at all).

 

[quiddlet]

Re: Big Wins And Bad Losses: TUC

In fact, there's a pretty large group of teams (I count 11 of them) on either side of the cut line that are separated by a whopping two hundredths of a point in RPI (this group includes #23 MN all the way to #33 BU).

Logically speaking, does it make any sense to you that wins against a team should count more than wins against another just because it's in the top 25 of one statistical measure when a strong argument can be made that nearly a dozen teams near that cut-off point are essentially equal?

Really, we're in that group?!

I agree with you... that's why I said the philosophy (counting wins only against good teams) makes sense but I guess I failed to say that I also believe the methodology is terrible. KRACH clearly does a much better job.

 

[Slap Shot]

Re: Big Wins And Bad Losses: TUC

SS - KRACH already accomplishes that.

/broken record

Shows how much I follow it all.

Bakunin - are you saying you'd prefer they continue to use the PWR rather than the KRACH or that both should be scrapped in place of something else?

 

[Priceless]

Re: Big Wins And Bad Losses: TUC

The most misunderstood aspect of the PWR is that they are only accurate once - at the end of the season. That we're able to predict and analyze this far out isn't a flaw of the system, it shows how transparent that system is. There is no TUC cliff. It's an illusion because we're able to run the pairwise before the end of the season.

TUC is the top 25 in RPI because there has to be a cutoff somewhere. Basketball uses 50 and 100. Are those any more or less arbitrary because they're bigger numbers?

 

[Federal League]

Re: Big Wins And Bad Losses: TUC

The most misunderstood aspect of the PWR is that they are only accurate once - at the end of the season. That we're able to predict and analyze this far out isn't a flaw of the system, it shows how transparent that system is. There is no TUC cliff. It's an illusion because we're able to run the pairwise before the end of the season.

TUC is the top 25 in RPI because there has to be a cutoff somewhere. Basketball uses 50 and 100. Are those any more or less arbitrary because they're bigger numbers?

The bouncy ball system is also flawed. KRACH isn't. Using KRACH, a win against the #26 team is worth slightly less than a win against the #25 team. Using PWR, a win against the #26 team is worth A LOT less than a win against the #25 team. Which one makes more sense?

 

[snwbrdr]

Re: Big Wins And Bad Losses: TUC

IMHO KRACH makes sense and PWR does not

Although my team will likely benefit from the PWR this year since they seem to like to lose to the weaker teams and beat some good ones

 

[The Sicatoka]

Re: Big Wins And Bad Losses: TUC

KRACH tries to be analytic and objective.

RPI is full of subjective, man-made, arbitrary weightings.

 

[blockski]

Re: Big Wins And Bad Losses: TUC

The most misunderstood aspect of the PWR is that they are only accurate once - at the end of the season. That we're able to predict and analyze this far out isn't a flaw of the system, it shows how transparent that system is. There is no TUC cliff. It's an illusion because we're able to run the pairwise before the end of the season.

TUC is the top 25 in RPI because there has to be a cutoff somewhere. Basketball uses 50 and 100. Are those any more or less arbitrary because they're bigger numbers?

Sorry, this line of reasoning is a load of crap.

It's true, the PWR only matters once, at the end of the season. What the day to day changes show you, however, is the massive impacts that very small chances in the dataset have on the final outcomes. It exposes all the flaws in the methodology.

 

[Slap Shot]

TUC is the top 25 in RPI because there has to be a cutoff somewhere. Basketball uses 50 and 100. Are those any more or less arbitrary because they're bigger numbers?

Why does there have to be a cutoff? Couldn't it be possible that in X sport a win against the nth place team from Y conference that is just outside the cutoff is legitimately more difficult than a win against the nth place team from Z conference several notches below? Who is to say that win should not count at all? If the system is weighted down the line, I'd imagine it would not reward a cupcake schedule nor punish one that is more difficult, and more importantly it would increase the sampling size making it more truly reflective of the results.

The most misunderstood aspect of the PWR is that they are only accurate once - at the end of the season. That we're able to predict and analyze this far out isn't a flaw of the system, it shows how transparent that system is.

That a system is tranparent in and of itself means nothing when evaluating its validity.

 

[Patman]

Re: Big Wins And Bad Losses: TUC

The most misunderstood aspect of the PWR is that they are only accurate once - at the end of the season. That we're able to predict and analyze this far out isn't a flaw of the system, it shows how transparent that system is. There is no TUC cliff. It's an illusion because we're able to run the pairwise before the end of the season.

TUC is the top 25 in RPI because there has to be a cutoff somewhere. Basketball uses 50 and 100. Are those any more or less arbitrary because they're bigger numbers?

The TUC cliff exists because it can happen at the end of the season and it is a hard cut. We notice it more since we apply it as we see it up to the current point in time.

The idea of the TUC component is to consider games against similar caliber teams... ideally of those quality of teams who should be "under consideration"... of course this is a mis-nomer... there are no teams that are "under consideration" as that implies a ponderance of human thought. The ideal here is that you throw out games against what is percieved to be weaker competition and instead only consider teams which are more in common to the ones playing. Its one way to deal with the strength of scheduling issue.

edit: the entire pairwise is geared to enhance local effects... CoP measures behavior against common opponents to generalize how they compare against each other through proxies, H2H is self-explanatory, and TUC is the topic of this thread. All of these are done to remove other results from the process for individual pairs. Its not meant to be a fair reading of the entire season or else it wouldn't be self-contradicting. Its meant to summarize performance against similar groups.

 

[Goon]

Re: Big Wins And Bad Losses: TUC

One of the biggest reasons college hockey should switch from PWR to KRACH. Yep, I'm going there already in this thread.

I think the KRACH is more accurate than the PWR.

 

[Priceless]

Re: Big Wins And Bad Losses: TUC

I apologize for intruding. I'll leave you to the KRACH circle-jerk.

Let's just call off the tournament and give the trophy to whomever KRACH says is #1.

 

[blockski]

Re: Big Wins And Bad Losses: TUC

I apologize for intruding. I'll leave you to the KRACH circle-jerk.

Let's just call off the tournament and give the trophy to whomever KRACH says is #1.

You and your strawman can enjoy that together.

Come on now - are you honestly defending the methodology of the PWR? I'm not talking about the results, or the ideas behind it, or anything else - but you're actually going to defend the methods used to come up with this ranking?

 

[ScoobyDoo]

Re: Big Wins And Bad Losses: TUC

You and your strawman can enjoy that together.

Come on now - are you honestly defending the methodology of the PWR? I'm not talking about the results, or the ideas behind it, or anything else - but you're actually going to defend the methods used to come up with this ranking?

It's defensible. Since the sweet 16 has the pairwise picked the wrong field? KRACH which is defended so highly about oh, 3 years ago or so, would have placed 8 WCHA schools in the sweet 16. To me that seems more out of whack than anything the pairwise has done.