Bracketology (11/6)

[bbg95]

Nov 11, 2024 15:23:09 GMT -5 bluepenquin said:

Nov 11, 2024 14:36:09 GMT -5 bigtenchamp1997 said:

Bluepenguin
The Massey ratings I used are from this morning and are through Sunday night.

Massey has a couple rankings.  RAT is probably the one you were using.  If I remember - this is more based on wins and losses.  PWR is the other rating right next to this - and this is the one they describe as relating to how good a team is - as in future matches.  Most people want to use RAT, but if we are looking at who Massey thinks is better - we should use PWR.

Totally disagree. Massey doesn't use PWR as his rating. It seems to be a starting point, and it's adjusted to account for actual results.

[horns1]

I know we've discussed in some prior Bracketology threads what other NCAA sports do as far as rules preventing teams from the same conference being matched up prior to the Final Four. Women's soccer released their bracket this afternoon.  A couple of intra-conference potential matchups in the quarterfinals, and none in the Sweet 16 (as far as seeded teams):

1 Miss. State vs 2 Arkansas
1 Florida State vs 2 North Carolina


www.ncaa.com/brackets/soccer-women/d1/2024

[trojansc]

Nov 11, 2024 16:44:08 GMT -5 bbg95 said:

Nov 11, 2024 15:43:41 GMT -5 BigDigEnergy said:

And Pablo has Penn State at 11 this week!

That kind of discredits Pablo tbh.

It's really hard for any ranking system that takes sets and/or points into account to NOT knock Penn State for dropping sets to Iowa, Northwestern, Maryland, Michigan State, Indiana, Yale, and going 5 with OSU/UCLA.

I think it's fair to have one system say none of that matters - it's only wins/losses. But taking into account margin of victory and trying to rank teams based on 'power' skill as Massey would call it - it makes sense.

But there's also probably outliers in every ranking system. Pablo sometimes has them. But as far as predictive measures, there's been no other method as accurate as Pablo.

[bbg95]

Nov 11, 2024 18:22:21 GMT -5 trojansc said:

Nov 11, 2024 16:44:08 GMT -5 bbg95 said:

That kind of discredits Pablo tbh.

It's really hard for any ranking system that takes sets and/or points into account to NOT knock Penn State for dropping sets to Iowa, Northwestern, Maryland, Michigan State, Indiana, Yale, and going 5 with OSU/UCLA.

I think it's fair to have one system say none of that matters - it's only wins/losses. But taking into account margin of victory and trying to rank teams based on 'power' skill as Massey would call it - it makes sense.

But there's also probably outliers in every ranking system. Pablo sometimes has them. But as far as predictive measures, there's been no other method as accurate as Pablo.

Sure. I was only half serious.

[n00b]

THIS IS THE WAY

[bluepenquin]

Nov 11, 2024 16:47:41 GMT -5 bbg95 said:

Nov 11, 2024 15:23:09 GMT -5 bluepenquin said:

Massey has a couple rankings.  RAT is probably the one you were using.  If I remember - this is more based on wins and losses.  PWR is the other rating right next to this - and this is the one they describe as relating to how good a team is - as in future matches.  Most people want to use RAT, but if we are looking at who Massey thinks is better - we should use PWR.

Totally disagree. Massey doesn't use PWR as his rating. It seems to be a starting point, and it's adjusted to account for actual results.

PWR is the measure they use when determining win probabilities, not RAT.  

Let us look at Wisconsin and Penn State

Wisconsin has a RAT ranking of 6 and a PWR ranking of 4.  Penn State has a RAT ranking of 4 and a PWR ranking of 9.

If Massey was using RAT has their measure for win probabilities (who is the better team), then we would expect Penn State to be favored in a 'matchup'.  If Massey is using PWR - we would expect Wisconsin.

Go to matchup and put in Wisconsin vs. Penn State - and Massey shows Wisconsin with a 61% chance of beating Penn State.  Therefore - PWR is the measure Massey uses to determine the better team.  RAT is just adjusting for how they did in terms of wins and losses Bayesian formula.  And this is consistent with what everyone else is seeing - Penn State has been very fortunate in close matches this season and have gotten absolutely hammered in their losses.  

[bluepenquin]

Nov 11, 2024 23:54:09 GMT -5 n00b said:

THIS IS THE WAY

For the record - I think winning matches should be what counts for tournament seeding, not winning by more points or having close losses.  

Also - shout out to Evollve (which I keep misspelling) for potentially being the best D1 volleyball site for analytics.  

[n00b]

Nov 12, 2024 9:01:51 GMT -5 bluepenquin said:

Nov 11, 2024 23:54:09 GMT -5 n00b said:

THIS IS THE WAY

For the record - I think winning matches should be what counts for tournament seeding, not winning by more points or having close losses.  

Also - shout out to Evollve (which I keep misspelling) for potentially being the best D1 volleyball site for analytics.  

Agreed. That Strength of Record metric itself doesn’t consider scores.

(Although it does use a score based rating system to produce match win probabilities. But a team can’t help their own rating by winning by more)

[c4ndlelight]

Nov 11, 2024 18:22:21 GMT -5 trojansc said:

Nov 11, 2024 16:44:08 GMT -5 bbg95 said:

That kind of discredits Pablo tbh.

It's really hard for any ranking system that takes sets and/or points into account to NOT knock Penn State for dropping sets to Iowa, Northwestern, Maryland, Michigan State, Indiana, Yale, and going 5 with OSU/UCLA.

I think it's fair to have one system say none of that matters - it's only wins/losses. But taking into account margin of victory and trying to rank teams based on 'power' skill as Massey would call it - it makes sense.

But there's also probably outliers in every ranking system. Pablo sometimes has them. But as far as predictive measures, there's been no other method as accurate as Pablo.

The other thing is that matches for Top 10 teams against Iowa, Northwestern, Maryland, Michigan State and Yale don't really matter (unless they lose - and even then it's usually aberration/illness related like when Penn St. lost to Oregon St. that one time).

Having tight margins against an Indiana/OSU/UCLA might mean something, but not against those really terrible teams.

One of the problems with mathematical rating systems like this is that you can't tell the machine to just turn off what it looks at.

[bbg95]

Nov 12, 2024 8:56:02 GMT -5 bluepenquin said:

Nov 11, 2024 16:47:41 GMT -5 bbg95 said:

Totally disagree. Massey doesn't use PWR as his rating. It seems to be a starting point, and it's adjusted to account for actual results.

PWR is the measure they use when determining win probabilities, not RAT.  

Let us look at Wisconsin and Penn State

Wisconsin has a RAT ranking of 6 and a PWR ranking of 4.  Penn State has a RAT ranking of 4 and a PWR ranking of 9.

If Massey was using RAT has their measure for win probabilities (who is the better team), then we would expect Penn State to be favored in a 'matchup'.  If Massey is using PWR - we would expect Wisconsin.

Go to matchup and put in Wisconsin vs. Penn State - and Massey shows Wisconsin with a 61% chance of beating Penn State.  Therefore - PWR is the measure Massey uses to determine the better team.  RAT is just adjusting for how they did in terms of wins and losses Bayesian formula.  And this is consistent with what everyone else is seeing - Penn State has been very fortunate in close matches this season and have gotten absolutely hammered in their losses.  

My point is that Massey uses RAT as his actual rating. PWR is just a component and not equivalent to Pablo, which is the actual rating for that system. I don't think PWR is as refined as Pablo is, so I don't really see the point of using it when its creator uses RAT as the primary ranking.

[bluepenquin]

Nov 12, 2024 9:32:59 GMT -5 bbg95 said:

Nov 12, 2024 8:56:02 GMT -5 bluepenquin said:

PWR is the measure they use when determining win probabilities, not RAT.  

Let us look at Wisconsin and Penn State

Wisconsin has a RAT ranking of 6 and a PWR ranking of 4.  Penn State has a RAT ranking of 4 and a PWR ranking of 9.

If Massey was using RAT has their measure for win probabilities (who is the better team), then we would expect Penn State to be favored in a 'matchup'.  If Massey is using PWR - we would expect Wisconsin.

Go to matchup and put in Wisconsin vs. Penn State - and Massey shows Wisconsin with a 61% chance of beating Penn State.  Therefore - PWR is the measure Massey uses to determine the better team.  RAT is just adjusting for how they did in terms of wins and losses Bayesian formula.  And this is consistent with what everyone else is seeing - Penn State has been very fortunate in close matches this season and have gotten absolutely hammered in their losses.  

My point is that Massey uses RAT as his actual rating. PWR is just a component and not equivalent to Pablo, which is the actual rating for that system. I don't think PWR is as refined as Pablo is, so I don't really see the point of using it when its creator uses RAT as the primary ranking.

In terms of win probabilities - which was the point of the discussion, who is the better team and who is more likely to win a hypothetical match - PWR is clearly what Massey uses and not RAT.  And when wanting Massey's comparable to Pablo - it would have to be PWR.  Pablo directly creates win probabilities.  Massey also directly creates win probabilities from its PWR rating and not the RAT rating.

[bluepenquin]

Nov 12, 2024 9:22:11 GMT -5 c4ndlelight said:

Nov 11, 2024 18:22:21 GMT -5 trojansc said:

It's really hard for any ranking system that takes sets and/or points into account to NOT knock Penn State for dropping sets to Iowa, Northwestern, Maryland, Michigan State, Indiana, Yale, and going 5 with OSU/UCLA.

I think it's fair to have one system say none of that matters - it's only wins/losses. But taking into account margin of victory and trying to rank teams based on 'power' skill as Massey would call it - it makes sense.

But there's also probably outliers in every ranking system. Pablo sometimes has them. But as far as predictive measures, there's been no other method as accurate as Pablo.

The other thing is that matches for Top 10 teams against Iowa, Northwestern, Maryland, Michigan State and Yale don't really matter (unless they lose - and even then it's usually aberration/illness related like when Penn St. lost to Oregon St. that one time).

Having tight margins against an Indiana/OSU/UCLA might mean something, but not against those really terrible teams.

One of the problems with mathematical rating systems like this is that you can't tell the machine to just turn off what it looks at.

Sure - but this goes the other way also.  We cannot just ignore things that doesn't fit our narrative and assume they are meaningless.  i.e., there are always going to be exceptions and outliers - but they aren't really that common with large sample sizes (like total points played YTD in the season).  

I can get behind a narrative that Wisconsin was doing some weird stuff that led to them losing their first 3 matches - that wouldn't likely happen if they played today (and even with this - Pablo has them as the #5 team and not #11 like Penn State).  But for Penn State to consistently play to their competition and raise the level in crunch time (unless they are getting blown out) is a heavy lift and not very plausible.  

[n00b]

Nov 12, 2024 10:11:13 GMT -5 bluepenquin said:

Nov 12, 2024 9:22:11 GMT -5 c4ndlelight said:

The other thing is that matches for Top 10 teams against Iowa, Northwestern, Maryland, Michigan State and Yale don't really matter (unless they lose - and even then it's usually aberration/illness related like when Penn St. lost to Oregon St. that one time).

Having tight margins against an Indiana/OSU/UCLA might mean something, but not against those really terrible teams.

One of the problems with mathematical rating systems like this is that you can't tell the machine to just turn off what it looks at.

Sure - but this goes the other way also.  We cannot just ignore things that doesn't fit our narrative and assume they are meaningless.  i.e., there are always going to be exceptions and outliers - but they aren't really that common with large sample sizes (like total points played YTD in the season).  

I can get behind a narrative that Wisconsin was doing some weird stuff that led to them losing their first 3 matches - that wouldn't likely happen if they played today (and even with this - Pablo has them as the #5 team and not #11 like Penn State).  But for Penn State to consistently play to their competition and raise the level in crunch time (unless they are getting blown out) is a heavy lift and not very plausible.  

I generally disagree with c4ndlelight 's premise that failing to be dominant against Iowa, Northwestern, Maryland, Michigan State and Yale is irrelevant. I think those are useful data points when comparing Penn State to other top teams.

[bbg95]

Nov 12, 2024 10:23:46 GMT -5 n00b said:

Nov 12, 2024 10:11:13 GMT -5 bluepenquin said:

Sure - but this goes the other way also.  We cannot just ignore things that doesn't fit our narrative and assume they are meaningless.  i.e., there are always going to be exceptions and outliers - but they aren't really that common with large sample sizes (like total points played YTD in the season).  

I can get behind a narrative that Wisconsin was doing some weird stuff that led to them losing their first 3 matches - that wouldn't likely happen if they played today (and even with this - Pablo has them as the #5 team and not #11 like Penn State).  But for Penn State to consistently play to their competition and raise the level in crunch time (unless they are getting blown out) is a heavy lift and not very plausible.  

I generally disagree with c4ndlelight 's premise that failing to be dominant against Iowa, Northwestern, Maryland, Michigan State and Yale is irrelevant. I think those are useful data points when comparing Penn State to other top teams.

I don't know if I'd say they're completely irrelevant, but I think they're significantly less relevant than Penn Sate's performance against top opponents. Anyway, we'll see how Penn State does against Purdue. If they lose, I'll reevaluate. But right now, they have just two road losses to top 10 opponents.

[c4ndlelight]

Nov 12, 2024 10:11:13 GMT -5 bluepenquin said:

Nov 12, 2024 9:22:11 GMT -5 c4ndlelight said:

The other thing is that matches for Top 10 teams against Iowa, Northwestern, Maryland, Michigan State and Yale don't really matter (unless they lose - and even then it's usually aberration/illness related like when Penn St. lost to Oregon St. that one time).

Having tight margins against an Indiana/OSU/UCLA might mean something, but not against those really terrible teams.

One of the problems with mathematical rating systems like this is that you can't tell the machine to just turn off what it looks at.

Sure - but this goes the other way also.  We cannot just ignore things that doesn't fit our narrative and assume they are meaningless.  i.e., there are always going to be exceptions and outliers - but they aren't really that common with large sample sizes (like total points played YTD in the season).  

I can get behind a narrative that Wisconsin was doing some weird stuff that led to them losing their first 3 matches - that wouldn't likely happen if they played today (and even with this - Pablo has them as the #5 team and not #11 like Penn State).  But for Penn State to consistently play to their competition and raise the level in crunch time (unless they are getting blown out) is a heavy lift and not very plausible.  

And I’m saying it’s not very plausible that Penn St. is #11 (behind a couple of teams they absolutely TROUNCED) when they have only lost to two Top 5 teams.