Bart Torvik's T-Rankings Now Used by Selection Committee

[BlueSince92]

Another metric to keep an eye on.

This isn't brand new news (about nine months old) but I don't recall us discussing it here. Replaces Jeff Sagarin's ratings, which Jeff retired. Bart seems like a stand-up guy. One day we'll see our own @Aike in an interview like this.



We are currently #8 by this metric.

Top Ten:

1. Houston
2. Duke
3. Auburn
4. Tennessee
5. Gonzaga
6. Iowa State
7. Alabama
8. MARK POPE'S KILLA GANGSTAS
9. Kansas
10. Illinois

Link to T-Rank

 

[Richard Cory]

Another metric to keep an eye on.

This isn't brand new news (about nine months old) but I don't recall us discussing it here. Replaces Jeff Sagarin's ratings, which Jeff retired. Bart seems like a stand-up guy. One day we'll see our own @Aike in an interview like this.

View embedded media

We are currently #8 by this metric.

Top Ten:

1. Houston
2. Duke
3. Auburn
4. Tennessee
5. Gonzaga
6. Iowa State
7. Alabama
8. MARK POPE'S KILLA GANGSTAS
9. Kansas
10. Illinois

Link to T-Rank

I had just asked about the old Sargin ratings in another thread earlier today. Thanks for the info and link. Wonder what metrics it uses that makes it like Houston and Duke so much.

 

[NociHTTP]

And yet we beat two of those teams above us, lol.

 

[Smashcat]

I had just asked about the old Sargin ratings in another thread earlier today. Thanks for the info and link. Wonder what metrics it uses that makes it like Houston and Duke so much.

Houston is the new Gonzaga. Often overvalued.

 

[ukcatz12]

Houston is the new Gonzaga. Often overvalued.

Where do you guys get these opinions? Neither is overvalued. Gonzaga has been to nine straight S16s, and in that time has gotten to two FFs and a title game. Houston has been to five straight S16, a FF, and lost to 3 against Duke last year in a game where they lost Shead to injury. Their team last year was elite.

It's damn hard to win a title, and just because a team can't break through and do it doesn't mean they're overrated or overvalued.

 

[BlueSince92]

I had just asked about the old Sargin ratings in another thread earlier today. Thanks for the info and link. Wonder what metrics it uses that makes it like Houston and Duke so much.

Right? Especially Houston which, while I admit none of their losses are bad losses at all, what good wins do they have to balance that? Some combination of being high in the initial guesswork, looking good in pure statistics, and having only "good" losses, I guess.

 

[Lempface]

Where do you guys get these opinions? Neither is overvalued. Gonzaga has been to nine straight S16s, and in that time has gotten to two FFs and a title game. Houston has been to five straight S16, a FF, and lost to 3 against Duke last year in a game where they lost Shead to injury. Their team last year was elite.

It's damn hard to win a title, and just because a team can't break through and do it doesn't mean they're overrated or overvalued.

What does past season performance have to do with current season rankings? It shouldn't anyways, just like a 3 loss Uconn in the top 25 after only winning against 200+ ranked teams and Baylor is a joke.

 

[ukcatz12]

What does past season performance have to do with current season rankings? It shouldn't anyways, just like a 3 loss Uconn in the top 25 after only winning against 200+ ranked teams and Baylor is a joke.

Nothing really, but the comment I replied to said "often overvalued" when referring to Houston and Gonzaga, implying there were many previous seasons where they were overvalued. And these metric rankings are usually pretty accurate come tournament time. I remember last year Alabama took a lot of non-conference losses and people complained they were still highly ranked on Kenpom. Turned out they were actually good after all.

 

[yabbadabbadoo]

Houston 1
Duke 2

Sorry, this metric seems faulty and stupid

 

[H00SIER-Cat]

Aike's model is the only one I recognize.

 

[Lempface]

Nothing really, but the comment I replied to said "often overvalued" when referring to Houston and Gonzaga, implying there were many previous seasons where they were overvalued. And these metric rankings are usually pretty accurate come tournament time. I remember last year Alabama took a lot of non-conference losses and people complained they were still highly ranked on Kenpom. Turned out they were actually good after all.

I'd like to see rankings that more accurately reflect what you've done by who you've beaten and not used as a prediction model. Any team can get hot at the right moment and go on a run, it doesn't justify a higher than earned ranking IMO. Losing 3 games back to back to back like UConn did in their first non-trivial competition should certainly have them out of the top 25 in any metric that is used by the NCAA for tournament selection and seeding IMO.

Predictive models are fine to use to see what teams are likely to go on a run and for betting purposes I just don't like them being used for selection criteria.

 

[Jeff Sykes]

Any system or metric that doesn't put the vast majority of its weight on head to head match up results and ranks those who lost and have more losses ahead of the team and or teams that have beaten them is flawed and not to be taken seriously.

 

[Jim Malone Lt.]

Any system or metric that doesn't put the vast majority of its weight on head to head match up results and ranks those who lost and have more losses ahead of the team and or teams that have beaten them is flawed and not to be taken seriously.

So you think Clemson should be above Kentucky?

 

[Fox2monk]

Another metric to keep an eye on.

This isn't brand new news (about nine months old) but I don't recall us discussing it here. Replaces Jeff Sagarin's ratings, which Jeff retired. Bart seems like a stand-up guy. One day we'll see our own @Aike in an interview like this.

View embedded media

We are currently #8 by this metric.

Top Ten:

1. Houston
2. Duke
3. Auburn
4. Tennessee
5. Gonzaga
6. Iowa State
7. Alabama
8. MARK POPE'S KILLA GANGSTAS
9. Kansas
10. Illinois

Link to T-Rank

Yeah @Aike lets go! Your time to shine has arrived.

 

[Fox2monk]

And yet we beat two of those teams above us, lol.

That bodes well for our future as it shows how strong they are.

 

[Fox2monk]

Houston is the new Gonzaga. Often overvalued.

This year probably but they are recruiting well for next year and have had it rolling with a great culture.

 

[TankedCat]

Another metric to keep an eye on.

This isn't brand new news (about nine months old) but I don't recall us discussing it here. Replaces Jeff Sagarin's ratings, which Jeff retired. Bart seems like a stand-up guy. One day we'll see our own @Aike in an interview like this.

View embedded media

We are currently #8 by this metric.

Top Ten:

1. Houston
2. Duke
3. Auburn
4. Tennessee
5. Gonzaga
6. Iowa State
7. Alabama
8. MARK POPE'S KILLA GANGSTAS
9. Kansas
10. Illinois

Link to T-Rank

for a moment I was all mad - "I DON"T SEE KENTUCKY ON THE LIST!"

 

[Jeff Sykes]

So you think Clemson should be above Kentucky?

So do you think Duke and Zaga should be ranked ahead of UK? And not just ranked ahead of them but ranked several spots ahead of them. Do you think that is not flawed in so way?

 

[Jim Malone Lt.]

So do you think Duke and Zaga should be ranked ahead of UK? And not just ranked ahead of them but ranked several spots ahead of them. Do you think that is not flawed in so way?

I don't think they are manually manipulating their formulas in a grand conspiracy against Kentucky so I guess not. If you dig into Torvik's stats Kentucky has not shot the ball really well against the best competition they played which is holding them back a bit. Gonzaga's Baylor destruction has really helped their stats with a limited plate. Duke shot very well in their win against Auburn. Kentucky really hasn't had a shooting game like that on the resume in a big game yet. Watch them jump up in a big way once they do, which they should be more inclined to do once they get some of these teams at Rupp Arena. Duke got Auburn at home, so that is an early advantage for them. If Kentucky had Clemson at home, they might have already had theirs as well. These statistics are devoid of emotion so most of the time they tell you the truth. Last year they foretold that Kentucky's defense would be their undoing, and they were right as rain.

 

[KAB Cats]

I have a problem with two teams Kentucky has beat being ranked ahead of Kentucky, both who have a worse record.

Like I get the subjective algorithms used to rank team performances, but I don't agree that any dimension should be held paramount to wins/losses in direct head to head competition.

 

[TankedCat]

in fairness to Duke - Kentucky, Kansas and Auburn before December 9 should count for something.

But Duke, Clemson and Gonzaga is no slouch schedule by any means. Really, I can't believe Cal agreed to that. He must have known he was on the hot seat with low attendance and fan apathy

 

[Jeff Sykes]

I don't think they are manually manipulating their formulas in a grand conspiracy against Kentucky so I guess not. If you dig into Torvik's stats Kentucky has not shot the ball really well against the best competition they played which is holding them back a bit. Gonzaga's Baylor destruction has really helped their stats with a limited plate. Duke shot very well in their win against Auburn. Kentucky really hasn't had a shooting game like that on the resume in a big game yet. Watch them jump up in a big way once they do, which they should be more inclined to do once they get some of these teams at Rupp Arena. Duke got Auburn at home, so that is an early advantage for them. If Kentucky had Clemson at home, they might have already had theirs as well. These statistics are devoid of emotion so most of the time they tell you the truth. Last year they foretold that Kentucky's defense would be their undoing, and they were right as rain.

None of that matters is my point. What you may think is the most important thing, what someone else may think is the most important thing, and what I may think is the most important thing to use to have a system isn't always correct. I personally put more stock into who won the head to head match up. You can spout numbers and stats all day but at the end of it the only stat that mattered was who won the game. UK's shooting numbers are down no doubt but outside of the Clemson game, they still hung 80 to 90 points on the scoreboard so really how does that hurt their ranking? UK just put 90 points on the scoreboard against the #7 team in the country and you want to penalize them for shooting what you or whoever thinks is a subpar shooting percentage? They out rebounded the #7 team in the country but yet you want to penalize them for missing shots? I'm not arguing that UK needs to be higher ranked. That is not what I'm saying and nowhere did I say it was a conspiracy theory. NOWHERE. I don't believe in them. I am just saying these systems are flawed. Until the people who can do this come up with a set in stone metrics to use they are all flawed because they are based on the bias of the person or persons who created the system. You keep talking about low shooting percentage holding UK's ranking down but outside of 1 game, they are putting big points on the board even against the highly ranked teams they have played. You explain to me how a system can penalize a team for scoring 90 points against the #7 team. UK scored 77 points against the # 5 team when they played Duke. How can you hold that against their ranking? The most important stat in both of those games? UK won. That is the only stat that truly matters.

 

[Jim Malone Lt.]

None of that matters is my point. What you may think is the most important thing, what someone else may think is the most important thing, and what I may think is the most important thing to use to have a system isn't always correct. I personally put more stock into who won the head to head match up. You can spout numbers and stats all day but at the end of it the only stat that mattered was who won the game. UK's shooting numbers are down no doubt but outside of the Clemson game, they still hung 80 to 90 points on the scoreboard so really how does that hurt their ranking? UK just put 90 points on the scoreboard against the #7 team in the country and you want to penalize them for shooting what you or whoever thinks is a subpar shooting percentage? They out rebounded the #7 team in the country but yet you want to penalize them for missing shots? I'm not arguing that UK needs to be higher ranked. That is not what I'm saying and nowhere did I say it was a conspiracy theory. NOWHERE. I don't believe in them. I am just saying these systems are flawed. Until the people who can do this come up with a set in stone metrics to use they are all flawed because they are based on the bias of the person or persons who created the system. You keep talking about low shooting percentage holding UK's ranking down but outside of 1 game, they are putting big points on the board even against the highly ranked teams they have played. You explain to me how a system can penalize a team for scoring 90 points against the #7 team. UK scored 77 points against the # 5 team when they played Duke. How can you hold that against their ranking? The most important stat in both of those games? UK won. That is the only stat that truly matters.

Advanced metrics might be above your pay grade.

 

[Jeff Sykes]

Advanced metrics might be above your pay grade.

Maybe actually watching a game with your own eyes and believing what they see is above your pay grade. Numbers are not the end all be all and they never ever tell the whole story. I gave you a prime example that you yourself said. UK is getting hurt in the rankings because they shot a low percentage but yet they still scored 80 to 90 points. Now I admit I'm not the smartest person in the room unlike some who seem to think of themselves but I can actually see with my own eyes what the scoreboard said. Advanced metrics be damned. Metrics are good for predictive reasoning. They are not so good at ranking anything. So you can shovel that crap somewhere else. Maybe next time don't be a smartass.

 

[Jim Malone Lt.]

Maybe actually watching a game with your own eyes and believing what they see is above your pay grade. Numbers are not the end all be all and they never ever tell the whole story. I gave you a prime example that you yourself said. UK is getting hurt in the rankings because they shot a low percentage but yet they still scored 80 to 90 points. Now I admit I'm not the smartest person in the room unlike some who seem to think of themselves but I can actually see with my own eyes what the scoreboard said. Advanced metrics be damned. Metrics are good for predictive reasoning. They are not so good at ranking anything. So you can shovel that crap somewhere else. Maybe next time don't be a smartass.

Glad we agree then. Advanced metrics are just not for you. That's fine.

 

[Bluegrass79]

Yeah whatever it takes to get DUKE to the top.
Lol.

 

[BlueSince92]

A lot of you are missing the value of a metric system. It condenses numbers. The selection committee already knows we beat Duke head to head. This is an easy shorthand way for them to know things like tempo and points per touch so they can use that too in perspective. You go tinkering with that and subordinate everything to head-to-head results, now you only have one metric. And it’s not a very data-rich metric because the top teams won’t play each other all that much. It’s better to have multiple windows on a team and a season.

It’s fine to want head-to-head matchups weighted a little heavier. Build that into your own system and we’ll all have more tools available. It’s also fine to object to the initial preseason ordering because that has inertia and is still affecting things at this point. It’s made by flawed humans with biases. Not a lot of other kinds of humans available to set that preseason order and it has to be set somehow. But talking about it’s shortcomings makes perfect sense.

Any accusations about conspiracy or bias week to week though, when we’re talking about a metric system, is making us all dumber.

 

[Jeff Sykes]

Glad we agree then. Advanced metrics are just not for you. That's fine.

Never said that. You really need to stop assuming. You aren’t good at it. Again the numbers just don’t always tell the whole truth. They are a tool to use to help you make a decision but not the end all be all tool. Coaches will plainly tell you that yes they do use it but they trust thier gut and their eyes more than numbers. Analytics are a tool to help you make reasoned decisions not rank those decisions. You can’t discuss nothing without insulting someone. I never once insulted you, belittled you, or accused you of anything. You apparently didn’t even bother to read my original post because you assumed I was talking about it being a conspiracy against UK just the man tryng to keep them down. I never said anything like that you assumed and was wrong. You assumed that I wouldn’t use analytical information but yet again you would be wrong. It is flawed but can be useful if you can acknowledge that it is flawed.

 

[KAB Cats]

A lot of you are missing the value of a metric system. It condenses numbers. The selection committee already knows we beat Duke head to head. This is an easy shorthand way for them to know things like tempo and points per touch so they can use that too in perspective. You go tinkering with that and subordinate everything to head-to-head results, now you only have one metric. And it’s not a very data-rich metric because the top teams won’t play each other all that much. It’s better to have multiple windows on a team and a season.

It’s fine to want head-to-head matchups weighted a little heavier. Build that into your own system and we’ll all have more tools available. It’s also fine to object to the initial preseason ordering because that has inertia and is still affecting things at this point. It’s made by flawed humans with biases. Not a lot of other kinds of humans available to set that preseason order and it has to be set somehow. But talking about it’s shortcomings makes perfect sense.

Any accusations about conspiracy or bias week to week though, when we’re talking about a metric system, is making us all dumber.

We shouldn't question algorithms?

 

[BlueSince92]

We shouldn't question algorithms?

Quote the exact sentence of mine which makes you think I said you shouldn’t question algorithms.

 

[Jeff Sykes]

A lot of you are missing the value of a metric system. It condenses numbers. The selection committee already knows we beat Duke head to head. This is an easy shorthand way for them to know things like tempo and points per touch so they can use that too in perspective. You go tinkering with that and subordinate everything to head-to-head results, now you only have one metric. And it’s not a very data-rich metric because the top teams won’t play each other all that much. It’s better to have multiple windows on a team and a season.

It’s fine to want head-to-head matchups weighted a little heavier. Build that into your own system and we’ll all have more tools available. It’s also fine to object to the initial preseason ordering because that has inertia and is still affecting things at this point. It’s made by flawed humans with biases. Not a lot of other kinds of humans available to set that preseason order and it has to be set somehow. But talking about it’s shortcomings makes perfect sense.

Any accusations about conspiracy or bias week to week though, when we’re talking about a metric system, is making us all dumber.

Exactly. It is a tool to use not a end all be all ranking tool. Also these systems are flawed simply because there is no baseline set for them. What one system deems most likely mportant may not be in another one. They are flawed simply because they are being feed the standards and numbers that each person thinks is the most important. They should never release these rankings until like February. That way they might have enough current information to input.

 

[Jim Malone Lt.]

Never said that. You really need to stop assuming. You aren’t good at it. Again the numbers just don’t always tell the whole truth. They are a tool to use to help you make a decision but not the end all be all tool. Coaches will plainly tell you that yes they do use it but they trust thier gut and their eyes more than numbers. Analytics are a tool to help you make reasoned decisions not rank those decisions. You can’t discuss nothing without insulting someone. I never once insulted you, belittled you, or accused you of anything. You apparently didn’t even bother to read my original post because you assumed I was talking about it being a conspiracy against UK just the man tryng to keep them down. I never said anything like that you assumed and was wrong. You assumed that I wouldn’t use analytical information but yet again you would be wrong. It is flawed but can be useful if you can acknowledge that it is flawed.

I'm not insulting you. I'm stating the obvious. Advanced metrics aren't for you. You like the eye test. That's fine.

 

[LineSkiCat14]

in fairness to Duke - Kentucky, Kansas and Auburn before December 9 should count for something.

But Duke, Clemson and Gonzaga is no slouch schedule by any means. Really, I can't believe Cal agreed to that. He must have known he was on the hot seat with low attendance and fan apathy

Clemson is turning out to be pretty legit.

I would say that schedule is pretty par for Cal. We always had one of Duke/Kansas/MSU.. we'd then have one more of like UNC, Ucla, Gonzaga or Kansas (if not in the CC).. and Clemson kind was a surprise. I imagine when scheduling they figured it was a "good but not TOO good" P5 opponent.

 

[CrimsonCats]

There’s no one correct methodology for ranking 352 basketball teams, many of whom will never play each other in a given season. If you just want rankings based on records and the eye test, we already have the AP poll for that. Just because an analytic algorithm comes up with different results than that system doesn’t necessarily mean the data is worthless. It’s just another lens to look through if you’re trying to compare so many different teams.

A potential issue is that there’s no ties, and human voters often overreact to wins and losses. Basketball is a variable sport, and even lopsided matchups don’t have a 100% win rate for the better team. As an example for Kentucky vs Duke or Gonzaga, we’re talking about total coin flip games, one that was tied with about 20 seconds left and one that was decided by 1 point in OT. Someone had to win, and I’m glad those games ended up going our way, but there’s absolutely no statistical separation gained in games that close, so it’s not actually outrageous that a model could think those are still slightly better teams than us based on their performance in other games and that the result could tip the other way if we played again. It’s just a different logic than the one most poll voters use.

 

[The_Answer1313]

The thing about this is although the committee gets these numbers, it's always secondary information and it's questionable just how much the committee uses it

The fact is the NET is what everything is based on. Your SOS, the actual resume numbers on the team sheets......everything.

Ken Pomroy said they receive his figures but he questions how much they actually use them.

 

[The_Answer1313]

There’s no one correct methodology for ranking 352 basketball teams, many of whom will never play each other in a given season. If you just want rankings based on records and the eye test, we already have the AP poll for that. Just because an analytic algorithm comes up with different results than that system doesn’t necessarily mean the data is worthless. It’s just another lens to look through if you’re trying to compare so many different teams.

A potential issue is that there’s no ties, and human voters often overreact to wins and losses. Basketball is a variable sport, and even lopsided matchups don’t have a 100% win rate for the better team. As an example for Kentucky vs Duke or Gonzaga, we’re talking about total coin flip games, one that was tied with about 20 seconds left and one that was decided by 1 point in OT. Someone had to win, and I’m glad those games ended up going our way, but there’s absolutely no statistical separation gained in games that close, so it’s not actually outrageous that a model could think those are still slightly better teams than us based on their performance in other games and that the result could tip the other way if we played again. It’s just a different logic than the one most poll voters use.

Exactly. So maybe the systems thought Duke and Gonzaga were better to begin with and after close UK wins they still feel that Duke and Gonzaga is better just the gap closed somewhat.

That's how these things work.

Another thing tho I wish most would understand these teams are all closer than most want to admit. If someone wants to say that UK is better than Duke and Gonzaga especially considering we beat the two teams, it's a valid argument. An equal argument could be made tho for them.

Neither is technically "incorrect"

 

[Jeff Sykes]

I'm not insulting you. I'm stating the obvious. Advanced metrics aren't for you. You like the eye test. That's fine.

Above your pay grade isn't saying they aren't for you. So just stop. You can think and believe what you want it isn't any of my business because it doesn't affect me personally. To be honest I don't really care just as I believe you could say the same to me. But I will not be talked down to or insulted especially when I did neither to you to start with. I guess I'm just too old because I believe in respecting others.

 

[Jim Malone Lt.]

Above your pay grade isn't saying they aren't for you. So just stop. You can think and believe what you want it isn't any of my business because it doesn't affect me personally. To be honest I don't really care just as I believe you could say the same to me. But I will not be talked down to or insulted especially when I did neither to you to start with. I guess I'm just too old because I believe in respecting others.

Sorry, didn't know that above your pay grade would offend you.

 

[runt#69]

My eye is on MARK POPE'S KILLA GANGSTAS.

scary stuff there. if less equals more, then what happens if we got BOTH Butler and Kriisa out ? Beat the Boston Celtics in a best of 7 with what we got left ?

with all seriousness, the Duke and Gonzaga wins will have to rank up there throughout the season as two of the best wins for any team. Mark Pope killa Gangstas might go postal on the rest of the tournament field by then ....

 

[UKFBFan78]

I don't think they are manually manipulating their formulas in a grand conspiracy against Kentucky so I guess not. If you dig into Torvik's stats Kentucky has not shot the ball really well against the best competition they played which is holding them back a bit. Gonzaga's Baylor destruction has really helped their stats with a limited plate. Duke shot very well in their win against Auburn. Kentucky really hasn't had a shooting game like that on the resume in a big game yet. Watch them jump up in a big way once they do, which they should be more inclined to do once they get some of these teams at Rupp Arena. Duke got Auburn at home, so that is an early advantage for them. If Kentucky had Clemson at home, they might have already had theirs as well. These statistics are devoid of emotion so most of the time they tell you the truth. Last year they foretold that Kentucky's defense would be their undoing, and they were right as rain.

Wrong, these metrics are are flawed. Why does how well you shot the ball have more weight than wins against top 10 teams. Thats ridiculous. And any one who watched any game last year could have told you the defense would be their downfall.