Cuse only #9 in Sagarin ratings, still behind Nova #6 | Syracusefan.com

Cuse only #9 in Sagarin ratings, still behind Nova #6

Yeah. I know there are diehard supporters of Sagarin, Kenpom, etal on here (and everywhere) but they're really, really flawed. They happen to get some teams right but then get others way wrong.

I think the main purpose of Sagarin is not pure rankings per se, but for betting purposes against the spread.
 
Yeah. I know there are diehard supporters of Sagarin, Kenpom, etal on here (and everywhere) but they're really, really flawed. They happen to get some teams right but then get others way wrong.

As JB says, sometimes one has to actually watch the games and forget the stats and the mathematical formulas.

My eyes tell me that Arizona and Syracuse are the class of the OOC portion of this season's teams. Will they remain so as the season progresses through the in conference schedule? Who knows? That's why they play the games. :)

Cheers,
Neil
 
Yeah. I know there are diehard supporters of Sagarin, Kenpom, etal on here (and everywhere) but they're really, really flawed. They happen to get some teams right but then get others way wrong.

Any time you rank 340 or however many teams it is, you're going to miss on some. But for a system that ranks every team in the country against every other one, I think they do pretty well. Every ranking system you can think of, computer or human, is going to miss on teams. Its the nature of the beast.

I think I would like to see less emphasis on results against really bad teams though. Louisville is #1 right now in Pomeroy, mainly because they have killed some terrible teams (their first 5 games were all against teams ranked 200 or higher in the Pomeroy ratings). They've since piled up 2 more blowout wins against teams 200 or higher. In general I think MOV is important in evaluating teams, but maybe there should be some attempt to cap it against teams that are truly terrible.
 
it's still early in the season, let's see where the ratings end up
 
it's still early in the season, let's see where the ratings end up
That's another part of it - KenPom, RPI, whatever, are not really worth anything until later in the season.
 
Any time you rank 340 or however many teams it is, you're going to miss on some. But for a system that ranks every team in the country against every other one, I think they do pretty well. Every ranking system you can think of, computer or human, is going to miss on teams. Its the nature of the beast.

I think I would like to see less emphasis on results against really bad teams though. Louisville is #1 right now in Pomeroy, mainly because they have killed some terrible teams (their first 5 games were all against teams ranked 200 or higher in the Pomeroy ratings). They've since piled up 2 more blowout wins against teams 200 or higher. In general I think MOV is important in evaluating teams, but maybe there should be some attempt to cap it against teams that are truly terrible.
RPI has Louisville at 25
 
RPI is pretty bad, at least this early in the season. I'd take Pomeroy over the RPI easily. Hell I'd probably take the AP Poll over the RPI.
 
That's another part of it - KenPom, RPI, whatever, are not really worth anything until later in the season.

This is correct about RPI, but incorrect about Pomeroy. I don't know enough about Saragin's model to say. In 2012 and 2013 SU's rank barely changed throughout the season, the difference between high and low rankings was 6 in 13 and 3 in 12. This year and 11 saw more volatility due to poor performances against weak teams early in the seasons.
 
Here is my TLDR post.

I like to use them all to complement each other - the AP, my eyes, KenPom (don't use Sagarin), and the RPI because its a necessary evil. To me the deeper you want to rank/follow feams outside of the top 10 in the country the more you need to rely on something like KP.

How many of us actually watch every team (or at least the top 75 teams) enough to from judgments on how they should be ranked, or at least what level that team is at? Nobody here. We can rank top 5 / top 10 based on what we have seen and results early on fairly well. But after that?

Lets get the semi garbage out of the way first - RPI. Its an awful ranking system, technically unsound, yet obviously relevant due to tournament implications. For Syracuse less so, but if you want to understand what conferences are lining up well for tourney seeds the RPI is extremely important in November and December. I am not looking at who is #1 in RPI, but what conferences are doing really well. The amusing thing is that conferences can game or just luck their way into better RPI rankings - it doesn't matter that the RPI stinks, the RPI matters (especially for tourney level teams not expecting top seeds)

The AP Poll / Eyes: Not necessarily the same, but I group them together because they have similar limitations. As I mentioned before you cant see them all. But at the same time most are able to focus and see the results of the top teams. We can come up with a logical top 5/top 10 with some viewing and some simple comp of records. We can come up with arguments who is #1 or #2, or top 10. Especially by this time of year, the teams with 0-1-2 losses are thinning out. But the rankings are often impacted by pre-season ranking - some good teams play no tough road games, while others play more. A factor not adequately captured in AP Polls, esepcialy as we move further from #5.

But will your eyes tell you who should be in the top 20? Top 30? Top 40? Do you trust the AP fully in these ranges with all their biases? Teams change a lot from year to year, and we don't see enough of most of them. Our observations are typically based on prior years - we have nothing else to go on. But teams can change a lot quickly in the NCAA.

So to me this is where KenPom came in, especially at this time of year. I think a lot of people struggle with KenPom because its not based on W/L. Its basically based on how many points you score and allow per possession adjusted to schedule.
For example. one 4 point loss vs a 4 point win will not change your KP ranking much. After all its 8 points over the course of all your possessions. But in the eyes of AP rankers (and in my vote if I had one), and especially when we are discussing the top 10 today, that decision (win or loss) has a huge impact on your ranking, and rightly so. KP does not rank top teams in the way we have been framed to evaluate the top 5/top 10 teams in America. It doesn't mean its flawed - the difference between a number 1 and number 5 team is typically minute, and the ranking system is meant to rank 350 teams not 10 teams.

How do you evaluate your more middle of the pack teams? There are a tonne of teams that are 8-3, 7-4, 9-3 for example, but givens schedules, there can be big difference in those teams. How do you know (or try to understand) what teams are in line for the tourney, before the conference season actually starts. How do you determine who may be underachievers without relying on prior observations/biases? Well the AP and eyes wont really help you past #15 IMO. So this is where a good ranking system comes in that has no biases.

Many of you don't really care beyond this level which is fine - teams outside of #15 are not the competitors of Syracuse at a national level. But how can you make blanket statements that Clemson sucks for example. None of us have seen enough of the middle of the pack.

Do I think Syracuse/Arizona or Louisville is #1? Syracuse/Arizona should be #1. Either way the cream has risen to the top in the poll or the KP system. But I would also argue that teams that are ranked way more higher in AP than KP, are the teams that are likely going to fade (for example Storrs) I am sure if KP was an AP voter his rankings would not be the same - he would also consider a close W or L more than his system. He would also say his rankings are not meant to be a full basis for his AP vote.

In order to adequately rank teams outside of the top 15 and to rank 350 teams, you need such a margin / efficiency based systems. W/L are not possible or you end up with ridiculous garbage like the RPI.
 
RPI is pretty bad, at least this early in the season. I'd take Pomeroy over the RPI easily. Hell I'd probably take the AP Poll over the RPI.

Yep the RPI sucks on an individual team basis, and sucks to rank teams. Y

Yet it is really, really important right now in terms of where you are as a conference as a whole, because that framcs everything for your RPI and the number of top 50 opportunities you have. There is nothing more fake than a power conference team getting a top 50 win at home (and that top 50 team only being there because its in a good conference). But thats the RPI and it matters.
 
In order to adequately rank teams outside of the top 15 and to rank 350 teams, you need such a margin / efficiency based systems. W/L are not possible or you end up with ridiculous garbage like the RPI.

One of the reasons RPI is officially used and Pomeroy is not is because there is a rule that the NCAA won't use any "computer" scoring system that factors in margin of victory. It is set up to prevent teams from running up the score.
 
That's another part of it - KenPom, RPI, whatever, are not really worth anything until later in the season.

That's not really accurate. Its pretty accurate right now in telling us what range teams are in, and where teams are headed vis a vis the tourney. There will be some deviations of course.

If you want the system to be accurate in telling you who will be #1, #2, #3, rather than who will be amongst the elite teams, your asking it to do things it was not intended to. As I said below, to accurately rate 350 teams you need to use an efficiency based system. Its going to create a few inconsistencies in terms of what we view as #1-#5, as the system cannot differentiate between a close W or close L.
 
Last edited:
How do you evaluate your more middle of the pack teams? There are a tonne of teams that are 8-3, 7-4, 9-3 for example, but givens schedules, there can be big difference in those teams. How do you know (or try to understand) what teams are in line for the tourney, before the conference season actually starts. How do you determine who may be underachievers without relying on prior observations/biases? Well the AP and eyes wont really help you past #15 IMO. So this is where a good ranking system comes in that has no biases.

Many of you don't really care beyond this level which is fine - teams outside of #15 are not the competitors of Syracuse at a national level. But how can you make blanket statements that Clemson sucks for example. None of us have seen enough of the middle of the pack.

Do I think Syracuse/Arizona or Louisville is #1? Syracuse/Arizona should be #1. Either way the cream has risen to the top in the poll or the KP system. But I would also argue that teams that are ranked way more higher in AP than KP, are the teams that are likely going to fade (for example Storrs) I am sure if KP was an AP voter his rankings would not be the same - he would also consider a close W or L more than his system. He would also say his rankings are not meant to be a full basis for his AP vote.

In order to adequately rank teams outside of the top 15 and to rank 350 teams, you need such a margin / efficiency based systems. W/L are not possible or you end up with ridiculous garbage like the RPI.


Good points, all. What Sagarin can do for bettors is provide a basis for evaluating such middle-of-the -pack teams on a relative basis. Let's not forget that one can make just as much money betting Clemson vs NC State as one can betting SU vs Duke. Arguably, even more, since there is less information about the middle-ranked teams, so the "market" is more inefficient.

And let's give Sagarin credit where it is due - in his PURE_ELO (which disregards MOV), Syracuse is #1.
 
One of the reasons RPI is officially used and Pomeroy is not is because there is a rule that the NCAA won't use any "computer" scoring system that factors in margin of victory. It is set up to prevent teams from running up the score.

Very true. You don't want teams running up the score. I would just like some tweaks to RPI, or some sort of tweaking of margins. Easier said than done.

I don't like some of the artificial things that come out of the RPI.
 
Its going to create a few inconsistencies in terms of what we view as #1-#5, as the system cannot differentiate between a close W or close L.
Right, and this is where I think it causes a lot of problems for some people. If you have 2 teams that play on a neutral court and one team wins by a point, a rating system is going to essentially say that the teams are even. But a lot of people are going to look at it and say Team A won, so they're clearly better than Team B, even if the margin is essentially nothing. A team like Kansas is a good example of this; they've lost close games to good teams on neutral/road locations, so just looking at the record shows 3 losses and a team that shouldn;t be considered elite, but I bet that is going to look wrong in 2 months.
 
The good teams—the truly good teams—manage to win games and manage to not lose games to inferior teams. Regardless of how much they might win by or lose by. For my money, that is all that matters—who did you beat and who did you lose to.
KennyP, Vegas and Sagarin be damned.
:)
 
The good teams—the truly good teams—manage to win games and manage to not lose games to inferior teams. Regardless of how much they might win by or lose by. For my money, that is all that matters—who did you beat and who did you lose to.
KennyP, Vegas and Sagarin be damned.
:)

Fair enough. That is how you have the rank the elite, the great teams like Syracuse.

But how else will we know where UConn ranks in a few years? Somebody has to rank the irrelevant teams in the 100's
 
But how else will we know where UConn ranks in a few years? Somebody has to rank the irrelevant teams in the 100's
:D
Well played, sir.
 
Sagarin has little meaning until later in the season when many more games have been played and teams have been "connected".
 
Sagarin has little meaning until later in the season when many more games have been played and teams have been "connected".
but even at the end of the year the data set is still too small, better, but still too small
 

Forum statistics

Threads
167,141
Messages
4,682,374
Members
5,900
Latest member
DizzyNY

Online statistics

Members online
301
Guests online
1,403
Total visitors
1,704


Top Bottom