Dork Poll tracking (Pomeroy-Sagarin)

[BattierBattalion]

As I wrote in an email earlier today--

Sullinger is the nation's best freshman. If I had to put money on anyone other than Duke to win the national title, it would be The Ohio State University. I think they are the clear #2 and the team I fear the most.

-Jason "Thad has really built a quality program... and his recruiting is studly" Evans

They scored 93 points on 65 possessions at Florida. That's incredible.

[loran16]

Well. After Yesterday Duke remains #1 (and #1 in both Offensive and Defensive efficiency), while UNC dropped to #8 (they'd been bumped up due to the Hofstra game.)

Meanwhile, potential opponent KState dropped to #20 (#24 on both O and D) after it's horrible game to Presbyterian.

[MChambers]

Well. After Yesterday Duke remains #1 (and #1 in both Offensive and Defensive efficiency), while UNC dropped to #8 (they'd been bumped up due to the Hofstra game.)

Meanwhile, potential opponent KState dropped to #20 (#24 on both O and D) after it's horrible game to Presbyterian.

You're right, as far as Pomeroy goes, but somehow Duke dropped in Sagarin, at least in the blended rating. Still #1 in the Predictor, but #3 in the Rating.

It's over.

[budwom]

Pomeroy and Sagarin are fun to check out, but seriously, this early in the season these kind of ratings mean zipsquat. In another ten days we'll have some much more meaningful data to pick over.

[77devil]

You're right, as far as Pomeroy goes, but somehow Duke dropped in Sagarin, at least in the blended rating. Still #1 in the Predictor, but #3 in the Rating.

It's over.

Likely because the ELO Chess component considers the strength of the opponents but not the margin of victory. Pitt and WVU played strong teams in the their last game(or two for Pitt), while Duke has played cupcakes thus far. That will change dramatically starting Monday.

As Budworm notes below, as does Sagarin, the ratings are not statistically meaningful(aka theory of "zipsquat") at this point.

Pomeroy and Sagarin are fun to check out, but seriously, this early in the season these kind of ratings mean zipsquat. In another ten days we'll have some much more meaningful data to pick over.

[tieguy]

Jason "sadly, Ken has no archived numbers for the woeful 2002 UNC club" Evans

Google cache has partial caches of Ken's data from 2002. Looks like they were 136 at the end of the season, though I seem to recall his methodology changed significantly in 200(4?) and so these numbers aren't directly comparable.

[tieguy]

Likely because the ELO Chess component considers the strength of the opponents but not the margin of victory.

As Sagarin puts it, ELO is "politically correct"; I seem to recall that he only does it because otherwise he can't participate in the BCS rankings.

Note, though, that all he considers in the "better" rankings is the score. There is no understanding of the underlying structure of the game, which makes his numbers much less useful when comparing two teams who play the game at very different paces. Pomeroy takes that into account. So really Pomeroy > Sagarin Predictor >> Sagarin ELO.

[JasonEvans]

Rankings as of Monday 11/22

Team         Sag   Pom   Dif
Duke           7     1    -6
UNC           25    13   -12
Clemson       34    30    -4
Fla St        36    27    -9
Maryland      43    29   -14
NC State      49    37   -12
Va Tech       55    39   -16
Miami         94    60   -34
Virginia      97    68   -29
BC           113    52   -61
Ga Tech      121    81   -40
Wake         146   138    -8

What jumps out is how much higher Pomeroy thinks of the ACC than Sagarin does. There is not a single team that is ranked higher in Sag than it is in Pom. In some cases, the differences are quite stark. Pom sees Miami and BC as a bubble teams while Sag has them barely in the hunt for an NIT bid. Time will tell, but I have to say that the performance of the ACC thus far does not make me confident about the quality of the conference.

Now, here is how each team moved from week to week, starting with Sagarin.

Sagarin     Today   Last wk   Change
Duke           7       1        -6
UNC           25       9       -16
Clemson       34      20       -14
Fla St        36      32        -4
Maryland      43      29       -14
NC State      49      70       +21
Va Tech       55      51        -4
Miami         94      46       -48
Virginia      97      58       -39
BC           113      63       -50
Ga Tech      121      53       -69
Wake         146      90       -56

Wow! Every single team in the conference moved down except NC State. That is a bad week! Many of the moves were dramatic. There are 5 ACC teams, almost half the conference, ranked lower than 90. That's stunning and quite unusual for this conference.

Pomeroy     Today   Last wk      Change
Duke           1        1          --
UNC           13       16          +3
Fla St        27       27          --
Maryland      29       23          -6
Clemson       30       29          -1
NC State      37       41          +4
Va Tech       39       25         -14
BC            52       38         -14
Miami         60       46         -14
Virginia      68       57         -11
Ga Tech       81       78          -3
Wake         138      132          -6

I don't even know what to say about a ranking that thinks more highly of Carolina today than it did a week ago. I generally trust Ken's rankings a great deal but I find it mystifying that he could still be ranking UNC this high. Still, the obvious trend for the conference as a whole is clear. It was a bad week for the conference as most ACC teams went down in the rankings.

Well, that's all for now. As always, I welcome more analysis and discussion of this fun (but dorky) subject. Lets hope the conference fares a bit better in the coming week!

-Jason "Wake sucks royally, at least we can all agree on that!" Evans

[JasonEvans]

One more thing worth noting -- the ACC was the #1 conference in Pomeroy's rankings last week. They are now #2 and the gap to the Big Ten is quite large. Sagarin had the ACC as the #2 conference a week ago. We are now #5. Blech!!

-Jason "aside from Duke, this is not looking like a very strong league" Evans

[COYS]

Rankings as of Monday 11/22

Team         Sag   Pom   Dif
Duke           7     1    -6
UNC           25    13   -12
Clemson       34    30    -4
Fla St        36    27    -9
Maryland      43    29   -14
NC State      49    37   -12
Va Tech       55    39   -16
Miami         94    60   -34
Virginia      97    68   -29
BC           113    52   -61
Ga Tech      121    81   -40
Wake         146   138    -8

What jumps out is how much higher Pomeroy thinks of the ACC than Sagarin does. There is not a single team that is ranked higher in Sag than it is in Pom. In some cases, the differences are quite stark. Pom sees Miami and BC as a bubble teams while Sag has them barely in the hunt for an NIT bid. Time will tell, but I have to say that the performance of the ACC thus far does not make me confident about the quality of the conference.

Now, here is how each team moved from week to week, starting with Sagarin.

Sagarin     Today   Last wk   Change
Duke           7       1        -6
UNC           25       9       -16
Clemson       34      20       -14
Fla St        36      32        -4
Maryland      43      29       -14
NC State      49      70       +21
Va Tech       55      51        -4
Miami         94      46       -48
Virginia      97      58       -39
BC           113      63       -50
Ga Tech      121      53       -69
Wake         146      90       -56

Wow! Every single team in the conference moved down except NC State. That is a bad week! Many of the moves were dramatic. There are 5 ACC teams, almost half the conference, ranked lower than 90. That's stunning and quite unusual for this conference.

Pomeroy     Today   Last wk      Change
Duke           1        1          --
UNC           13       16          +3
Fla St        27       27          --
Maryland      29       23          -6
Clemson       30       29          -1
NC State      37       41          +4
Va Tech       39       25         -14
BC            52       38         -14
Miami         60       46         -14
Virginia      68       57         -11
Ga Tech       81       78          -3
Wake         138      132          -6

Wow. I don't even know what to say about a ranking that thinks more highly of Carolina today than it did a week ago. I generally trust Ken's rankings a great deal but I find it mystifying that he could still be ranking UNC this high. Still, the obvious trend is clear. It was a bad week for the conference as most ACC teams went down in the rankings.

Well, that's all for now. As always, I welcome more analysis and discussion of this fun (but dorky) subject. Lets hope the conference fares a bit better in the coming week!

-Jason "Wake sucks royally, at least we can all agree on that!" Evans

I think Pomeroy at least gives UNC credit for blowing out Hofstra and posting decent defensive and offensive efficiency stats against other weak competition. Pomeroy is less concerned with wins and losses than overall efficiency. Minnesota, for example, has been pretty poor on offense and just decent enough on defense to have a pretty mediocre offensive/defensive efficiency differential, despite beating UNC. I think Pomeroy still sees a higher ceiling for UNC, which is probably true. That being said, if UNC continues to struggle against quality opponents, Kenpom's preseason predictors and the Heels' early season success against cupcakes will stop bailing them out.

The historical data Pomeroy uses is probably what is keeping the ACC afloat despite a horrible early season. It doesn't help that NCState had to play their toughest game without their veteran star.

[DoubleDuke Dad]

Rankings as of Monday 11/22

Pomeroy     Today   Last wk      Change
Duke           1        1          --
UNC           13       16          +3
Fla St        27       27          --
Maryland      29       23          -6
Clemson       30       29          -1
NC State      37       41          +4
Va Tech       39       25         -14
BC            52       38         -14
Miami         60       46         -14
Virginia      68       57         -11
Ga Tech       81       78          -3
Wake         138      132          -6

I don't even know what to say about a ranking that thinks more highly of Carolina today than it did a week ago.

-Jason "Wake sucks royally, at least we can all agree on that!" Evans

How many more losses in a row will it take before they are ranked #1?

[pfrduke]

The historical data Pomeroy uses is probably what is keeping the ACC afloat despite a horrible early season. It doesn't help that NCState had to play their toughest game without their veteran star.

This is the first season that the Pomeroy ratings have involved "bias" early in the season, rather than rely only on this season's numbers. My guess is that there will be some kinks that need to work themselves out, and it will improve over the next couple of years. But I think you're right that the bias is inflating the ACC's numbers.

[tieguy]

I don't find the Pomeroy/Sagarin delta all that interesting (especially not at this point in the season), but a top-25 poll differential would be very interesting- there must be some teams that the coaches and media are over/underrating in the opinion of the dorks.

FWIW, I'm also working on shoving all this into Google Docs to visualize the change over the course of the season, but so far not enough data to make it work.

[sagegrouse]

Congratulations to the OP and the posters for keeping perspective and making reflective and not heated comments. What Sagarin and KenPom are doing is to rig up early season projections for their data-driven models, when -- quite frankly -- there is no data.

Tune back in January 15.

sagegrouse

[camion]

At this point in the season the dorks are using EWAGs in their rankings just like the pollsters. I start paying attention to Kenpom and Sagarin about the time serious conference play starts in January.

Sagarin's version of the current dork disclaimer is:
"For the first few weeks of the season, the starting ratings have weight
in the process(BAYESIAN), but once the teams are all WELL CONNECTED, then
the starting ratings are no longer used and all teams are started equal
and the ratings are then done in an UNBIASED manner from that point on.
The teams are now NOT WELL CONNECTED and so the ratings are BAYESIAN."


Finally, I don't consider the RPI an accurate indicator of teams' relative strengths, ever. That isn't it's purpose. The formula shown below tells you something about a team's strength, but mainly it penalizes teams (like Va Tech) that play cupcake schedules.

From Wikipedia:
The current and commonly used formula for determining the RPI of a team at any given time is as follows.
RPI = (WP * 0.25) + (OWP * 0.50) + (OOWP * 0.25)
where WP is Winning Percentage, OWP is Opponents' Winning Percentage and OOWP is Opponents' Opponents' Winning Percentage.

[loran16]

At this point in the season the dorks are using EWAGs in their rankings just like the pollsters. I start paying attention to Kenpom and Sagarin about the time serious conference play starts in January.

Sagarin's version of the current dork disclaimer is:
"For the first few weeks of the season, the starting ratings have weight
in the process(BAYESIAN), but once the teams are all WELL CONNECTED, then
the starting ratings are no longer used and all teams are started equal
and the ratings are then done in an UNBIASED manner from that point on.
The teams are now NOT WELL CONNECTED and so the ratings are BAYESIAN."


Finally, I don't consider the RPI an accurate indicator of teams' relative strengths, ever. That isn't it's purpose. The formula shown below tells you something about a team's strength, but mainly it penalizes teams (like Va Tech) that play cupcake schedules.

From Wikipedia:
The current and commonly used formula for determining the RPI of a team at any given time is as follows.
RPI = (WP * 0.25) + (OWP * 0.50) + (OOWP * 0.25)
where WP is Winning Percentage, OWP is Opponents' Winning Percentage and OOWP is Opponents' Opponents' Winning Percentage.

No one is suggesting the Pomeroy or Sagarin rankings are in any way predictive right now, but they're interesting. Also, while they do take into account right now early season projections, at least Pomeroy's is based upon a statisical system, and both of their adjustments right now are done not by gut reaction (like the pollsters) but due to their mathematical formulas.

So if anything, they're at least as relevant as the current Top 25 polls, and Pomeroy is even more interesting, because it makes evaluations of specific factors of teams (O and Defensive Efficiency, Tempo, etc.).

No one is suggesting that they're in any way definitive or a really accurate method of classifying who the best team is yet, but they're better than the polls and the movements are interesting.

[MChambers]

No one is suggesting the Pomeroy or Sagarin rankings are in any way predictive right now, but they're interesting. Also, while they do take into account right now early season projections, at least Pomeroy's is based upon a statisical system, and both of their adjustments right now are done not by gut reaction (like the pollsters) but due to their mathematical formulas.

So if anything, they're at least as relevant as the current Top 25 polls, and Pomeroy is even more interesting, because it makes evaluations of specific factors of teams (O and Defensive Efficiency, Tempo, etc.).

No one is suggesting that they're in any way definitive or a really accurate method of classifying who the best team is yet, but they're better than the polls and the movements are interesting.

Pomeroy's new system for starting ratings tries to take into account new players. One potential weakness in that is that freshmen in most cases will need time to adjust to college play. In other words, they'll be a lot better in March than November, but Pomeroy's new system may not reflect that in November.

[loran16]

Pomeroy's new system for starting ratings tries to take into account new players. One potential weakness in that is that freshmen in most cases will need time to adjust to college play. In other words, they'll be a lot better in March than November, but Pomeroy's new system may not reflect that in November.

This applies for all players, not just freshmen. If pomeroy is overrating the freshmen at first, the teams in question will drop down as their results show that and then rise up as the freshmen reach potential. Same with other players. Really shouldn't make a difference...the ratings are NOT supposed to indicate how good teams are in March right now, but how good they are right now. They're not particularly reliable right now, but they're better than other things.

[JasonEvans]

Another thing to note is that Pomeroy has indicated that as he gets more and more data, he weights the pre-season SWAG less and less. He says the SWAG won't be completely gone until early January but that it will account for just a tiny fraction of a percent of a team's total ranking by the time we get to January.

At this point, with most teams on game #4 or so, there is obviously a strong weighting of the SWAG, but it is diminishing and, I would bet, it won't be a very significant factor any more in just a couple weeks.

I can absolutely compare the Sag and KenPom numbers to the AP and ESPN polls. I'll start that next week... or maybe sooner.

I also love that tieguy (a welcome old-timer who is coming back STRONG!!) is going to try to put all this into a graph. In think it could be very interesting to watch. I wonder a bit about how it is going to look though with one team at #1 and another at #150. I would imagine it is going to be hard to see real movement on a graph where there is that wide a gap in the values of the data.

--Jason "as always, thanks to everyone for participating in this dorky exercise " Evans

[loran16]

Another thing to note is that Pomeroy has indicated that as he gets more and more data, he weights the pre-season SWAG less and less. He says the SWAG won't be completely gone until early January but that it will account for just a tiny fraction of a percent of a team's total ranking by the time we get to January.

At this point, with most teams on game #4 or so, there is obviously a strong weighting of the SWAG, but it is diminishing and, I would bet, it won't be a very significant factor any more in just a couple weeks.

I can absolutely compare the Sag and KenPom numbers to the AP and ESPN polls. I'll start that next week... or maybe sooner.

I also love that tieguy (a welcome old-timer who is coming back STRONG!!) is going to try to put all this into a graph. In think it could be very interesting to watch. I wonder a bit about how it is going to look though with one team at #1 and another at #150. I would imagine it is going to be hard to see real movement on a graph where there is that wide a gap in the values of the data.

--Jason "as always, thanks to everyone for participating in this dorky exercise " Evans

Eh, last week The Preseason projections counted for 5 games. He says he's reducing the impact every week. So by next week, the preseason projections are gong to be outweighed byt he acutal games.