Sagarin and Pom ratings (updated 3/4/09)

[robed deity]

From looking at numbers over the years, it seems rare that a national champion does not have a high KenPom ranking. In fact, since 1999, only once did a team win the national championship and end up ranked outside of the top 2 (Syracuse, 5th in 2003). Granted, these rankings I am looking at from Pomeroy's site were the final rankings, and therefore a result of the tournament games, so it is difficult to say exactly where these teams were before the start of the NCAA's. But, then again, its hard to climb up through the efficiency rankings so quickly in just 6 games.

Also, the eventual national champion usually has a top ten ranking in both offense and defense. I seem to remember an article on CNNSI (Luke Winn maybe?) that illustrated this with numbers from the past decade or so. On Pomeroy's site, it only goes back 4 years or so with specific offense and defense ratings, so Im not sure where Winn got his info. Just looking at the past 4 years, Florida from '07 was the only team without a top ten rating in O AND D, and its D was on the cusp at 12th.

Interestingly, guess who is the only team presently with an O and D rating in the top 10?
Thats right, our Blue Devils (at 6th and 8th, respectively). But then again, if any year was to break the trend of having a high efficiency rating and winning the title, it would be this one. There just aren't any great teams out there, and the current KenPom numbers show this. For instance, Oklahoma's current D rating (48th) in most years would eliminate them from championship or final four talk, but I think any reasonable mind would contend they have a legit chance.

[rsvman]

For the last two seasons I have filled out four brackets. One goes straight seeds, one on RPI, on one Sagarin, and one on my own picks. Each year my own picks have beat all 3 other methods.

I have not yet tried using Pomeroy. Guess I'll have to start that this year. Possibly Pomeroy would finally be the system that could out-predict me.

[mehmattski]

As a follow up, here are the results of Pomeroy's system making predictions just prior to the tournament in 2007:

South: Pomeroy made the correct Ohio State pick for the final four. Memphis was given a slightly worse chance of making the Elite Eight than Texas A&M. Those three teams and Tennessee were given the best chances at Sweet Sixteen berths, and all were correct.

East: North Carolina (49%) and Georgetown (32%) were given the bulk of the odds to make the final four, and indeed did match up in a classic elite eight game. The other sweet sixteen participants, USC and Vanderbilt, were not heavily favored to make that round; Texas and WSU were given the nod.

Midwest: Pomeroy predicted this would be a "wide open region, with six teams given at least a 5% chance of making Atlanta." In the end, only two major upsets happened in that bracket (Winthrop over Notre Dame and UNLV over Wisconsin). Had you filled out your bracket according to Pomeroy, the sweet sixteen would have been Florida-Maryland-Wisconsin-Notre Dame, and only one of those teams made it through. Pomeroy did not see UNLV coming.

West: The top four seeds in this region made the Sweet Sixteen, but Pomeroy did not get it right. His top four percentages for that round went to Kansas-UCLA-Duke-Southern Illinois, and in fact Pitt was given a better chance than the Salukis.

North Carolina was given the best chance at a title, followed by Kansas. Pomeroy got half the final four right, and six of the elite eight.

For the past two years, that gives him 6 of 8 final four teams (75%) and 12 of 16 elite eight teams (also 75%). How did pre-tournament Sagarin and RPI do?

[MChambers]

The rankings at the top are more compressed, if you will, than has been the case since 2006. That is, the top teams have lower Pythagorean scores than in 2007 and 2008. The difference looks significant to me, but what do I know. I assume this compression means that the NCAA tournament will be more wide open, with fewer favorites reaching the Final Four.

[hurleyfor3]

I think some people here are confusing cause and effect.

The prior Pomeroy polls include the NCAA Tournament games, and are weighted towards the most recent games played. So naturally the national champion will end up with a high rating. Also, the 2007 and 2008 FFs had all high-seeded teams playing (1s and 2s), whereas no #1 made the Final Four in 2006. So the Pomeroy season-end ratings from these years will be top-heavy.

This year's tournament hasn't happened yet, so it's not entirely valid to compare the current rankings to those of prior years.

[MChambers]

I think some people here are confusing cause and effect.

The prior Pomeroy polls include the NCAA Tournament games, and are weighted towards the most recent games played. So naturally the national champion will end up with a high rating. Also, the 2007 and 2008 FFs had all high-seeded teams playing (1s and 2s), whereas no #1 made the Final Four in 2006. So the Pomeroy season-end ratings from these years will be top-heavy.

This year's tournament hasn't happened yet, so it's not entirely valid to compare the current rankings to those of prior years.

You are right, of course, but I still think the top teams at this point in the season are closer together and to the rest of the world than often has been the case. I guess there is no way to prove it, without some exhausting research.

[MIKESJ73]

Pre Tournament RPI of the last five final fours:

2004: #1, #5, #6, #16 #5 won
2005: #2, #6, #12, #21 #6 won
2006: #9, #13, #15, #26 #15 won
2007: #1, #2, #6, #9 #6 won
2008: #1, #3, #4, #5 #5 won

2008 was as much of an anamoly as 2006.

On a side note, Duke is now #1 in RPI

[camion]

But, but, but, the RPI is not designed to rank teams according to their relative strength or to predict winners. That is not its function. Sagarin and Pomeroy are better designed for that purpose.

RPI is an indicator that a team plays pretty well and plays a strong schedule. For the RPI it's the schedule that really matters.

[gotham devil]

Pre Tournament RPI of the last five final fours:

2004: #1, #5, #6, #16 #5 won
2005: #2, #6, #12, #21 #6 won
2006: #9, #13, #15, #26 #15 won
2007: #1, #2, #6, #9 #6 won
2008: #1, #3, #4, #5 #5 won

2008 was as much of an anamoly as 2006.

On a side note, Duke is now #1 in RPI

http://realtimerpi.com/rpi_Men.html

I was glad to see we've now passed Pitt for #1 in RPI.

[mehmattski]

I think some people here are confusing cause and effect.

The prior Pomeroy polls include the NCAA Tournament games, and are weighted towards the most recent games played. So naturally the national champion will end up with a high rating. Also, the 2007 and 2008 FFs had all high-seeded teams playing (1s and 2s), whereas no #1 made the Final Four in 2006. So the Pomeroy season-end ratings from these years will be top-heavy.

This year's tournament hasn't happened yet, so it's not entirely valid to compare the current rankings to those of prior years.

That is why I cited Pomeroy's own pre-tourney predictions for 2008 and 2007. He was pretty darn good at predicting from the Sweet Sixteen on in both years, based on his own rankings. These would be based on a team's performance immediately following the conference tournaments in those years, which avoids the bias you're talking about.

[dcarp23]

One really interesting trend is that since Elliot Williams' introduction into the starting lineup, Duke's defensive efficiency has gone down and its offensive efficiency has gone up. I'd bet if you asked the casual observer, he would have said otherwise. Duke is now the fifth most efficient offensive team in the country according to Pomeroy.

If I had to guess the reason for the change, I'd point to two factors: 1) Scheyer at the point and the complete disappearance of turnovers. 2) An increase in steals on the defensive end leading to a significant increase in runouts and easy baskets that were nonexistent earlier in the season.

[MChambers]

One really interesting trend is that since Elliot Williams' introduction into the starting lineup, Duke's defensive efficiency has gone down and its offensive efficiency has gone up. I'd bet if you asked the casual observer, he would have said otherwise. Duke is now the fifth most efficient offensive team in the country according to Pomeroy.

If I had to guess the reason for the change, I'd point to two factors: 1) Scheyer at the point and the complete disappearance of turnovers. 2) An increase in steals on the defensive end leading to a significant increase in runouts and easy baskets that were nonexistent earlier in the season.

Has defensive efficiency gone down? Williams seems to be playing good man defense, although perhaps he is not so good on help defense (most freshmen aren't). Or is Nolan that much better? Or is it that Zoubek's minutes have decreased? Or something else more subtle?

[tbyers11]

One really interesting trend is that since Elliot Williams' introduction into the starting lineup, Duke's defensive efficiency has gone down and its offensive efficiency has gone up. I'd bet if you asked the casual observer, he would have said otherwise. Duke is now the fifth most efficient offensive team in the country according to Pomeroy.

If I had to guess the reason for the change, I'd point to two factors: 1) Scheyer at the point and the complete disappearance of turnovers. 2) An increase in steals on the defensive end leading to a significant increase in runouts and easy baskets that were nonexistent earlier in the season.

It's all relative.

Duke's defensive efficiency in the last 5 starting with the St John's game has averaged 105.72, according to Ken Pom. While this is a considerably worse defensive efficiency that the season as a whole at 88.7, it is considerably better than the 114.88 mark in the 4 games (@Clem, UVa, UNC, @BC) that preceded St. Johns and were the impetus for the lineup change.

I don't think that we are playing as good of defense in the last 5 games as earlier in the season, but when you factor in the level of competition and the familiarity of conference foes with each other I don't think the difference is that large and I definitely feel it is a big improvement over the 4 games before Elliot's insertion into the starting lineup.

I agree with you and am excited by the somewhat unexpected and significant bump in offensive efficiency in the last 5 games.

[dcarp23]

Has defensive efficiency gone down? Williams seems to be playing good man defense, although perhaps he is not so good on help defense (most freshmen aren't). Or is Nolan that much better? Or is it that Zoubek's minutes have decreased? Or something else more subtle?

Great question--it would be nice to be able to look at the stats before the St. John's game (or before the Clemson game) and see what stats changed. It seems to me that the change began when teams made it their game plan to get into the paint and to the rim, which had not happened at the beginning of the year. The cause of that has been debated here ad nauseam, obviously.

That said, Duke's D still ranks 12th in the country, which isn't too bad.

[Chitowndevil]

Great question--it would be nice to be able to look at the stats before the St. John's game (or before the Clemson game) and see what stats changed. It seems to me that the change began when teams made it their game plan to get into the paint and to the rim, which had not happened at the beginning of the year. The cause of that has been debated here ad nauseam, obviously.

That said, Duke's D still ranks 12th in the country, which isn't too bad.

It isn't good either, considering (1) Duke had been ranked either #1 or #2 in DE pretty much continuously through the end of January, and (2) through their last nine games, Duke has had only one (at VT) in which an opponent had an effective FG% (like FG%, but gives extra weight to made 3s) below 50%. Before that, Duke had allowed an opponent eFG% above 50% only 5 times in 21 games.

Of course you can't argue with the offensive numbers since the lineup change: in its last five games, Duke has turned the ball over on less than 15% of possessions, and shot 2 free throws for every field goal attempted. Both of those are phenomenal figures and well above Duke's season averages (yes it's a sample of only five games, but to put up those kind of numbers late in the season against quality competition is still very encouraging).

My own opinion is the decline in Duke's D has less to do with Williams and Scheyer than with Nolan Smith, who was tremendously effective pressuring the ball early in the year but fell off in mid January. I suspect this has more to do with Nolan's health than his game, but that is conjecture. But since I'm stating opinion at this point, I might as well say I don't think Duke will make a deep March run with the kind of defensive numbers they've put up recently. However, if Nolan Smith can get back to his early January form and Duke can use Williams and Smith interchangably and/or together to pressure the ball, I like Duke's chances.

[calltheobvious]

It isn't good either, considering (1) Duke had been ranked either #1 or #2 in DE pretty much continuously through the end of January, and (2) through their last nine games, Duke has had only one (at VT) in which an opponent had an effective FG% (like FG%, but gives extra weight to made 3s) below 50%. Before that, Duke had allowed an opponent eFG% above 50% only 5 times in 21 games.
Of course you can't argue with the offensive numbers since the lineup change: in its last five games, Duke has turned the ball over on less than 15% of possessions, and shot 2 free throws for every field goal attempted. Both of those are phenomenal figures and well above Duke's season averages (yes it's a sample of only five games, but to put up those kind of numbers late in the season against quality competition is still very encouraging).

My own opinion is the decline in Duke's D has less to do with Williams and Scheyer than with Nolan Smith, who was tremendously effective pressuring the ball early in the year but fell off in mid January. I suspect this has more to do with Nolan's health than his game, but that is conjecture. But since I'm stating opinion at this point, I might as well say I don't think Duke will make a deep March run with the kind of defensive numbers they've put up recently. However, if Nolan Smith can get back to his early January form and Duke can use Williams and Smith interchangably and/or together to pressure the ball, I like Duke's chances.

I'm not yet near fluency in Pomerese, but isn't there a metric, maybe 'adjusted efg%' that takes possessions into account, i.e. even if your opponent is converting a high percentage of its actual attempts, there should be a stat that deflates that number if say, the team is turning the ball over a lot.

[fogey]

Our defensive ratings are much more influenced by the quality of the opposition than to our lineup changes, in my view. Elliot is a terrific defender, and his minutes on the floor have certainly not hurt our defense. But the early season patsies are long gone, and it is much tougher to defend better teams successfully.

[tbyers11]

I'm not yet near fluency in Pomerese, but isn't there a metric, maybe 'adjusted efg%' that takes possessions into account, i.e. even if your opponent is converting a high percentage of its actual attempts, there should be a stat that deflates that number if say, the team is turning the ball over a lot.

Here is Ken Pom's Duke Game Plan page and here is the explanation of what the column headings mean.

Defensive efficiency (Eff. under Defense) is the metric that I think addresses overall defense the best. It is points allowed/# of possessions. Anything below 100 in Pomeroyese is generally considered good. Anything below 90 is really good.

The four columns to the left (effective FG% (eFG%), turnover rate (TO%), offensive rebounds allowed rate (OR%) and free throws allowed rate (FTR) break down the defense into specific areas that can tell you more specifically why the team is excelling or failing at defense. TO% directly answers your question about by showing how of the opposing teams possessions become turnovers. While everyone wants to have as low an eFG% as possible just using that stat doesn't give you the whole defensive picture, TO%, OR% and FTR fill out that picture.

On Ken Pom's scouting report for Duke you can look at the four factors for the defense (in the upper left) and see how Duke ranks amongst all DI teams. The data suggests that while Duke is good at eFG% defense (81st is still top 25% of all teams) is it is really good at forcing turnovers (20th best in D1).

Hope that helps a bit