NET News (March 4): The "top 8" separates itself, and a comparison to the AP

[scottdude8]

Despite Duke's loss to Va. Tech this week, the Blue Devils have held steady at No. 3 in the NET Rankings (https://www.ncaa.com/rankings/basketball-men/d1/ncaa-mens-basketball-net-rankings). Despite the chaos of Saturday (MSU losing to lowly Indiana, Tennessee throttling Kentucky, and mid-major stalwarts Nevada and Houston falling), the rest of the top 10 hasn't shifted too much either. That, combined with another loss by Marquette weakening their resume (and showing that the Big East as a conference is down pretty significantly this season) leads me to believe that the top eight teams have separated themselves from the field when it comes to NCAA seeding, at least according to the NET.

FWIW, the Bracketologists seem to agree, at least according to Bracket Matrix (http://www.bracketmatrix.com/): the lowest 2 seed (Michigan) there is an average 2.16 seed (more or less meaning that they're a 2 seed in almost every bracket, with a few 3 seed projections), while the highest 3 seed (Houston) is an average 2.99 seed (meaning that they're almost always a 3 seed). That separation will probably get larger as updated projections reflecting Houston's loss come out.

This separation is also seen in the NET rankings, in which the top 8 seeds (Virginia, Duke, Gonzaga, Kentucky, Tennessee, UNC, MSU and Michigan) have also distinguished themselves. These teams are all in the top 9 of the NET (Houston still finds itself at No. 6 even after the bad home loss). Excluding Gonzaga (as their quadrant resume is skewed by their bad conference), these teams all have seven or more Q1 wins (it's actually 8 or more if you exclude Tennessee, the low man at 7). The only other teams to approach that level of a Q1 resume all have major flaws on their team sheets: Purdue has 7 Q1 wins, but 6 Q1 losses, not to mention a bad Q3 loss to Notre Dame; LSU has 9 Q1 wins, but have 3 Q2 losses and 5 overall; any other team with a comparable number of Q1 wins has at least 6 losses overall (Marquette is the team at 6 losses, if you eliminate them it's actually 7). I'd speculate that of these teams, LSU has the best shot at sneaking onto the two line if they can make some noise in the SEC tourney, especially considering that would likely include a victory over UK and/or Tennessee. However, the fact that the NET has LSU at No. 13 despite their strong Q1 record (and record overall) implies that the analytics don't love the Tigers. Nonetheless, the NET, the team sheets, and the Bracketologists all seem to agree that, at the moment, there's a clear separation between the No. 1 and No. 2 seeds and the rest of the field.

With that in mind, I thought it'd be illustrative to take another look at how the NET compares to the AP rankings and see which system we think best matches up with our intuition as to the relative strength of these teams. Here are some illustrative examples:

UNC is No. 3 in the AP poll, but No. 7 in the NET. This one seems fairly obvious to me: the human bias present in the AP poll means that more recent results are weighed more heavily. Thus, a UNC team with five losses is ranked higher in the AP poll than numerous teams with less losses, not to mention better tourney resumes, than the Tar Heels. In my view the NET trumps the AP in this case because it evenly weighs the whole season as opposed to focusing on more recent results. FWIW, UNC would be No. 5 in the RPI.

Houston is No. 12 in the AP poll, but No. 6 in the NET. Again, there's a fairly obvious explanation here: Houston has been able to inflate a lot of their "analytic" measurements playing in an inferior conference, which affects their NET ranking but can be corrected for by human observation in the AP poll. Here the human AP poll is probably superior as it is better at identifying teams that may be outliers based on their schedules. FWIW, Houston would be No. 6 in the RPI.

Marquette is No. 16 in the AP poll, but No. 26 in the NET. This one is a bit more interesting: Marquette has a multitude of quality wins on their team sheet, and 6 losses puts them right with the "second tier" of teams. But the NET doesn't seem to like them nearly as much as the human rankings. My theory: the computers are doing a better job of picking up that the Big East is having a very down year, while humans have been impressed by Marcus Howard and the "name recognition" of the Big East. FWIW, Marquette would be No. 20 in the RPI.

Finally, Kansas State is No. 18 in the AP poll, but No. 28 in the NET. I think this example brings together a few of the ideas mentioned above: Kansas State has two bad losses, at Tulsa and home against Texas, which happened months ago an are probably out of the minds of human voters. Meanwhile, KSU has won 7 of their last 9 in the Big 12, a conference which still brings with it a lot of "name recognition" but according to the NET metrics has struggled this year (the best team is NET 10 Texas Tech, and there are only three teams in the NET Top 25). There's also the factor of Dean Wade's injury which may make some voters disregard some of their losses. So here a combination of recency bias and human bias towards big name conferences may be inflating KSUs AP ranking. FWIW, KSU would be No. 19 in the RPI.

These are just a handful of dozens of examples you could analyze comparing the NET, AP, and RPI. But I find it interesting that in most of these cases the AP poll tracks better with the reviled RPI than the NET (there are some major outliers of course that I didn't go into heree... Kansas would STILL somehow be No. 1 in the RPI if it were in use this year). How you view the success of the NET at this point probably hinges on what your opinions are of teams like these where there's a big difference in their NET rankings from the RPI/AP. I would argue that the NET is a clear, but imperfect, improvement over the RPI, and seems to be doing its job in creating a more objective tool for analyzing teams as opposed to the subjectivity involved in human rankings.

I imagine that last point will be debated heartily. Have at it!

[Ian]

It's not surprising that RPI tracks better with AP poll. Both are more W/L based and don't really take margin of victory, efficiency, etc into account.

[BandAlum83]

Is there a scenario where Duke, UVA and uNC all get #1 seeds in the NCAAT?

[Wahoo2000]

Is there a scenario where Duke, UVA and uNC all get #1 seeds in the NCAAT?

Prob requires some weird losses by a combo of KY/Tenn/MSU/Michigan combined with Duke/UNC/UVA only losing to each other from here on out, and the loser of UNC/Duke regular season finale wins the ACCT, beating both UVA and Duke/UNC along the way.

[BandAlum83]

Prob requires some weird losses by a combo of KY/Tenn/MSU/Michigan combined with Duke/UNC/UVA only losing to each other from here on out, and the loser of UNC/Duke regular season finale wins the ACCT, beating both UVA and Duke/UNC along the way.

I think somehow, a UVA loss to either Syracuse or Louisville might help.

Remaining pre-conference tournament games:

Kentucky: @ ole miss, Florida
Tennessee: MS State, @Auburn
Michigan: @Mich State
Mich State: Nebraska, Michigan

UNC: @ BC, Duke
Duke: Wake, @ UNC
UVA: @Syracuse, Louisville

So everyone outside of the ACC loses at least one of their remaining conference game and flames out in conference tournaments.

UVA drops one regular season game - Probably away vs Syr would be a better loss to keep them on the #1 line, Duke wins out for 3 way tie at top of ACC.

UNC beats UVA in tournament and Duke beats UNC for ACCT title. Maybe that would do it?

[uh_no]

Is there a scenario where Duke, UVA and uNC all get #1 seeds in the NCAAT?

only 2 will get 1 seeds. A loss in the semis knocks out duke or UNC. I think UVA can skirt even with a loss there (people will complain about duke beating UVA twice, but a loss in the semis for us W/o zion means we haven't established that he will be back, and then we're graded on our performance without him, and a loss WITH zion means we're simply not as good. getting to the finals is critical for us).

If UNC and Duke both make the finals, whoever wins gets it IMO

So pretty much it's uva and either duke or UNC.

[thedukelamere]

only 2 will get 1 seeds. A loss in the semis knocks out duke or UNC. I think UVA can skirt even with a loss there (people will complain about duke beating UVA twice, but a loss in the semis for us W/o zion means we haven't established that he will be back, and then we're graded on our performance without him, and a loss WITH zion means we're simply not as good. getting to the finals is critical for us).

If UNC and Duke both make the finals, whoever wins gets it IMO

So pretty much it's uva and either duke or UNC.

Unless Cuse and the Cards can help us out this week :cool:

[BandAlum83]

Unless Cuse and the Cards can help us out this week :cool:

Syracuse is tough, but UVA has some zone busters in Guy and Jerome. Let's hope for a bad day from behind the arc for UVA!

C'mon Cuse fans! Bring you "A" game to the dome!

[JasonEvans]

Is there a scenario where Duke, UVA and uNC all get #1 seeds in the NCAAT?

It ain't easy but here is how I see it...

Provided Virginia doesn't lose 2 games between now and selection Sunday, they have a #1. Heck, they might even be able to lose twice and still get a #1.

Is there a scenario where Gonzaga does not get a #1 if they lose in the WCC tourney, especially if they drop a game to a less team (not St. Mary's)? Probably not.

The path to the ACC getting three depends to a large extent on other contenders. The committee is going to be reluctant to reward any one conference with three #1s so we need other contenders to look less than worthy. Here is how I see that playing out...

1) Neither Kentucky nor Tennessee win the SEC tourney. If LSU or Auburn or someone else wins the SEC tourney (not exactly a huge longshot) that would help a lot. It is worth noting that if the committee is deciding between Duke and Kentucky for the final #1 seed, we did beat the @&^#%*~ out of them earlier this year (much earlier, so it carries less weight). Kentucky also beat UNC, so that would seem to give the Cats the tiebreaker with Carolina if those two are vying for the last #1 seed.

2) Michigan and MSU each find their way to one more loss. They play this weekend, so one of them will lose. It would be good if the loser (or someone else) went on to win the BigTen tourney. It would be bad for the 3 ACC #1s scenario for Michigan to beat MSU and then win the Big Ten Tourney. Like Kentucky, Michigan also has a win against UNC so I feel like the most likely ACC scenario would involve Duke as the 4th #1 seed, not Carolina.

So, if both of those happen, here are the scenarios I see where the ACC gets 3 #1s...

* Whoever loses on Saturday must win the ACC tourney (and probably must beat Virginia in the ACC finals).

* I think the easiest path is if Duke loses at Carolina (preferably with an absent or somewhat limited Zion) but then roars back and wins the ACC tourney in dominating fashion. Duke gets a #1 because the non-Zion games are deeply discounted but Carolina gets credit for beating Duke twice. That said, there might be a scenario where Carolina loses to Duke on Saturday but then beats both Duke and Virginia to win the ACC and all three teams get a #1. But, I think the "Zion excuse" is the easiest path to making this happen.

-Jason "all this is highly unlikely, but not at all impossible" Evans

[scottdude8]

I think there's a lot of unknowns this year brought about by the NET. Case and point: I don't think in past years a 2-loss non power conference team like Gonzaga would ever be considered a one seed lock... look at the Zags team sheet and outside of beating us it isn't that impressive. But being No. 1 in the NET is a new piece of support we haven't had before. And because of all the mystery surrounding the NET we don't know how much a loss in their conference tourney would drop them.

But going to the UNC question: I really think that scenario is more about other teams falling than anything that happens in the ACC. For there to be any shot at three ACC No. 1 seeds, I would imagine that all of Tennessee, Kentucky, MSU and Michigan would need another loss, and probably a bad one. If that happens, a six or seven loss UNC team could definitely get a No. 1 seed alongside us and UVA. It isn't unprecedented (the old Big East got 3 No. 1 seeds at least once, IIRC), and a six or seven loss No. 1 seed is also quite common, as I've mentioned a lot. But I would imagine because of the optics the committee would need it to be a near-bulletproof case to take that on.

[Kedsy]

I think there's a lot of unknowns this year brought about by the NET. Case and point: I don't think in past years a 2-loss non power conference team like Gonzaga would ever be considered a one seed lock... look at the Zags team sheet and outside of beating us it isn't that impressive. But being No. 1 in the NET is a new piece of support we haven't had before. And because of all the mystery surrounding the NET we don't know how much a loss in their conference tourney would drop them.

It's hard to quantify what teams were considered locks, when and by whom, but Gonzaga has twice been a #1 seed, just in the past six seasons.

[HereBeforeCoachK]

Prob requires some weird losses by a combo of KY/Tenn/MSU/Michigan combined with Duke/UNC/UVA only losing to each other from here on out, and the loser of UNC/Duke regular season finale wins the ACCT, beating both UVA and Duke/UNC along the way.

...and all of these matchups being close games as well....

[scottdude8]

It's hard to quantify what teams were considered locks, when and by whom, but Gonzaga has twice been a #1 seed, just in the past six seasons.

True. But they were a 1 loss team in 2017 and in 2013 as a two loss team they were the last 1 seed, and probably only got it thanks to us getting upset in the ACC tourney. The first case they had a better resume, and the second they were FAR from a lock at any point. All that’s to say a third loss for Gonzaga would be a situation that’s hard to predict, especially with the new NET.

[Kedsy]

True. But they were a 1 loss team in 2017 and in 2013 as a two loss team they were the last 1 seed, and probably only got it thanks to us getting upset in the ACC tourney. The first case they had a better resume, and the second they were FAR from a lock at any point. All that’s to say a third loss for Gonzaga would be a situation that’s hard to predict, especially with the new NET.

So you're saying a team from a non-power conference has to be a 1-loss team or better to be a "lock" to get a #1? Maybe. 2014 Wichita State (0 losses), 2008 Memphis (1 loss), and 2004 St. Joseph's (1 loss) support that theory. But 2006 Memphis (3 losses) and 2002 Cincinnati (3 losses) don't necessarily support it. My guess is this year's Gonzaga team, with a win over Duke, no losses outside the AP top 5, #2 in KenPom and #5 in the RPI, would certainly be a strong contender for a #1. Not sure about "lock," but I'm not sure anybody's ever a lock.

[weezie]

... Not sure about "lock," but I'm not sure anybody's ever a lock.

Yet one R. Pitino sees a sure thing coming down the pike...https://thespun.com/more/top-stories/rick-pitino-thinks-1-team-will-definitely-make-the-final-four-its-not-duke

[CDu]

So you're saying a team from a non-power conference has to be a 1-loss team or better to be a "lock" to get a #1? Maybe. 2014 Wichita State (0 losses), 2008 Memphis (1 loss), and 2004 St. Joseph's (1 loss) support that theory. But 2006 Memphis (3 losses) and 2002 Cincinnati (3 losses) don't necessarily support it. My guess is this year's Gonzaga team, with a win over Duke, no losses outside the AP top 5, #2 in KenPom and #5 in the RPI, would certainly be a strong contender for a #1. Not sure about "lock," but I'm not sure anybody's ever a lock.

I would probably not classify 2002 Cincy in with those other teams. They had Memphis, Louisville, and Marquette in their conference. While not a Power-5, the CUSA was probably the sixth-best conference in D-1 at that time.

I know that you said “non-power-5”, but I would suggest that the gradations are not binary. A 3-loss CUSA team from 2002 is a different animal than a 3-loss WCC team this year. Just very different level of competition.

[scottdude8]

So you're saying a team from a non-power conference has to be a 1-loss team or better to be a "lock" to get a #1? Maybe. 2014 Wichita State (0 losses), 2008 Memphis (1 loss), and 2004 St. Joseph's (1 loss) support that theory. But 2006 Memphis (3 losses) and 2002 Cincinnati (3 losses) don't necessarily support it. My guess is this year's Gonzaga team, with a win over Duke, no losses outside the AP top 5, #2 in KenPom and #5 in the RPI, would certainly be a strong contender for a #1. Not sure about "lock," but I'm not sure anybody's ever a lock.

I think you have to take the competition into account this year. What if Gonzaga loses one and Michigan wins out? Do you put a 3 loss Gonzaga as a 1 over a 4 loss B1G champion? I’d think given the SOS that’d be a hard argument to make unless the analytics/NET are strongly in favor in Gonzaga. What if Gonzaga loses and both Tennessee and UK finish with 5 losses or less? Or MSU wins out (god forbid)? I think the same argument holds. Historically, to earn a No. 1 seed a “mid-major” (maybe that’s a better word than non-Power 5 or whatever) has to have a nearly infallible resume, i.e very few losses and all of them justifiable. A loss to a Mountain West team punches a hole in that infallible resume, for sure. If this was like last year where the only alternative was a 6-7 loss major conference team, then maybe Gonzaga gets the benefit of the doubt. But there are going to be 2 seeds with VERY strong resumes this year whichever way the cookie crumbles. So yes, I don’t think Gonzaga is a lock... or for those who don’t like that term, I don’t think that Gonzaga necessary can lose a game and still be comfortable that they have a 1 seed.

[Wander]

I know that you said “non-power-5”, but I would suggest that the gradations are not binary. A 3-loss CUSA team from 2002 is a different animal than a 3-loss WCC team this year. Just very different level of competition.

Agreed. Look at the "power" Pac-12 conference this year. It is far closer to Gonzaga's WCC than it is to the ACC or Big 10.

IMO Gonzaga this year will just depend on how heavily the committee weights margin of victory (and metrics that take margin of victory into account). If that is given a lot of value, Gonzaga is a lock for a 1 seed no matter what else happens, as they have been crushing their opponents by a crazy margin, even correcting for strength of schedule. If not, then they still may be a 1 seed, but I wouldn't call it a lock yet.

[Wahoo2000]

Syracuse is tough, but UVA has some zone busters in Guy and Jerome. Let's hope for a bad day from behind the arc for UVA!

C'mon Cuse fans! Bring you "A" game to the dome!

This man is a seer!

I will note though that you could've added Hunter, who is actually UVa's highest 3pt% shooter on any kind of volume at 49%

[duke2x]

Agreed. Look at the "power" Pac-12 conference this year. It is far closer to Gonzaga's WCC than it is to the ACC or Big 10. IMO Gonzaga this year will just depend on how heavily the committee weights margin of victory (and metrics that take margin of victory into account). If that is given a lot of value, Gonzaga is a lock for a 1 seed no matter what else happens, as they have been crushing their opponents by a crazy margin, even correcting for strength of schedule. If not, then they still may be a 1 seed, but I wouldn't call it a lock yet.

There's not a lot of head-head to distinguish the teams that are competing for #1 seeds. Gonzaga has the advantage that nobody wants that #1 seed, but they can't play quality competition after January 1. I exclude MI despite having 4 losses because their NCSOS is #310 on kenpom and NET is #9.

TN 2-1.
Duke 3-2. Injuries are a factor here--but how much?
KY 2-2. Injuries are less of a factor than Duke.
UNC 2-2. With 5 losses, 6 is too many to be a #1 seed this year. Go BC.
UVA 1-2. Almost a lock in my book but not necessarily for the East.
Gonzaga 1-2. Almost a lock for the West since the other 5 are guaranteed at least 4 more losses between them.

[DukeDevil]

Syracuse is tough, but UVA has some zone busters in Guy and Jerome. Let's hope for a bad day from behind the arc for UVA!

C'mon Cuse fans! Bring you "A" game to the dome!

This post is hilarious in retrospect

[BandAlum83]

I think there's a lot of unknowns this year brought about by the NET. Case and point: I don't think in past years a 2-loss non power conference team like Gonzaga would ever be considered a one seed lock... look at the Zags team sheet and outside of beating us it isn't that impressive. But being No. 1 in the NET is a new piece of support we haven't had before. And because of all the mystery surrounding the NET we don't know how much a loss in their conference tourney would drop them.

But going to the UNC question: I really think that scenario is more about other teams falling than anything that happens in the ACC. For there to be any shot at three ACC No. 1 seeds, I would imagine that all of Tennessee, Kentucky, MSU and Michigan would need another loss, and probably a bad one. If that happens, a six or seven loss UNC team could definitely get a No. 1 seed alongside us and UVA. It isn't unprecedented (the old Big East got 3 No. 1 seeds at least once, IIRC), and a six or seven loss No. 1 seed is also quite common, as I've mentioned a lot. But I would imagine because of the optics the committee would need it to be a near-bulletproof case to take that on.

Now that's funny!

Have they ever felt they needed to justify what they do?

[BandAlum83]

This man is a seer!

I will note though that you could've added Hunter, who is actually UVa's highest 3pt% shooter on any kind of volume at 49%

I don't know them as well as you. ;)

UVA is formidable. I thought that going on the road to Syracuse would be our best hope for a UVA stumble that might allow us to have a shot at tying for the regular season championship. I just don't see it at home on Senior day against Louisville since we broke them for y'all.

[scottdude8]

Now that's funny!

Have they ever felt they needed to justify what they do?

Touche, good sir. Touche.

[JasonEvans]

Have they ever felt they needed to justify what they do?

We can wink and joke about how the all-powerful NCAA selection committee does whatever it wants, but I think that is far from the case. Every year, the committee chair appears on national TV to discuss and explain the decisions made. Many times in the past, the committee has invited journalists to conduct their own pretend seeding a few weeks before the actual tournament, to educate journalists and their audience about how complicated the process can be. Years of loud criticism forced the committee to abandon the RPI and put the new NET metric in place.

None of these are the actions of a group that disdainfully considers itself beyond reproach. In fact, they appear to be the actions of a group that embraces transparency and understanding in the hope it will alleviate some of the inevitable criticism leveled at it.

-Jason "that said, I am sure there will be aspects of the seeding that I will find stupider than if they had thrown darts with seed numbers at names of random teams... and I reserve the right to call the committee all kinds of bad names after the draw is announced" Evans

[uh_no]

-Jason "that said, I am sure there will be aspects of the seeding that I will find stupider than if they had thrown darts with seed numbers at names of random teams... and I reserve the right to call the committee all kinds of bad names after the draw is announced" Evans

The thing is, a lot of us here think far more about these things than the committee does....these guys are ADs and conference execs and what not....they don't necessarily live and breathe CBB like we do. Sometimes decisions are made which are questionable when viewed by slightly more informed eyes, and they are rightly lambasted here. The lack of intense observation exhibited as an over-reliance on the flawed RPI. Thank god that's gone.

I don't think the committee goes and sets out to make crappy decisions. I also think that if ANY of us made a bracket, there would probably be at least one thing or two that would be widely questioned or inconsistent with other choices.

In the end I don't mind that their decisions aren't self consistent. In one case, there is no reason why they can't value one factor in one case, and another factor in another. This, of course, ticks of joe lunardi who then is "wrong".

[Indoor66]

The thing is, a lot of us here think far more about these things than the committee does...these guys are ADs and conference execs and what not...they don't necessarily live and breathe CBB like we do. Sometimes decisions are made which are questionable when viewed by slightly more informed eyes, and they are rightly lambasted here. The lack of intense observation exhibited as an over-reliance on the flawed RPI. Thank god that's gone.

I don't think the committee goes and sets out to make crappy decisions. I also think that if ANY of us made a bracket, there would probably be at least one thing or two that would be widely questioned or inconsistent with other choices.

In the end I don't mind that their decisions aren't self consistent. In one case, there is no reason why they can't value one factor in one case, and another factor in another. This, of course, ticks of joe lunardi who then is "wrong".

Sometimes "slightly more informed eyes" fail to see the forest for the trees, IMO.

(That is not aimed at you, uh_no, or any other poster. Just a general observation.)

[English]

We can wink and joke about how the all-powerful NCAA selection committee does whatever it wants, but I think that is far from the case. Every year, the committee chair appears on national TV to discuss and explain the decisions made. Many times in the past, the committee has invited journalists to conduct their own pretend seeding a few weeks before the actual tournament, to educate journalists and their audience about how complicated the process can be. Years of loud criticism forced the committee to abandon the RPI and put the new NET metric in place.

None of these are the actions of a group that disdainfully considers itself beyond reproach. In fact, they appear to be the actions of a group that embraces transparency and understanding in the hope it will alleviate some of the inevitable criticism leveled at it.

-Jason "that said, I am sure there will be aspects of the seeding that I will find stupider than if they had thrown darts with seed numbers at names of random teams... and I reserve the right to call the committee all kinds of bad names after the draw is announced" Evans

They may embrace self-improvement and public relations (and good on them for that), but transparency? If that's true, please explain to us how the NET works. Or how the Committee intends to use it.

Anyway, the human element of the Committee's role actually appeals to me more than a computer pumping out the bracket in a rankings/S-curve way devoid of drama and nuance.

[weezie]

After a worn out drive home from Durham, I'm left with one over-riding concern...

Pooooooooor Bill Self and waaaaahhhhhhh for the jayhawk fans. Sniff, sniff, boooo-hooo.

Some wag suggested it was time for Billy to skip out on ku and take over for Popovich in San Antonio because why not?! How do the radio idiots come up with this stuff?