Revisiting NET and Talking Seeds

Submitted by TrueBlue2003 on March 3rd, 2019 at 11:00 PM

Thought I'd get a diary up to talk seeds as Michigan has a six day layoff before a massive game Saturday.


As soon as the NCAA released the elements of the formula that went into the NET rankings, I thought that the inclusion of unadjusted efficiencies was absurd and would favor good teams in mid-major conferences.  Now that we're almost finished with conference play and those good teams have been feasting on their lesser conference brethren, that appears to be absolutely the case with NET.

Compare the mid-majors in the top 25 of NET to their Kenpom ranks:

Gonzaga NET: 1; Kenpom: 2

Houston NET: 6; Kenpom: 15

Buffalo NET: 14; Kenpom: 20

Wofford NET: 15; Kenpom: 22

Cincinnati NET: 23; Kenpom: 31

Every single one is ranked higher in NET and kenpom and by an average of more than 6 spots.  Again, the reason for this is that the NET treats a 30 point win over a 30 point win over the last place MAC team as the same as a 30 point win over the first place Big Ten team in the efficiency portion of the formula.  There is no opponent adjustment like kenpom and the predictive advanced stats make.

What could potentially impact Michigan here is that disparity for Houston.  In the committee's initial seeding on Feb 9, they had Houston as 3 seed (which is in the overall 9-12 range).  At the time they were 7th in NET and 19th in kenpom, suggesting that committee seemed to stick surprisingly close to the NET ranking rather than using it mostly a tool determine the quality of a teams opponents.

After Houston lost a home game to UCF yesterday they only fell to 6th in NET.

Wofford and Buffalo have climbed significantly since that initial seeding, going from 29th to 15th and 23rd to 14th respectively as they've feasted on weak conference foes.  Will be interesting to see if they sniff protected top 4 seeds as a result.


One thing that I think really stood out in the initial seedings in terms of where the committee deviated from NET rankings was Michigan's old bugaboo: non-conference SOS.

I think that's the reason Gonzaga was a 1 seed at the time despite having the same record against Michigan against a significantly worse overall schedule.  The committee seems to be rewarding teams for schedule difficult games with the scheduling they control.

Kansas is the other example as the were a 4 seed despite being 18th in the NET with a 17-6 record.  They were the only team included in the top 16 that weren't top 16 NET and my only guess is their SoS, particularly NC SoS

Just some food for thought and dicsussion.




March 4th, 2019 at 12:29 PM ^

I'm hoping for a two seed with Gonzaga as the 1 seed out west again in SoCal.  Staples was Crisler West last year.  Honda Center would be this year.

Houston that bracket would be fine with me as well. I would take Va Tech or Kansas too. Texas Tech is the team I don't want to see on that line.

Michigan Arrogance

March 4th, 2019 at 7:13 AM ^

I was also worried about this seeding stuff with the NET rankings - I find it inconceivable that M could be a 3 given the wins in the OOC schedule and the strength of the B10. 3 road losses (2 to tournament teams, the 3rd to a top 65 team) and the MSU loss at home.

Lots of teams with more losses than that and to lower ranked teams. Someone posted a stat that M will finish with . no more than 7 losses and therefore will have the fewest losses of any M team since 92-93. Sweeping a good OOC schedule and the strongest B10 in 6 years and they could get a 3? I'd be VERY interested in seeing a Quad 1-4 W-L analysis for the top 15 teams to see how seeding works out; M may end up with 20 Q1 wins if the BTT goes well and they beat MSU on Sat.


March 4th, 2019 at 12:37 PM ^

Well, M is squarely in two seed range right now.  But a loss to MSU puts them on the border and then it'd be up to them to get a couple wins in the BTT to stay comfortably in two seed range.  Anything but a deep run and they'd be at risk of Houston or Texas Tech or even Purdue or LSU passing them.

A win at MSU probably locks up a 2 seed and gives them a good chance at a 1 seed, especially if UK and/or Kentucky lose again this week.

The problem is Michigan's OOC was terrible outside of UNC and Nova.


March 4th, 2019 at 10:08 AM ^

what does "a 30 pt win over a 30pt win" mean?   

What does this mean "ranked higher in NET and kenpom and by an average of more than 6 spots."


March 4th, 2019 at 12:44 PM ^

Typed this up while bored on a plane and submitted real quick while landing before wi-fi went out.  Went back and read all the typos, and yikes, brutal.  Like a monkey typed this thing.  The fact these can't be edited is absurd, but I am embarrassed by how bad that was. I need to write gooder. I assume you're just being snarky, so yeah, thanks for pointing them out.

Focus on the chart!  The numbers!


March 4th, 2019 at 12:18 PM ^

This needed a thorough proofreading before posting... but I agree with the jist of the implications here that NET is a bit messed up on some of the inputs.  Seems like they are trying to validate eventual high seeds for top mid-majors teams over what we may think are better teams that finish 3rd-6th in a major conference.  Gonzaga may make sense as they've proven to hang with the top team, the rest probably not 


March 4th, 2019 at 12:56 PM ^

It'll be interesting to see if they do in fact end up giving those mid-majors high seeds.  I personally don't know if they did this on purpose or just didn't understand the implications of using unadjusted efficiency margins.  If the didn't, I could see them tweaking the formula. 

It is interesting to see how the rankings went from being a pretty good system midseason when schedules were more balanced across the sport to systematically favoring teams that have played weak schedules.

I still think it's better than RPI, but it has flaws that are easily preventable.  And the flaws aren't in the direction of "you play to win the game" which would be understandable.  The flaws are in the direction of "you play to beat bad teams by as much as possible".

That does two things that I don't think the committee intended: 1) encourages running up the score and 2) encourages easy scheduling to make it easier to run up scores.  The latter seems to be offset somewhat by the committee taking into account NC SOS and making an adjustment but you still have the former problem.


March 4th, 2019 at 6:30 PM ^

I've got to think a win in Lansing this weekend and a decent showing in the big ten tournament should land Michigan a 2 seed. 

Even with a loss against sparty I would think that winning the big ten tourney would solidify them as a 2.


March 5th, 2019 at 12:35 AM ^

2 or 3 seed is not much difference - just hoping we beat Snarty. 

We need to be in a region where fans are few for most of the teams and have a lot of UM grads around like Staples Center to make it feel like a home game.

The frigging key is getting Matthews back at 90%+ strength.