November 30th, 2018 at 9:27 AM ^

Sounds good. Computer models that factor in strength of schedule, W/L margins, etc. cannot possibly be accurate with a tiny sample size.  They're not supposed to. As long as they get it right when it matters (selection time) who cares what they say in November.  

Same applies for CFB polls 5 weeks into the season.  Besides the media, who cares? Play the games and it will all sort itself out.  


November 30th, 2018 at 9:26 AM ^

It's like this new system doesn't even look at basketball, it just says give me all the undefeated teams, a dart board, a dart, some beer and look out. Texas Tech? 


November 30th, 2018 at 9:45 AM ^

KenPom has them at #73.  They're your likely #1 seed in the OVC tournament.

They also have three road wins, including one at NET #42 Lipscomb.  The NET really likes road wins.

I mean, they're not the 9th best team in the country, but for a computer rating that's not taking pre-season expectations into account, it's not insane.  Let's try not to judge it until mid-February at the earliest.

PS: They seem to update the NET sometime in the late morning / early afternoon each day.  It hasn't actually been updated with Thursday's games yet.  There wasn't a lot of basketball activity last night, so it probably doesn't matter much, but I thought I'd point that out for the future.  I have a feeling many of us will eventually be checking this thing daily. :)


November 30th, 2018 at 9:42 AM ^

So even though 10 points was set as the margin of victory cap for "blowouts," the second component of NET factors in efficiency margin which is essentially margin of victory (more blowouts equals higher efficiency margin). So unless I'm oversimplifying or misunderstanding, their "margin of victory" cap is basically pointless and is peak NCAA.

This is relevant as Michigan has yet to play a game where they did not win by at least 17 points.


November 30th, 2018 at 9:50 AM ^

Yeah, I didn't quite get that either.  Unless they're going to adjust the efficiency to represent a 10-point win, which seems all sorts of wrong to me...

They also have the asinine 0.6/1.0/1.4 home/neutral/road thing that the RPI used.  (For computing a team's record, you use 0.6 for a home win or road loss, 1.4 for a home loss or road win, and 1.0 otherwise).  The intent was to reward road wins, which sounds benign enough.  The problem is that if you have two teams under tournament consideration -- say, both with 22-10 records -- if one team had a road win and a home loss, and the other team had a road loss and a home win, and all other items are equal, the team with the road win has a lower adjusted winning percentage.  I mean, it's not difficult to show mathematically (assume all other games were neutral site; 21.6-9.6 is a .692 winning percentage; 22.4-10.4 is .683), and it was a known problem with the RPI.. and they carried it over anyway.

So, I don't think the NCAA had its brightest minds working on the NET.  Or, it did -- which is terrifying. :)


November 30th, 2018 at 2:26 PM ^

Reading comprehension.  The team with the road win would have a lower win percentage per this formula which was exactly the way the RPI did it.

In those two games, they get awarded a 1.4-1.4 record for the road win and home loss.

The other team gets a 0.6-0.6 record in those two games.

So the first teams overall win % gets dragged down because that 0.500 record in those two games has a higher weight (1.4) on the overall record than the 0.500 record of the other team (0.6).

Boner Stabone

November 30th, 2018 at 10:00 AM ^

I like the NET.  I think these rankings will be very accurate after a large sample size.  It is still way to early to make conclusions off the small sample size, even though I like our #3 spot.


November 30th, 2018 at 2:50 PM ^

It is better than RPI at this point, probably.  The reason people are freaking out is that no one talked about the RPI in November so no one knew how bad it was at this point in the season.  Well, the replicator RPI sites still spit out RPI so let's check in there:

RPI top 20 per through Nov 29 games:

1. Kansas

2. Duke

3. Oklahoma (!!)

4. Tennessee

5. Wisconsin

6. Texas

7. Gonzaga

8. Iowa

9. Radford (!!!)

10. Michigan

11. Loyola Marymount (!!)

12. Belmont (!)

13. FSU

14. Auburn

15. Nevada

16. UCF

17. Texas Tech

18. UVA

19. Citadel (!!!) 3-2 Citadel!!!!!

20. Kansas St.

In theory, it seems like they threw a bunch of things at the wall for NET and to me it makes no sense that they would us an unadjusted efficiency margin because it does the last two things the committee has wanted in the past: 1) encourages easy home scheduling because it 2) simply rewards blowouts.

It is a bad ranking system no doubt.

But in practice so far, it seems (slightly) better than RPI.


November 30th, 2018 at 10:10 AM ^

At this stage of the season, I prefer the much more critical, analytical, unbiased assertions of a little beta program I've developed called "the Salinger eye test." And oh boy is Michigan good on that index. 



November 30th, 2018 at 10:18 AM ^

Color me dubious.  Until the selection committee actually uses something different than RPI to RPI teams into the tunamelt, I'm just going to assume that RPI will still be consulted.  Besides, I have no idea how different the NET actually is from RPI.  Perhaps the new 'RPI'ing into the tourney' will be 'Net'ing into the tourney'?