NET-RPI-Massey-AP rankings for TOP 40 teams | The Boneyard

NET-RPI-Massey-AP rankings for TOP 40 teams

DefenseBB

Snark is always appreciated!
Joined
Nov 10, 2016
Messages
7,863
Reaction Score
28,480
Ok, given the constant focus on the AP poll in particular and the actual tools used by the committee (NET, RPI Conference Champion, injuries), I thought I would try to show the analysis of the 4 ranking services at our disposal plus the Top 10 from HerHoopStats as displayed on their free home page. This conglomeration has SC #1, UConn #2 and then LSU/Stanford/Indiana fighting for #3.

We know NET is to be the primary source with RPI, Conference title and incorporating any injuries into the equation. This is what Charlie Creme tries to replicate and inform the interested fans. I think for the most part, he does an admirable job trying to replicate what he can. Here is the data and you will see that the AP voters do not use advanced metrics in their thinking as there are 5 teams whose AP ranking is not proportional to their NET or RPI-Texas, Oregon, Maryland, UCLA and Gonzaga. Texas and Oregon are under-rated by the voters. Maryland, UCLA and Gonzaga are OVER-RATED by the voters. The below chart shows the teams, record, Conference, NET, RPI, AP, Massey and top 10 Herhoopstat rankings. I have listed the top 40 from NET.

From a Conference perspective, the PAC-12 has 8 team in the top 40 of NET (7 in the RPI top 40), ACC, SEC and Big Ten have 7 each in NET, with Big 12 with 4 and Big East with 3. From an RPI perspective, the ACC/Big Ten have 7, SEC has 6 while Big East has 5 and Big 12 has 4 (swapping Baylor for Oklahoma).

Net image.PNG
 
Joined
Feb 8, 2016
Messages
5,060
Reaction Score
17,759
Excellent work. Thank you!

Nothing really matters but NET and I’m not sure how much it matters. The committee can deviate from the NET under guidelines that are so broad they really just leave it up to the say so of the committee’s discretion.

Basically, I think the committee does a pretty good job.
 

southie

Longhorn Lover
Joined
Apr 1, 2012
Messages
2,625
Reaction Score
5,893
I'm not convinced Creme uses NET as much as he should in his bracketology. He's had Texas as a #8 seed a couple of weeks ago before finally moving them up to #7; today he moved Texas up to a #5 seed (but still not in line with their NET ranking). And, he's had Ohio State really high up in his bracket despite their NET ranking being worse than that of a Top 1-2 seed. Now that OSU has lost 3 games in a row, I'll be curious to see where he re-seeds the Buckeyes.

 
Joined
Apr 24, 2022
Messages
5,190
Reaction Score
30,380
The committee can deviate from the NET under guidelines that are so broad they really just leave it up to the say so of the committee’s discretion.
I think that’s right. But I suspect they’re reluctant to depart too far from the NET for fear of seeming unfair. Of course, conference and geographic considerations force them to depart from it. But I imagine there’s a considerable incentive to minimize this.
 
Joined
Mar 7, 2020
Messages
193
Reaction Score
634
An alternative view would suggest that Texas, Oregon and Alabama?!?! are OVER RATED by the NET calculation and Gonzaga, Maryland and UCLA are UNDER-RATED by the NET rankings. UCLA had two close road losses last weekend and its NET improved from 27 to 25. I don’t think Oregon warrants an #14 ranking. #43 RPI seems a bit harsh while #20 Massey still seems generous. Oregon certainly has UNDERPERFORMED relative to its preseason expectations given its two returning all conference guards and #2 recruiting class.
 
Joined
Jan 30, 2017
Messages
3,041
Reaction Score
14,436
Been trying to figure out how the NET rankings are compiled. I have seen one image for NET (right side with 5 components), searching today I came across the NCAA updated and revised the NET ratings down to 2 components, Adjusted NET Efficiency and Team Value Index (left side effective 2020-2021 season). The link to the NCAA FAQ page helps to answer what these components are.

Providing this info to anyone that may have had questions regarding the NET but were afraid to ask.


20-2021 NET vs Original NET.jpg
 
Joined
Apr 24, 2022
Messages
5,190
Reaction Score
30,380
The puzzle of NET is how it manages SoS. It includes factors like location, which makes sense, but also winning percentage and score differential, both of which will tend to favor teams with weaker schedules. There’s also an efficiency factor on offense and defense which primarily reflects the team itself and only indirectly speaks to the quality of opponents. Finally, it includes the opponents’ NET ranking, which is the primary way it measures SoS, and the only factor that inhibits teams with cream puff schedules from getting into the higher rankings.

In the end, I think NET is a pretty good measure, though not perfect. I think you can see this by the way NET consistently ranks SC above UConn as 1 & 2. This looks pretty reasonable. By contrast, RPI gets this wrong, consistently ranking UConn above SC, I think because it puts too much weight on SoS. This is too jarring a departure from even a cursory eye test assessment of both teams.

The weakness of NET is clear from the way it has consistently ranked LSU too high, in my opinion. An extreme cream puff schedule made for lots of lopsided wins and created the appearance of efficiency, and this outweighed any SoS factors. This also departs too far from eye test results. This shows how the NET can be gamed, and we’re left with the current spectacle of pundits either grudgingly giving LSU a 2 seed or reluctantly giving them anything from a 4 or 5 to a 7 or 8 seed. Either result will produce howls from one camp or another.
 
Joined
Feb 8, 2016
Messages
5,060
Reaction Score
17,759
The puzzle of NET is how it manages SoS. It includes factors like location, which makes sense, but also winning percentage and score differential, both of which will tend to favor teams with weaker schedules. There’s also an efficiency factor on offense and defense which primarily reflects the team itself and only indirectly speaks to the quality of opponents. Finally, it includes the opponents’ NET ranking, which is the primary way it measures SoS, and the only factor that inhibits teams with cream puff schedules from getting into the higher rankings.

In the end, I think NET is a pretty good measure, though not perfect. I think you can see this by the way NET consistently ranks SC above UConn as 1 & 2. This looks pretty reasonable. By contrast, RPI gets this wrong, consistently ranking UConn above SC, I think because it puts too much weight on SoS. This is too jarring a departure from even a cursory eye test assessment of both teams.

The weakness of NET is clear from the way it has consistently ranked LSU too high, in my opinion. An extreme cream puff schedule made for lots of lopsided wins and created the appearance of efficiency, and this outweighed any SoS factors. This also departs too far from eye test results. This shows how the NET can be gamed, and we’re left with the current spectacle of pundits either grudgingly giving LSU a 2 seed or reluctantly giving them anything from a 4 or 5 to a 7 or 8 seed. Either result will produce howls from one camp or another.
Yes, I agree. I think both NET and RPI are too complex. I keep it much simpler in my Boneyard Poll ballots. I look only at Quad one wins minus total losses. Even then, I wish it didn't have the weighted home/netural/away game distinction. I would rather the quads be a straight (1)-top 64, (likely tournament teams)
(2) 65-128, (possible tournament teams)
(3) 129-192,(unlikely tournament teams) and
(4) the rest ( almost "no way") tournament teams).

But I'm too lazy to do the analysis each week. So I just go with the net quad One.

My ranked team ignored by most is Columbia U which has 4 quad one wins and only two losses for a net winning record of 2 while a team like Oregon has a net winning record of minus 4. And Oregon hasn't played many more quad one teams.
 
Joined
Apr 24, 2022
Messages
5,190
Reaction Score
30,380
Yes, I agree. I think both NET and RPI are too complex. I keep it much simpler in my Boneyard Poll ballots. I look only at Quad one wins minus total losses. Even then, I wish it didn't have the weighted home/netural/away game distinction. I would rather the quads be a straight (1)-top 64, (likely tournament teams)
(2) 65-128, (possible tournament teams)
(3) 129-192,(unlikely tournament teams) and
(4) the rest ( almost "no way") tournament teams).

But I'm too lazy to do the analysis each week. So I just go with the net quad One.

My ranked team ignored by most is Columbia U which has 4 quad one wins and only two losses for a net winning record of 2 while a team like Oregon has a net winning record of minus 4. And Oregon hasn't played many more quad one teams.
I like your quad analysis. I have an even shorter version of that for the preseason -- just look at how many Quad 4 teams have been scheduled for a team. In late December, look at how many Quad 3 + 4 teams have been scheduled. By mid-January, look at how many Quad 1+2+3 teams were on a team's schedule.

Of course, by March quad differences mostly don't matter anymore and the powerhouse teams should be more or less apparent. Surprises are always possible, and are even part of the fun for fans, if not for coaches. Last year, it was hard to tell whether UConn was a powerhouse or not and a lot of pre-tournament expectations were dashed. This year, it's LSU who will dash expectations, one way or another.

UConn, too, may also become a mystery depending on whether Caroline and Azzi return and how late in the season. Right now, the NET appears to favor giving UConn a 1 seed, and even Charlie Creme and Megan Gauer have gone along with this, and their doubts are reflected in which 1 seed that give us. The fact is many people, maybe most people, expect our weak bench to keep us from going very far.
 
Joined
Feb 8, 2016
Messages
5,060
Reaction Score
17,759
I like your quad analysis. I have an even shorter version of that for the preseason -- just look at how many Quad 4 teams have been scheduled for a team. In late December, look at how many Quad 3 + 4 teams have been scheduled. By mid-January, look at how many Quad 1+2+3 teams were on a team's schedule.

Of course, by March quad differences mostly don't matter anymore and the powerhouse teams should be more or less apparent. Surprises are always possible, and are even part of the fun for fans, if not for coaches. Last year, it was hard to tell whether UConn was a powerhouse or not and a lot of pre-tournament expectations were dashed. This year, it's LSU who will dash expectations, one way or another.

UConn, too, may also become a mystery depending on whether Caroline and Azzi return and how late in the season. Right now, the NET appears to favor giving UConn a 1 seed, and even Charlie Creme and Megan Gauer have gone along with this, and their doubts are reflected in which 1 seed that give us. The fact is many people, maybe most people, expect our weak bench to keep us from going very far.
Screw seeds difference between 1,2,2,4. I just want us both in Greenville but in different brackets.
 

UcMiami

How it is
Joined
Aug 26, 2011
Messages
14,056
Reaction Score
46,332
I think when looking at a tournament ranking, quads are just too broad to be valuable. There are 32 auto qualifiers but really only maybe 6-8 ish that deserve to be in the top 32 seed positions (8 seeds or better.) So really the seeding issues are about 24-26 at large teams plus the also ran at large bids of about 10-12 teams that fill in the 9-11 and the extra play-in teams. And for a top 4 seeding (16 teams) I really am not interested in how many wins they have over at large teams in the 30s or higher in whatever ranking you use. It just isn't that impressive and a loss to anyone outside the top 8 seeds is a real negative. All 'Quad 1' wins and losses are not equal and so using them muddies the water. A win or loss anywhere against say ND does not equate to a win or loss against Oregon State and yet they would both be in Net Quad1 I believe.

And while team efficiency is a good way of evaluation, efficiency against the Lipscombs of the world grossly inflates a resume - play their equivalent 10 times OOC and you have an efficiency balance that even a tough conference schedule will not overcome.

Same of course goes to win/loss records, SOS calculations, and RPI - how many P5 teams came into conference play with a .750 win percentage and now are struggling below .500 in conference play? That .750 win percentage helps their conference opponents in both SOS and RPI calculations but often meaningless. Texas Tech came into conference at 12-1 and are at 3-5 in conference but still sport a .714 win percentage for the year!
 

DefenseBB

Snark is always appreciated!
Joined
Nov 10, 2016
Messages
7,863
Reaction Score
28,480
I have been doing another statistical ranking of FG%, Defense, offense, assists and Turnovers to see how those various categories correlate to the eye test and ratings of the elite teams. Still trying to refine it for ease of viewing. At the moment, I am still scratching my head about UCLA as they are not anywhere near the top of the categories in ANYTHING and only supports my hypothesis of them being drastically over-rated. Same goes for Oregon.
 
Joined
Aug 13, 2013
Messages
8,406
Reaction Score
7,935
The RPI and NET seem to match up well with the ACC teams...6 in top 25 RPI and 6 in top 25 NET...
 
Joined
Jan 31, 2023
Messages
127
Reaction Score
959
The puzzle of NET is how it manages SoS. It includes factors like location, which makes sense, but also winning percentage and score differential, both of which will tend to favor teams with weaker schedules. There’s also an efficiency factor on offense and defense which primarily reflects the team itself and only indirectly speaks to the quality of opponents. Finally, it includes the opponents’ NET ranking, which is the primary way it measures SoS, and the only factor that inhibits teams with cream puff schedules from getting into the higher rankings.

In the end, I think NET is a pretty good measure, though not perfect. I think you can see this by the way NET consistently ranks SC above UConn as 1 & 2. This looks pretty reasonable. By contrast, RPI gets this wrong, consistently ranking UConn above SC, I think because it puts too much weight on SoS. This is too jarring a departure from even a cursory eye test assessment of both teams.

The weakness of NET is clear from the way it has consistently ranked LSU too high, in my opinion. An extreme cream puff schedule made for lots of lopsided wins and created the appearance of efficiency, and this outweighed any SoS factors. This also departs too far from eye test results. This shows how the NET can be gamed, and we’re left with the current spectacle of pundits either grudgingly giving LSU a 2 seed or reluctantly giving them anything from a 4 or 5 to a 7 or 8 seed. Either result will produce howls from one camp or another.
Everyone will know more about how “real” LSU is after they play SC. Do they win, stay close, or get blown out? Place your bets!
 

DefenseBB

Snark is always appreciated!
Joined
Nov 10, 2016
Messages
7,863
Reaction Score
28,480
First off, thank you @mbr33ct for linking the NCAAT Committee NET and outline. I have posted the specific set of criteria the committee is to use. None are RPI.

Now, while many of us are incredulous at how poor the LSU OOC was, the fact is they won most of those games handily as they should have, hence very little penalty within the confines of the NET calculation. UConn played at PC last night but only won by 10 and PC is NET 114. By an easy comparison, LSU beat Montana State NET 116 by 39 pts (91-52) and while at home, LSU did what it was supposed to do-win big or even bigger than expected.

The current NET has us #2 and LSU #3, Stanford is #4, Indiana #5. The question of a #1 seed will ultimately come down to how well both UConn and LSU play against SC. My own opinion is the 4 #1 seeds will be SC*, Stanford*, Indiana*, UConn*
LSU (with 2 losses), Iowa, Texas*, ND* will secure the #2 seeds.
The * denotes Conference Champ/Automatic bid. Here is the snippet of the NCAA criteria:

Criteria used by the Division I Women’s Basketball Committee to evaluate a team includes (alphabetically):

  • Availability of talent (injured or unavailable players)
  • Bad losses
  • Common opponents
  • Competitive in losses
  • Conference record
  • Early competition versus late competition
  • Head-to-head outcomes
  • NET ranking
  • Non-conference record
  • Overall record
  • Regional Advisory Committee region rankings
  • Significant wins
  • Strength of conference
  • Strength of schedule
 
Joined
Feb 3, 2020
Messages
339
Reaction Score
842
I have been doing another statistical ranking of FG%, Defense, offense, assists and Turnovers to see how those various categories correlate to the eye test and ratings of the elite teams. Still trying to refine it for ease of viewing. At the moment, I am still scratching my head about UCLA as they are not anywhere near the top of the categories in ANYTHING and only supports my hypothesis of them being drastically over-rated. Same goes for Oregon.
Part of it is P5 bias and Oregon is likely getting some credit for how good they were during the Sabrina Ionescu era. UCLA is currently ranked 12th in SOS and projected to have the 9th best SOS by Massey. PAC12 has 3 teams in the projected top 10 SOS and 5 in the projected top 20 SOS. There are ways to crunch the numbers to make the PAC 12 look like a weak conference this year, but both Net and Massey have the PAC12 as the #1 conference.

Not sure why Massey likes Oregon so much, but it has Oregon rated #21 and Charlie has Oregon as a projected 7 seed. Massey has UCLA rated #17 and Charlie has them as a projected 4 seed.

That the Net has Oregon as #14 and LSU #3 is why there are 13 other factors the committee looks at. Net also has UCLA rated as #25, while Massey/Creme has them ranked higher.
 
Last edited:
Joined
Apr 24, 2022
Messages
5,190
Reaction Score
30,380
First off, thank you @mbr33ct for linking the NCAAT Committee NET and outline. I have posted the specific set of criteria the committee is to use. None are RPI.

Now, while many of us are incredulous at how poor the LSU OOC was, the fact is they won most of those games handily as they should have, hence very little penalty within the confines of the NET calculation. UConn played at PC last night but only won by 10 and PC is NET 114. By an easy comparison, LSU beat Montana State NET 116 by 39 pts (91-52) and while at home, LSU did what it was supposed to do-win big or even bigger than expected.

The current NET has us #2 and LSU #3, Stanford is #4, Indiana #5. The question of a #1 seed will ultimately come down to how well both UConn and LSU play against SC. My own opinion is the 4 #1 seeds will be SC*, Stanford*, Indiana*, UConn*
LSU (with 2 losses), Iowa, Texas*, ND* will secure the #2 seeds.
The * denotes Conference Champ/Automatic bid. Here is the snippet of the NCAA criteria:

Criteria used by the Division I Women’s Basketball Committee to evaluate a team includes (alphabetically):

  • Availability of talent (injured or unavailable players)
  • Bad losses
  • Common opponents
  • Competitive in losses
  • Conference record
  • Early competition versus late competition
  • Head-to-head outcomes
  • NET ranking
  • Non-conference record
  • Overall record
  • Regional Advisory Committee region rankings
  • Significant wins
  • Strength of conference
  • Strength of schedule
Great analysis! Only thing I’d add is that the list of criteria are mostly just elements already included in NET. The primary considerations outside the NET will probably be:

available talent
Head to head
Regional Advisory
Good vs bad losses

Whatever else isn’t already included in NET is more or less eye test stuff, and I think they’ll try to avoid the appearance of relying on that. The more they can document the basis of their judgments, the fewer complaints of bias they’ll receive.
 
Joined
Feb 8, 2016
Messages
5,060
Reaction Score
17,759
Great analysis! Only thing I’d add is that the list of criteria are mostly just elements already included in NET. The primary considerations outside the NET will probably be:

available talent
Head to head
Regional Advisory
Good vs bad losses

Whatever else isn’t already included in NET is more or less eye test stuff, and I think they’ll try to avoid the appearance of relying on that. The more they can document the basis of their judgments, the fewer complaints of bias they’ll receive.
Does "Regional Advisory Committee" means how many tickets can they sell in what region?
 

Online statistics

Members online
629
Guests online
3,935
Total visitors
4,564

Forum statistics

Threads
155,784
Messages
4,031,481
Members
9,865
Latest member
Sad Tiger


Top Bottom