NET Update post Purdue

My only possible explanation for the movement is that the actual NET scores of the teams around us factored in more than our wins did.

Oregon was the team that passed us, so their efficient 12 point win over Utah (NET 101) was enough to pass our 5 point slug fest over NET ranking #47 team, Penn State.

As a former NET defender, these counter intuitive late season shifts are really inexplicable.

Yeah your probably right. Gotta factor in utah and okla state losing past couple of days, washington barely winning, etc
 

My only possible explanation for the movement is that the actual NET scores of the teams around us factored in more than our wins did.

Oregon was the team that passed us, so their efficient 12 point win over Utah (NET 101) was enough to pass our 5 point slug fest over NET ranking #47 team, Penn State.

As a former NET defender, these counter intuitive late season shifts are really inexplicable.

Penn St passed Lipscomb and TCU. TCU lost by 9, so that could make sense they drop more than Penn St. But Lipscomb didn't chance.
Now if these were lowly rated teams, I could see moving closer to the mean after a narrow loss, but moving up when in the top 15-20% of all teams or so is weird.

My mistake about Nebraska.


Creighton lost and moved up two spots.
Passed Indiana
And Memphis who didn't play yesterday (and has since won).
 

Comparison of team sheets of Iowa and Gophers shows gophers slightly better in quad 1,2 record, significantly better in wins over teams with a net better than 50 (8 to 4), significantly better in wins over teams with a net better than 20 (3 to 1), and significantly better in both conf. and nonconf. SOS. The only advantage for Iowa is away record (4-6 versus our 2-9). The only way Iowa can be seeded ahead of us is through their better NET score - which according to some will not be determinative. We will see.
 
Last edited:


That’s improvement, I expected us to go down again.
 



lol we didn't move up in the NET.. unbelievable

Are you sure it's been updated?

I agree though, it is trash. Too many of these teams like Penn State ranking up high when they are in fact just not good teams. I still think RPI does a WAY better job than this NET does. I would be interested to see the RPIs of all the tourney teams when the bracket comes out compared to their NETs. If the committee truly uses a team's entire resume, you should be able to see which formula does a better job of predicting who should get in and who shouldn't.
 

The last couple days have really soured me on the NET. At least last week one could make the argument that adjusted efficiency metrics explained our ranking (even if one believed those metrics to be overemphasized in the formula). Our Kenpom ranking had tracked pretty close to our NET, but now we’re up to 41 in kenpom while dropping two spots in NET despite defeating two higher ranked teams. Don’t get it...
 

The last couple days have really soured me on the NET. At least last week one could make the argument that adjusted efficiency metrics explained our ranking (even if one believed those metrics to be overemphasized in the formula). Our Kenpom ranking had tracked pretty close to our NET, but now we’re up to 41 in kenpom while dropping two spots in NET despite defeating two higher ranked teams. Don’t get it...

Sagarin not updated yet, but we were 42 before Purdue and will probably crack the top 40 now. NET is in need of a revamp or retirement.
 



Penn St passed Lipscomb and TCU. TCU lost by 9, so that could make sense they drop more than Penn St. But Lipscomb didn't chance.
Now if these were lowly rated teams, I could see moving closer to the mean after a narrow loss, but moving up when in the top 15-20% of all teams or so is weird.

My mistake about Nebraska.


Creighton lost and moved up two spots.
Passed Indiana
And Memphis who didn't play yesterday (and has since won).

It is not a matter of more data moving it closer to the mean, they weighted the factors wrong.

They made some assumptions, tested the factors with some data and then thought it was a good predictor. They obviously didn’t test the model throughly enough. You don’t make such a major change because you think it works, you need to prove it works.

How they could mess up the factors so bad that simply losing a head to head game can improve your rating over the team that beat you is a fatal flaw. Any data scientist that was involved should be fired. If you messed up such a critical model so badly for a Fortune 500 company, you would be looking for another occupation!


Sent from my iPad using Tapatalk
 

Are you sure it's been updated?

I agree though, it is trash. Too many of these teams like Penn State ranking up high when they are in fact just not good teams. I still think RPI does a WAY better job than this NET does. I would be interested to see the RPIs of all the tourney teams when the bracket comes out compared to their NETs. If the committee truly uses a team's entire resume, you should be able to see which formula does a better job of predicting who should get in and who shouldn't.

It says on the NET website that the Gophers have a 21-12 record, so it has been updated.
 

It is not a matter of more data moving it closer to the mean, they weighted the factors wrong.

They made some assumptions, tested the factors with some data and then thought it was a good predictor. They obviously didn’t test the model throughly enough. You don’t make such a major change because you think it works, you need to prove it works.

How they could mess up the factors so bad that simply losing a head to head game can improve your rating over the team that beat you is a fatal flaw. Any data scientist that was involved should be fired. If you messed up such a critical model so badly for a Fortune 500 company, you would be looking for another occupation!


Sent from my iPad using Tapatalk
Honestly it doesn’t seem like the NET even factors the outcomes of games. It seems to just compile all of a team’s possessions and subsequently churns out macro efficiency numbers. If it does look at game outcomes, it only looks at margin of victory.. I’m pretty convinced the NET does not calculate a win or loss in a given game
 

It's been updated.

Nebraska lost and moved up from 49 to 48.


Memphis actually won, and they moved from 54 to 46. (They won by 24 so that's highly valued in the NET we know).

Syracuse didn't play, moved up a spot.

Oklahoma and Texas... neither one played and they swapped spots. 37/38.
B12 plays double-round robin by the way.
At #39, Baylor didn't move however.
 



St Mary's moved up two spots by not playing.
They passed two teams who lost, so that makes a bit of sense. (NC State and VCU)
 

Are you sure it's been updated?

I agree though, it is trash. Too many of these teams like Penn State ranking up high when they are in fact just not good teams. I still think RPI does a WAY better job than this NET does. I would be interested to see the RPIs of all the tourney teams when the bracket comes out compared to their NETs. If the committee truly uses a team's entire resume, you should be able to see which formula does a better job of predicting who should get in and who shouldn't.

This. RPI May end up closer to the final seeding than the beloved, but trash NET.


Sent from my iPhone using Tapatalk
 

I still don't get this. I'm looking at Palm's bracket: https://www.cbssports.com/college-basketball/bracketology/

He has us (21-12) as a 8 seed and our NET ranking is 57 and our Quad 1 is 5-9

He has NC State as out at 22-11, with a NET ranking of 33 and Quad 1 of 3-9. Their RPI is 96 while ours is 37... Our SOS is 150 and their SOS is 352. That might be why...

Where is SelectionSunday when I need someone to explain this all...
 

I still don't get this. I'm looking at Palm's bracket: https://www.cbssports.com/college-basketball/bracketology/

He has us (21-12) as a 8 seed and our NET ranking is 57 and our Quad 1 is 5-9

He has NC State as out at 22-11, with a NET ranking of 33 and Quad 1 of 3-9. Their RPI is 96 while ours is 37... Our SOS is 150 and their SOS is 352. That might be why...

Where is SelectionSunday when I need someone to explain this all...

I know that Quad 1 wins have become an important metric. However, it still matters who those Quad 1 wins are.
 

I still don't get this. I'm looking at Palm's bracket: https://www.cbssports.com/college-basketball/bracketology/

He has us (21-12) as a 8 seed and our NET ranking is 57 and our Quad 1 is 5-9

He has NC State as out at 22-11, with a NET ranking of 33 and Quad 1 of 3-9. Their RPI is 96 while ours is 37... Our SOS is 150 and their SOS is 352. That might be why...

Where is SelectionSunday when I need someone to explain this all...

Moral of the story, NET is trash.
Bracket makers are following more in the lines of RPI.
 

I saw someone posted on here there were only a few times it was a big difference between the rpi and net (with respect to people making the tourney) What if instead of these play in games we currently do, we just make those teams play each other..Isnt the point of a play in the game because they are unsure which of the final teams should get in... At least this way they could be transpartent about it
 

I saw someone posted on here there were only a few times it was a big difference between the rpi and net (with respect to people making the tourney) What if instead of these play in games we currently do, we just make those teams play each other..Isnt the point of a play in the game because they are unsure which of the final teams should get in... At least this way they could be transpartent about it

I've posted that.

Teams that would be in based on RPI and Not in with NET.

Seton Hall
Minnesota
Arizona State
Saint John's
Alabama
Oregon



Teams in on NET and out with RPI.

Ole Miss
North Carolina State
Indiana
Texas
Nebraska
Penn State



Would be out on both, but will probably be in:

Ohio State
 

I've posted that.

Teams that would be in based on RPI and Not in with NET.

Seton Hall
Minnesota
Arizona State
Saint John's
Alabama
Oregon



Teams in on NET and out with RPI.

Ole Miss
North Carolina State
Indiana
Texas
Nebraska
Penn State



Would be out on both, but will probably be in:

Ohio State

Interesting thanks...So 12 teams for 6 spots...Then you could have the top net team play the lowest rpi and such
 

Interesting thanks...So 12 teams for 6 spots...Then you could have the top net team play the lowest rpi and such

Or just leave out the teams that have dog records like penn state and Texas since the point of the game is to., you know, win
 

Or just leave out the teams that have dog records like penn state and Texas since the point of the game is to., you know, win

I mean i get your point, but unless you can make a pretty much completely even schedule (like the nba) it makes zero sense. Might as well just throw tenn st in by those standards. As a gophers fan i would love it if they only scheduled awful teams for every single non conference game. That would be a lot of fun said no one ever. So yea winning does matter but its not everything in a completely unbalanced system
 
Last edited:

I mean i get your point, but unless you can make a pretty much completely even schedule (like the nba) it makes zero sense. Might as well just throw tenn st in by those standards. As a gophers fan i would love it if they only scheduled awful teams for every single non conference game. That would be a lot of fun said no one ever. So yea winning does matter but its not everything in a completely unbalanced system



For sure,
Penn state has seven fewer wins than Minnesota because of schedule differences.

Certainly not because Minnesota managed to win some close games and Penn state managed to lose close games.
Plus winning close games is luck. Not skill.





Sarcasm
 

Honestly it doesn’t seem like the NET even factors the outcomes of games. It seems to just compile all of a team’s possessions and subsequently churns out macro efficiency numbers.

Indeed, it seems as if the approach to all this was wrong from the start. To my eye, and I freely admit I might just not be understanding it correctly, NET seems to emphasize the "why" more than the "what".

That may be fine for predicting which team is better in a comparison, but it seems to fall short in awarding accomplishment. And accomplishment is supposed to be the goal. Because, as someone else said, the point of the game is to...you know...win.
 

Indeed, it seems as if the approach to all this was wrong from the start. To my eye, and I freely admit I might just not be understanding it correctly, NET seems to emphasize the "why" more than the "what".

That may be fine for predicting which team is better in a comparison, but it seems to fall short in awarding accomplishment. And accomplishment is supposed to be the goal. Because, as someone else said, the point of the game is to...you know...win.

This is a pretty solid take on NET
 

For sure,
Penn state has seven fewer wins than Minnesota because of schedule differences.

Certainly not because Minnesota managed to win some close games and Penn state managed to lose close games.
Plus winning close games is luck. Not skill.





Sarcasm

Picking up on your sarcasm, but there seem to be people out there who actually think performance in close games is pure luck. Probably people who never played sports at a level higher than t ball.
 

For sure,
Penn state has seven fewer wins than Minnesota because of schedule differences.

Certainly not because Minnesota managed to win some close games and Penn state managed to lose close games.
Plus winning close games is luck. Not skill.





Sarcasm

I dont like net...i think its dumb...its not put in place to compare penn st and minnesota and if you think that...well....its used as a tool to help rank teams that play on a completley different playing field..which is half the reason i think its dumb...but hey guess what we need some sort of tool to do that...i dont think any of them are perfect but everyone pretty much agrees that 60 or so of the 68 teams pretty much paved their own path....so what is so wrong about letting the final bubble teams batlle it out based on multiple ranking system rather than just one...there's a difference between accuracy and precision
 

I dont like net...i think its dumb...its not put in place to compare penn st and minnesota and if you think that...well....its used as a tool to help rank teams that play on a completley different playing field..which is half the reason i think its dumb...but hey guess what we need some sort of tool to do that...i dont think any of them are perfect but everyone pretty much agrees that 60 or so of the 68 teams pretty much paved their own path....so what is so wrong about letting the final bubble teams batlle it out based on multiple ranking system rather than just one...there's a difference between accuracy and precision

It literally is in place to compare Minnesota and penn state.


Team X and team Y are competing for a bubble spot.

Team X beat Minnesota
Team Y beat Penn state

Who has the better quality win.
According to NET team Y should get in because they beat quad 1 penn state on a neutral court. While team Y gets a quad 2 win for beating Minnesota on a neutral court.


This is obviously a hypothetical situation. But you can see why doing a bad job can have a major impact on many teams. Not just the teams misplaced in the rankings but the teams that played those teams and how they compare in seeding and placement in the tourney.
 

Up to #26 in Massey. NET is hopefully just the first step in a move to a better mathematical model like Massey. NET focuses way too much on efficiency regradless of opponent and caps margin of victory at 10.

Sent from my SM-G930T using Tapatalk
 




Top Bottom