Jan 4th Bracketology - Illinois #4 in first NET ranking

#26      
I wonder what metrics they're looking at.

But the rest of the number nerd models also have us top 10. It surprises me, but I'll take it.
 
#27      

Deleted member 380722

D
Guest
We have some really nice wins and no bad losses. Baylor was neutral and mizzou and Rutgers were losses on the road to good teams. 4-1 conference record along with knocking of duke on the road and Minnesota at home. Voila.
 
#28      

altenberger22

South Carolina
Colgate and Drake in the top 20.
Yep.

And our luck we'd sneak into the Top 4 to grab a #1 seed (in West Lafayette Regional) and then be forced to play the winner of the #4/#5 seed (Colgate/Drake) in a tough Regional semi-final game, before facing off against Boise State for a trip to the Final Four!

Brutal.
 
#29      
NET considers both the strength of opponent (Quadrant 1-4) and your Margin of Victory / Defeat. However, I believe that your margin is capped at 10 points.

From the linked .NET explanation post (dated 2018) "No longer will the NET use winning percentage, adjusted winning percentage and scoring margin."
 
#31      
From the linked .NET explanation post (dated 2018) "No longer will the NET use winning percentage, adjusted winning percentage and scoring margin."
The video in that link (around the 1:00 mark) states, that it now considers just 2 factors; one of which is your quality of win ("who'd you play, who'd you beat, and how good was that win"). I interpretted that to mean that "how good was that win" statement to mean that the amount you won by would still matter. At 1:16 in the video he also says that your Net Points per possession (your points per possession minus points your opponents get per possession) is also factored in still.

When I say "Margin of Victory" I mean how many points you won by, and from the video it sounds like that is still the case.
 
#34      
The video in that link (around the 1:00 mark) states, that it now considers just 2 factors; one of which is your quality of win ("who'd you play, who'd you beat, and how good was that win"). I interpretted that to mean that "how good was that win" statement to mean that the amount you won by would still matter. At 1:16 in the video he also says that your Net Points per possession (your points per possession minus points your opponents get per possession) is also factored in still.

When I say "Margin of Victory" I mean how many points you won by, and from the video it sounds like that is still the case.
The reference to how good was the win probably means the TVI.

The .NET has been reduced to two metrics.

1) Team Value Index (TVI).
This is a function with three inputs: Win/Loss, Quality of opponent, Location of game.

2) Adjusted net team efficiency.
This is the team efficiency adjusted for the quality of the opponent and pace. i.e. It is better to average 1.1pts/possession against Gonzaga on the road than against Chicago State at home.

The .NET used to consider the margin of victory. It was explicitly removed from the formula.
 
#35      
The reference to how good was the win probably means the TVI.

The .NET has been reduced to two metrics.

1) Team Value Index (TVI).
This is a function with three inputs: Win/Loss, Quality of opponent, Location of game.

2) Adjusted net team efficiency.
This is the team efficiency adjusted for the quality of the opponent and pace. i.e. It is better to average 1.1pts/possession against Gonzaga on the road than against Chicago State at home.

The .NET used to consider the margin of victory. It was explicitly removed from the formula.

Adjusted net team efficiency would naturally take into account margin of victory since it is points scored and points allowed per possession. Or am I missing something?
 
#36      
Adjusted net team efficiency would naturally take into account margin of victory since it is points scored and points allowed per possession. Or am I missing something?
I suppose the difference is margin per possession rather than margin per game (which would seem to penalize slower paced teams that typically have fewer possessions per game and thus likely smaller margins of victory)
 
#37      
I love seeing us ranked so high, but what kind of mathematical voodoo is involved to get us to number 4?

It's trying to solve a vexing problem. I mean, how do you rank teams? How do you understand quality of wins?

I'm not a big fan of moving to things like efficiency margin --I guess that makes me old school. It's relatively easy to construct non-intuitive results. Do you really want to reward a 30 point blowout and a 2 point loss, more than the team who played the same two games* and won both by 12?

That said, it's not a bad method, and predictive models are, on average, quite robust. And it seems like most experts agree that a hybrid model that uses win/loss and margin are a good compromise. Sagarin's ratings (for a while) used a simple 50-50 blend of "Predictor", based on point spread, and ELO, which was win/loss.

As long as they actually play the games, and have auto-bids to the tourney for winning games, everyone has a clear path to a championship regardless of margins. Still, we all enjoy seeing our teams fight for a good seed, and have good match-ups in the tourney.


*depends on the statistical model, but I'm assuming a simplified case.
 
#38      

danielb927

Orange Krush Class of 2013
Rochester, MN
I'm not a big fan of moving to things like efficiency margin --I guess that makes me old school. It's relatively easy to construct non-intuitive results. Do you really want to reward a 30 point blowout and a 2 point loss, more than the team who played the same two games* and won both by 12?

*depends on the statistical model, but I'm assuming a simplified case.

I agree it's more complicated than it seems on the surface.

For a predictive model? Almost certainly yes -- unless you have reason to believe that some parts of a 30-point win have less predictive value (like so-called "garbage time").

For a "ranking", which IMO seems to carry some notion of "earning" a given spot? Probably not.
 
#39      
Questionable given Lunardi's disdain for the Illini.

Speaking of Lunardi, today's Bracketology would have us in a very interesting bracket. We would be the 2-seed, and if we made it to the Elite Eight (i.e., beat 3-seed Creighton in the Sweet Sixteen), we'd presumably get a rematch in the Elite Eight of either #1 Baylor or #4 Mizzou ... I'd hope like hell for the latter! :p
 
#40      

drsmitty74

Rochester
Speaking of Lunardi, today's Bracketology would have us in a very interesting bracket. We would be the 2-seed, and if we made it to the Elite Eight (i.e., beat 3-seed Creighton in the Sweet Sixteen), we'd presumably get a rematch in the Elite Eight of either #1 Baylor or #4 Mizzou ... I'd hope like hell for the latter! :p
I was just going to post this exact same thing. I bet it killed Lunardi to put a #2 next to our name.
 
#41      
Speaking of Lunardi, today's Bracketology would have us in a very interesting bracket. We would be the 2-seed, and if we made it to the Elite Eight (i.e., beat 3-seed Creighton in the Sweet Sixteen), we'd presumably get a rematch in the Elite Eight of either #1 Baylor or #4 Mizzou ... I'd hope like hell for the latter! :p
A rematch with Mizzou in either Elite 8 or sweet 16 would be special....only thing that could make that better would be if fans were allowed to attend
 
#43      
I suppose the difference is margin per possession rather than margin per game (which would seem to penalize slower paced teams that typically have fewer possessions per game and thus likely smaller margins of victory)
Agreed. (Answering I11ini82)
 
#44      
It's trying to solve a vexing problem. I mean, how do you rank teams? How do you understand quality of wins?

I'm not a big fan of moving to things like efficiency margin --I guess that makes me old school. It's relatively easy to construct non-intuitive results. Do you really want to reward a 30 point blowout and a 2 point loss, more than the team who played the same two games* and won both by 12?
This is where the TVI comes into play. It measures wins/losses.

If one does not measure efficiency, how do you distinguish between the team that won the two games by 1pt, and the team that won against those same two teams by 30pts? One team is significantly better than the other. On the flip side, a team that loses each game by 1pt (The refs consistently make bad callsa against their center.) is much better than the teams getting blow out. Emphasizing wins-losses w/o taking into account the efficiency in some form penalizes hard schedules.
 
#45      
I suppose the difference is margin per possession rather than margin per game (which would seem to penalize slower paced teams that typically have fewer possessions per game and thus likely smaller margins of victory)

Yes, good point. But that still does reward margin of victory, to a fairly decent extent. Winning 75-60 in a 70 possession game is still better than winning 70-65 in a 70 possession game.

EDIT: Which for the record, in my opinion, is a good thing.
 
Last edited:
#46      
Speaking of Lunardi, today's Bracketology would have us in a very interesting bracket. We would be the 2-seed, and if we made it to the Elite Eight (i.e., beat 3-seed Creighton in the Sweet Sixteen), we'd presumably get a rematch in the Elite Eight of either #1 Baylor or #4 Mizzou ... I'd hope like hell for the latter! :p

I don't wanna see Baylor again unless we're playing them for the natty
 
#48      
Absolutely love the latest brackets! Big Ten gets all 4 two seeds and we only have to beat Creighton and Baylor to get to the Final Four!

I can't wait to see Baylor again! I honestly don't fear anyone with our team, as constructed we don't lose unless we f it up. Fts will get better I bet 70% or better will work.
 
Last edited:
#49      
I think I remember reading initially that the NCAA had partnered with someone like Microsoft to develop an “AI” or ML method for calculating the NET.

They could theoretically adjust their algorithm weekly to fit the teams to a let’s say a normal distribution. So it’s possible the factors are weighted differently week to week.
 
#50      
I think I remember reading initially that the NCAA had partnered with someone like Microsoft to develop an “AI” or ML method for calculating the NET.

They could theoretically adjust their algorithm weekly to fit the teams to a let’s say a normal distribution. So it’s possible the factors are weighted differently week to week.

AI (presumably) wouldn't be instructed to make a good television match-ups, or the highest revenue for the P5. Or maybe they'd find a way to sneak that stuff in?