NET and KenPom Tracker 23-24 (SU = 84 3/9/24)

[jncuse]

I generally think I understand how the NET and KP are working. Thank you for the education on this!

But then I see Michigan State, which is 11-7, 0-3 on the road, 1-2 at neutral sites, 2-5 in Quad 1, and 1-2 in Quad 2. How does that math add up to a 24 NET ranking?

See my latest post on NET just above this. NET is doing some crazy ass things right now with Road Blowout Wins against bad teams. And it shouldn't be in my opinion. Something is flawed with that aspect of their formula.

 

[jncuse]

Next two games are now Q3. Like I said the other day, we need to go 3-0 in this home stretch to dance.

We don't have enough Q1 opportunities to afford a bad loss, especially at home. Need to win every home game besides a split with UNC/Clemson and that might change if Clemson keeps falling towards the bubble.

Similar to Pitt, NC St is straddling that #75 spot (right now #78) and we play them twice.

It would be great if NC St does well the rest of the way (except us).

 

[nycfan]

Ask last year's Clemson team what they think about that! Clemson had 4 Q1 Wins, NC St had 1. Clemson had 5 bad losses, NC ST had none. NC St got in.

The fact that we on track to have Zero bad losses is a strong positive to have as a bubble team. (its very critical to our hopes that we sweep Louisville, Notre Dame and FSU at home)

Wow, your memory is better than mine but this certainly is very encouraging since we could very well end up being in a similar situation as NC St. That said, would feel a lot more comfortable if Oregon does in fact become a second Q1 (and Pitt stays there of course).

 

[CuseOrange15]

i think the "old RPI" site is busted ...the numbers are all wrong

it has SU with a record of 10-5, for example

where are you finding your RPI numbers?

I use this site for NET, RPI, ELO…

 

[jncuse]

Wow, your memory is better than mine but this certainly is very encouraging since we could very well end up being in a similar situation as NC St. That said, would feel a lot more comfortable if Oregon does in fact become a second Q1 (and Pitt stays there of course).

It's not all memory. I remember the generals but have to cheat for the specifics.

NET Ranking - Men's College Basketball | WarrenNolan.com

2023 Men's College Basketball NET Ranking: NCAA Evaluation Tool (NET) a ranking of teams based on five factors, team value index, net efficiency, winning percentage, adjusted winning percentage, and scoring margin.

www.warrennolan.com

NC St was 1-7 vs Q1 7-4 in Q2, and had a clean resume (no Q3 or Q4 losses)
That could be somewhere we are heading.

The biggest difference is that I their standalone NET was 45.. we will never get that high unless we go to Louisville and win by 35 (because apparently crushing bad teams on the road is huge under NET).

But standalone NET is not supposed to be that important so who knows.

 

[benjams77]

The one NET comparison to Cuse I am struggling with is Alabama

Bama is 9 - 12-6
Cuse is 69 - 12-5 (not counting Chaminade)

Bama has 4 wins vs top 100
vs Ind St. 28
vs S. Carolina 61
Neutral - Oregon 58 - shared win
@ Miss St.40 - quad 1

Cuse has 7 top 100 wins
Neutral - Oregon 58 - shared win
Pitt - Home and away - 68 - 1 quad 1 win
LSU - 90
Cornell - 88
Miami 65
Boston College 86

Bama played and lost to 2,3,4,11,33,(46 Neutral)
Cuse played and lost to 6, 9, 14, 35, (56 Road)

I get their Net losses are a "little" better by rankings but does it make up 60 spots in NET?

 

[Longtimesu]

Thanks jncuse for, as always, explaining (not defending) how the KP system works. I think most people here have a general understanding of it, but your posts have always greatly increased, at least, my understanding of the how/why.

Very much appreciated Jncuse. Especially for an ole timer like me who was never good at math. And now with all this metric stuff. Heck, I gotta 9 on my algebra final. Never could understand why adding a letter to a number makes sense. Or that adding a number with a letter to another letter gives you an altogether different letter.

So thanks for your explanations of this NET and Kenpom stuff. You make it easy to understand…Well almost.

 

[Cusefan0307]

According to the Matrix(I realize not up to date) Oregon and Miami are straddling each side of the cut line. Wake is just above Oregon. It would really benefit us to beat Wake and be 3-0 against those schools at the end. That’s the type of hair splitting the committee will do when they get down to the last teams.

 

[Cusefan0307]

The one NET comparison to Cuse I am struggling with is Alabama

Bama is 9 - 12-6
Cuse is 69 - 12-5 (not counting Chaminade)

Bama has 4 wins vs top 100
vs Ind St. 28
vs S. Carolina 61
Neutral - Oregon 58 - shared win
@ Miss St.40 - quad 1

Cuse has 7 top 100 wins
Neutral - Oregon 58 - shared win
Pitt - Home and away - 68 - 1 quad 1 win
LSU - 90
Cornell - 88
Miami 65
Boston College 86

Bama played and lost to 2,3,4,11,33,(46 Neutral)
Cuse played and lost to 6, 9, 14, 35, (56 Road)

I get their Net losses are a "little" better by rankings but does it make up 60 spots in NET?

Bama did a great job blowing out cupcakes and outside of yesterday their losses were close. Margin of victory and defeat has killed our metrics and helped Bama.

 

[anglerman]

Bama did a great job blowing out cupcakes and outside of yesterday their losses were close. Margin of victory and defeat has killed our metrics and helped Bama.

But it can’t be by 60 spots !

 

[Sbucketball]

According to the Matrix(I realize not up to date) Oregon and Miami are straddling each side of the cut line. Wake is just above Oregon. It would really benefit us to beat Wake and be 3-0 against those schools at the end. That’s the type of hair splitting the committee will do when they get down to the last teams.

One important thing I'm looking at is that we beat the other ACC bubble teams who may have similar resumes (no bad loses but no great wins), so if it's a pick-em for the committee, they look at the head-to-head, and also that helps us edge out teams in the ACC rankings.

 

[jordoo]

As many have pointed out, scoring margin is overweighted. I’m not saying it shouldn’t be considered, just that it’s something all of these algorithms should make a correction for.

 

[jncuse]

But it can’t be by 60 spots !

You wouldn't think, but I'll attempt to show the math. I'll use the KP as they have more discrete "value" on what the difference between #8 and #74 means... I know KP and NET are not exact parallels but they are treating the 2 teams relatively the same (Alabama 8, Syracuse 74).

KP views Alabama as being about 9.5 points better than us on a neutral floor (that is the difference between team #8 and #74) . So over 18 games, that means they have performed about 170 points better than us (relative to SOS).

Where do those 170 points come from?

Factor #1 - Blowing Out 3 Worst Teams at Home in OOC (96 of the 170)
This is what each team did against their 3 worst home opponents
Alabama won by 131 points (avg NET 240)
Syracuse won by 35 points (avg NET 218)
Difference of 96 points!

Factor #2 - Margin of Defeat in Losses (63 of the 170)
The losses of both teams are very strong -- Alabama's are probably a bit better, but for this assume
Alabama 61 points (6 games)
Syracuse 114 points (5 games)
Reduce Alabama to 51 to equalize games... so the difference is 63

For the remaining 9 games (8 for Bama) they then only be about 11 points better than us - so basically equal.. It makes sense, since amongst those 9 games we had nice 20 point wins against LSU and Oregon and almost all our wins against decent teams are all around that 10 point mark. Games against the middle have been where we have performed the best this year.

But as you can see from above, it is very much driven by 8 games (or 9 for Alabama):
1) How we did in our 3 "easiest" games in OOC compared to them.
2) How they did better in their losses.

 

[JustPherU]

I use this site for NET, RPI, ELO…

View attachment 236774

I forgot about this site, thanks for posting it.

 

[OrlandoCuse]

i think the "old RPI" site is busted ...the numbers are all wrong

it has SU with a record of 10-5, for example

where are you finding your RPI numbers?

RealtimeRPI site is busted but there’s several others that still produce the original RPI formula.

This one has us currently at 10. It’s pretty shocking to see the difference between the old RPI and current NET. It’s all driven by the move away from strength of schedule and its derivatives in favor of a more KP developed metric around offense/defense efficiency. IMO, the truth is somewhere in between, which would put us around 35 if you just used old RPI and NET. Layer in KP and we’re around 50.

NCAA RPI Rankings - RealGM

Check out the latest NCAA college basketball RPI rankings.

basketball.realgm.com

 

[OrlandoCuse]

The NET works in mysterious or dubious ways.

I think i will have to take back my comment about margin - at least for road games. Its seems the NET rewards blowing out cupcakes on the road, more than anything else right now (see the data below). It does clearly reward road wins and good wins, but apparently not as much as purely mutilating awful teams on the road. The Gonzaga, Drake, and Providence jumps in NET this week are baffling. (see below)

As crazy as a claim this might be, I think they might be applying their NET formula wrong. and have a sign going in the wrong way in their algo. They state they are trying to diminish the impact of blowouts, and in fact the data is showing the opposite. Blowout wins on the road (possibly not at home) appear to be getting the greatest reward under this system.

Biggest improvements in NET this week (for teams 70 or above), as pulled from Warren Nolan (NET Delta column). Some real strange movements - great to see we jumped so that is the biggest positive here. But the Gonzaga, Drake

NET Ranking - Men's College Basketball | WarrenNolan.com

Gonzaga +20 (2 Q4 Road Wins by 67 points)
Drake +20 (2 Q4 Wins, one on the road, won by 70 points)
Virginia +16 (2 Q2 Wins, one road, won by 17 points)
Colorado +14 (Q2 Win + Q4 Win, both home, won by 49 points)
Syracuse +12 (Q1+Q2 Win, one road, won by 14 points)
Providence +11 (Q4 Road Win, won by 38 points)
New Mexico +11 (Q1 Win + Q4 Road Win, won by 32 points)
St. Mary's +10 (Q1 Road Win at San Fran)

We can certainly see why all these teams improved in NET - they deserved to. But the two things that immediately jump out is the fact that Drake and Gonzaga had the biggest climb this week. Gonzaga did the equivalent of going on the road to Louisville twice this week and winning by 67 points total. Impressive, sure, most teams are not going to win by that much but reason to jump your NET from 50 to 30?

Drake basically did the exact same thing as Gonzaga, destroying 2 cupcakes by impressive margins, and moved from 64 to 44.

Providence played one game at #309 Depaul, They destroyed them by 38, OK that's nice, but should your NET increase from 68 to 57 because of that? Heck they improved by as much as we did, and our week was much more impressive in terms of what matters for selecting NCAA teams.

The moves of Virginia, Syracuse, New Mexico, St Mary's all make sense. They reward teams for quality wins, especially if its on the road. But yet they are still rewarded less than a Providence blowout win at Depaul.

Thoughts?

I think you nailed one of the flaws I was seeing with NET. Better to blowout a Q4 team on the road than beat a quality team at home by less than 10.

 

[jncuse]

I think you nailed one of the flaws I was seeing with NET. Better to blowout a Q4 team on the road than beat a quality team at home by less than 10.

For NET it is, but not for making the tournament where Q1 and Q2 wins reign supreme and standalone NET is still mostly ignored in your profile as long as its satisfactory. We have seen this week that NET still rewards you for winning.

The interesting thing though is that NET is valuing blowout Q4 wins even more than KP does. Providence and Gonzaga moved up more spots in NET than KP ... which is why I think they tried to make an adjustment and its working in the wrong direction.

 

[jncuse]

As many have pointed out, scoring margin is overweighted. I’m not saying it shouldn’t be considered, just that it’s something all of these algorithms should make a correction for.

Margin is fine for KP - but it shouldn't dominate the NET as much as it does. KP will tell you himself its ridiculous to pick tournament teams based on his system.

You have to go back to the roots of KenPom. KenPom was set up as a ranking system not only to rank the Duke's, Alabama's and Syracuse of the world, but to rank Oral Roberts, Lafayette and Norfolk St. It was general interest for some like myself, but also a great tool for degenerates. The only way to effectively rank those really bad teams (and all teams) who continuously lose and hardly ever win OOC, is to rely heavily on margin.

His site was being used by gambling degenerates who wanted to find an advantage in games like Coppin St vs Hampton. There was a time about 15-20 years ago, where people would seek out large line deviations in any NCAA game using KP and then auto-bet them ... and it was a very succesful strategy apparently... although i didn't gamble at the time. Now KP can't really be used like that, because Vegas changed their lines to most of the time reflect his rankings. So it's hard to say it didn't work - it did.

But it will still spit out anomolies. And these will tend to show up in "Luck". Its not because of how he viewed them to start the year which some speculate. Its typically because of blowouts all going one way and not balancing... or being really lucky or unlucky in close games.

 

[CardiacCuse]

RealtimeRPI site is busted but there’s several others that still produce the original RPI formula.

This one has us currently at 10. It’s pretty shocking to see the difference between the old RPI and current NET. It’s all driven by the move away from strength of schedule and its derivatives in favor of a more KP developed metric around offense/defense efficiency. IMO, the truth is somewhere in between, which would put us around 35 if you just used old RPI and NET. Layer in KP and we’re around 50.

NCAA RPI Rankings - RealGM

Check out the latest NCAA college basketball RPI rankings.

basketball.realgm.com

going 13-5 with a top 20 schedule should not put a team down in the 60s!

argggh

 

[CardiacCuse]

You wouldn't think, but I'll attempt to show the math. I'll use the KP as they have more discrete "value" on what the difference between #8 and #74 means... I know KP and NET are not exact parallels but they are treating the 2 teams relatively the same (Alabama 8, Syracuse 74).

KP views Alabama as being about 9.5 points better than us on a neutral floor (that is the difference between team #8 and #74) . So over 18 games, that means they have performed about 170 points better than us (relative to SOS).

Where do those 170 points come from?

Factor #1 - Blowing Out 3 Worst Teams at Home in OOC (96 of the 170)
This is what each team did against their 3 worst home opponents
Alabama won by 131 points (avg NET 240)
Syracuse won by 35 points (avg NET 218)
Difference of 96 points!

Factor #2 - Margin of Defeat in Losses (63 of the 170)
The losses of both teams are very strong -- Alabama's are probably a bit better, but for this assume
Alabama 61 points (6 games)
Syracuse 114 points (5 games)
Reduce Alabama to 51 to equalize games... so the difference is 63

For the remaining 9 games (8 for Bama) they then only be about 11 points better than us - so basically equal.. It makes sense, since amongst those 9 games we had nice 20 point wins against LSU and Oregon and almost all our wins against decent teams are all around that 10 point mark. Games against the middle have been where we have performed the best this year.

But as you can see from above, it is very much driven by 8 games (or 9 for Alabama):
1) How we did in our 3 "easiest" games in OOC compared to them.
2) How they did better in their losses.

ok, yeah, "the math"...

anyone can twist statistics however they want...

and these are twisted all kindsa ways, imo

 

[baggerbob]

I think you nailed one of the flaws I was seeing with NET. Better to blowout a Q4 team on the road than beat a quality team at home by less than 10.

Point differential should never be used to rate a team.
You can't take into account missing players, sick, or players with minor injuries.

 

[mark79]

But our luck jump from 6th to 1st after the win. It is all about luck.

Rather be lucky than good I always say, particularly on the golf course!

 

[jncuse]

RealtimeRPI site is busted but there’s several others that still produce the original RPI formula.

This one has us currently at 10. It’s pretty shocking to see the difference between the old RPI and current NET. It’s all driven by the move away from strength of schedule and its derivatives in favor of a more KP developed metric around offense/defense efficiency. IMO, the truth is somewhere in between, which would put us around 35 if you just used old RPI and NET. Layer in KP and we’re around 50.

NCAA RPI Rankings - RealGM

Check out the latest NCAA college basketball RPI rankings.

basketball.realgm.com

TLDR Version - The RPI Sort of Stank and could be gamed, or was just bad scheduling luck - it needed to be tweaked. The NET fixed those gaming issues but created the new problem of making some cupcake games that everybody is going to win, mean more if you blow them out more than others do.

They will need to find a balance of the two.
Maybe as you suggest, run the NET as it is now, run the RPI, average them and call that the new NET.

Here were the biggest problem with the RPI
1) The RPI was more easily gamed by some conferences who tended to try to schedule heavily in the 125-200 sweet spot and add some of them as road games. The MVC was really good at this.

2) The RPI formula really punished teams that played a couple sub 300 teams, and sometimes that was just an element of bad luck as the game was scheduled well in advance and that team now sucking is beyond your control.

3) For mid majors and small conferences the RPI generally threw out garbage that the committee did not trust. You would see a 24-7 Texas St team at RPI #38, and the committee would auto toss the ranking as meaningless. Margin based rankings viewed as better ways to find the best of these teams.

So on to the NET we went.

Net Fixed #1 and #2, by basically saying, OK you can play a bunch of sub 2 teams instead of 150 teams, but you are going to have to beat them by 20 instead of 12, to earn the same rank In theory that makes sense. But it created a new problem. Whether its RPI #175 or RPI #325 all tourney level teams are going to win 98% of those games.

The problem though as we are finding now, is that outperforming that "20" target, is just as important as outperforming a "pick Em" game for NET. So winning that game by 30 is just as good , beating a Q2 team by 10 points.

As for factor #3, I don't think NET has really solved that issue. The committee gives no credit to small conference teams in that #35-#50 range... it still seems hesitant to do so, although those 2 or 3 smaller conference teams identified as #35-#50 in the NET each year, are better now than what the RPI spit out.

 

[CardiacCuse]

The one NET comparison to Cuse I am struggling with is Alabama

me too! look at the list of teams higher than SU in NET, though, its a lot more than just Alabama

Auburn is 16-2 (great job) and they are 6th in NET...but guess what ? ZERO quad 1 wins! (0-2)

Mcneese is 65th (4 spots higher than Cuse) - Zero quad 1 wins, 2-2 against quad 2

Pitt is 1 spot higher than Cuse at 11-7 on the year

Creighton is 14-5 on the year 6-5 combined vs quad 1 and 2...meanhwile SU is 4-5 combined in Q 1 and 2...pretty similar right? nope. Creighton is 11th, SU 69th. Marquette is almost the same metrics.

Mich St has a garbage record vs quad 1 and 2...yet in the top 25

Gonzaga did wipe the floor with the Cuse...but guess what? ZERO quad 1 wins! yet 39 places higher than the orange at 12-5 on the season.

Villanova, Texas AM and St Johns all have 7 losses on the season...and well above SU.

Xavier is 10-8...and 20 spots higher than 13-5 cuse...meanwhile cuse has played a top 20 schedule in the NCAA!

I could go on...

Overall record should matter more...margin of victory should matter less

Obviously NET is incorporating KenPom...

I wonder if their algorithm is messed up honestly...I mean...who would really be smart enough to make sure it isnt???? highly possible that whoever is in charge of the NET is not that good at math, imo

 

[jncuse]

Point differential should never be used to rate a team.
You can't take into account missing players, sick, or players with minor injuries.

For the same reason, couldn't you then say W-L record shouldn't be used to rate a team, because you can have "Fake"*** Wins and Losses caused by missing players or injuries.

Margin is a generally effective way to rate a team over a large enough sample of games. There will be some one off variations, for most teams they will largely balance, but there will be the anomalies too like SU, Pitt and Bama this year who consistently performed one way in certain games.

The key thing though is that NCAA tournament is largely just selected by W and L, and the quality of those teams. It doesn't care about your margin.

*** Fake Wins is a term I stole from a Raptors board I am on. In 2021/2022 it was used by bitter fans who wanted us to tank, while we went 48-34. They would label each win as fake because somebody was usually out on that other team (especially in that Covid year where basically everybody in the league had to sit 3-4 games at various points of the year)