Serious question about the NCAA's "Net Rankings" | The Boneyard

Serious question about the NCAA's "Net Rankings"

willie99

willie99

Loving life & enjoying the ride, despite the bumps
Joined
Aug 26, 2011
Messages
8,034
Reaction Score
28,078
How can PC only be 28? With 3 Big East teams ahead of them?

Looking at their overall record, their Quad 1 record, their Quad 2 record, it makes absolutely no sense. Looks like the math majors need to tweek their formula
 
willie99 said:
How can PC only be 28? With 3 Big East teams ahead of them?

Looking at their overall record, their Quad 1 record, their Quad 2 record, it makes absolutely no sense. Looks like the math majors need to tweek their formula
Expand ...
Close wins and close losses are treated the same. Winning doesn’t matter for the computers
 
willie99 said:
How can PC only be 28? With 3 Big East teams ahead of them?

Looking at their overall record, their Quad 1 record, their Quad 2 record, it makes absolutely no sense. Looks like the math majors need to tweek their formula
Expand ...
Let's wait until we see their post-season performance before we suggest they need to tweak anything. The season isn't over.
 
tzznandrew said:
Let's wait until we see their post-season performance before we suggest they need to tweak anything. The season isn't over.
Expand ...

I'm not suggesting anything about how good they are, only that some math formula seems messed up as of 2/24/2022
 
.-.
willie99 said:
How can PC only be 28? With 3 Big East teams ahead of them?

Looking at their overall record, their Quad 1 record, their Quad 2 record, it makes absolutely no sense. Looks like the math majors need to tweek their formula
Expand ...
agreed. i dont like being rewarded for beating a cupcake like LIU by 50+ points and only playing our 3 freshmen a combined 25 minutes in the process. This incentivizes teams to embarrass the competition when i'd rather have given our bench more run.
 
Last edited:
willie99 said:
How can PC only be 28? With 3 Big East teams ahead of them?

Looking at their overall record, their Quad 1 record, their Quad 2 record, it makes absolutely no sense. Looks like the math majors need to tweek their formula
Expand ...
The problem Providence has with all the statistical rankings is that they keep playing to the exact expectation of those statistical models. The models have them ranked somewhere between the low 20’s and high 40’s. So when they barely beat a Butler, or some other average team, the models think hey, that is exactly what should happen with a team ranked at that level, so I’m correct and I will keep them at that ranking.

They are just a bizarre team. How many 23-3 teams in a legit major conference ever have a 18 point loss to a team not close to making the tournament, and a 32 point loss to a team that’s a solid 6-9 seed. Then you add in winning basically every close game. The models just don’t know what to do.
 
bendm said:
agreed. i dont like being rewarded for beating a cupcake like LIU by 50+ points and only playing our 3 freshmen a combined 25 minutes in the process. This incentivizes teams to embarrass the competition when i'd rather have given our bench more run.
Expand ...

I'd like it 20x if I could. Incentivizing poor sportsmanship is ridiculous. A 20 point and 50 point win are the same thing for all intents and purposes.
 
7774 said:
The problem Providence has with all the statistical rankings is that they keep playing to the exact expectation of those statistical models. The models have them ranked somewhere between the low 20’s and high 40’s. So when they barely beat a Butler, or some other average team, the models think hey, that is exactly what should happen with a team ranked at that level, so I’m correct and I will keep them at that ranking.

They are just a bizarre team. How many 23-3 teams in a legit major conference ever have a 18 point loss to a team not close to making the tournament, and a 32 point loss to a team that’s a solid 6-9 seed. Then you add in winning basically every close game. The models just don’t know what to do.
Expand ...
Exactly. This is the AI problem, the training data needs to be complete, and cover every situation you expect to encounter. One or two anomalies should not skew your results. However, here the data are incomplete (e.g., Is the opponent at full strength?) and the output reflects it.
 
There are 2 general rules to the NET:

1. Blow out inferior opponents, and
2. Don't ever get blown out.

Providence didn't listen.
 
.-.
FYI, Providence ranks 11th in the latest RPI, UConn 25th. And if you think the RPI makes sense, UConn dropped from 24th to 25th after beating Villanova.

NCAA College Basketball RPI Rankings (updated today)

NCAA college basketball RPI rankings and ratings for 2026, updated daily for all teams and available historically for every day of the 2026 season.
www.teamrankings.com www.teamrankings.com
 
7774 said:
The problem Providence has with all the statistical rankings is that they keep playing to the exact expectation of those statistical models. The models have them ranked somewhere between the low 20’s and high 40’s. So when they barely beat a Butler, or some other average team, the models think hey, that is exactly what should happen with a team ranked at that level, so I’m correct and I will keep them at that ranking.

They are just a bizarre team. How many 23-3 teams in a legit major conference ever have a 18 point loss to a team not close to making the tournament, and a 32 point loss to a team that’s a solid 6-9 seed. Then you add in winning basically every close game. The models just don’t know what to do.
Expand ...
Seems odd that in a season of nearly 30 games, two bad performances (but Virginia and Marquette are not Quad 3 or 4, mind you) could have such a negative impact. Most good teams have 2 or 3 clunkers each year. Heck, Auburn barely escaped against Georgia and Missouri.

And isn't the point differential for a particular game capped in the model? I heard it's capped at 10 points. Or maybe that's KenPom.

PC has had a lot of close games--some of which are against average or subpar teams. I reckon that hurts their ranking.

Then you also have to consider adjusted efficiency of PC. Not sure where they stand with that stat.

**The NET, Mon Amour
Like that special someone, you capture my heart every February. And your mysterious ways only draw me closer...**
 
There are predictive models and resume models. The former is mostly margin of victory/defeat adjusted by location and opponent. Examples include KenPom, BPI, and Sagarin. The latter is just who did you beat and where, examples include Strength of Record (SOR), Wins Above Bubble (WAB), and KPI. Predictives are better at telling how good a team is. Resumes are better at telling you who had a better season, since in that case all we care about is wins and losses.

Providence has had a great season. But they aren't necessarily a great team. They're 8th in SOR (great season), and 41st in KenPom (good not great team).

The NET is a hybrid. It's something like 80% predictive and 20% resume, with likely a boost for beating high end teams. It's trying to say who is good, with a little extra credit for those who have had a great season. That's why Providence is 28th, in the middle of the two other systems.
 
Sphinx said:
FYI, Providence ranks 11th in the latest RPI, UConn 25th. And if you think the RPI makes sense, UConn dropped from 24th to 25th after beating Villanova.

NCAA College Basketball RPI Rankings (updated today)

NCAA college basketball RPI rankings and ratings for 2026, updated daily for all teams and available historically for every day of the 2026 season.
www.teamrankings.com www.teamrankings.com
Expand ...
I don't like the RPI either. I do prefer eye test mixed in with a little common sense. Never understood nonathetes and bureaucrats trying to decide things with computer programs written by people who never played the game
 
auror said:
There are predictive models and resume models. The former is mostly margin of victory/defeat adjusted by location and opponent. Examples include KenPom, BPI, and Sagarin. The latter is just who did you beat and where, examples include Strength of Record (SOR), Wins Above Bubble (WAB), and KPI. Predictives are better at telling how good a team is. Resumes are better at telling you who had a better season, since in that case all we care about is wins and losses.

Providence has had a great season. But they aren't necessarily a great team. They're 8th in SOR (great season), and 41st in KenPom (good not great team).

The NET is a hybrid. It's something like 80% predictive and 20% resume, with likely a boost for beating high end teams. It's trying to say who is good, with a little extra credit for those who have had a great season. That's why Providence is 28th, in the middle of the two other systems.
Expand ...
thanks for the explanation
 
.-.
If we're trying to be predictive. We already have the best in the business for that

Let Vegas rank the teams :)
 
VAHuskyFan said:
Also, how the heck is Houston #4 in NET with exactly zero Quad 1 wins on the season (0-3)?

They need to tone the margin of victory component down from wherever they have it. The rankings for Houston and Providence are just foolish.
Expand ...

Even worse, Houston is top 4 overall in NET, but the committee didn't put them in the top 4 SEED LINES on their mock bracket.
 
husky429 said:
I'd like it 20x if I could. Incentivizing poor sportsmanship is ridiculous. A 20 point and 50 point win are the same thing for all intents and purposes.
Expand ...
These are college men (and women). Blowing a team out is not poor sportsmanship IMO.

When I was growing up, I was on bad soccer teams. We lost every game one year. We made a choice to get better and work hard, and guess what happened? My HS team won the state championship back-to-back years, with a core group of the same kids who got annihilated when we were in 5th/6th grade. We ran right through some of the same towns that beat up on us growing up. I am for blow-outs. Sometimes in life you have to lose to learn how to win.
 
.-.
I posted a longer version of this article on another thread, but I think it explains the situation accurately.

Bleacher Report
Kerry Miller
2/15/22

Their 10 Q1/Q2 wins have come by a combined 59 points, and the two losses were by a combined 50 points.

Providence fans want you to believe all those close wins are emblematic of great coaching and a winning culture, but what the metrics see is a 21-2 team that is about 10 shots away from a .500 record.
 
SoFla Steve said:
I posted a longer version of this article on another thread, but I think it explains the situation accurately.

Bleacher Report
Kerry Miller
2/15/22

Their 10 Q1/Q2 wins have come by a combined 59 points, and the two losses were by a combined 50 points.

Providence fans want you to believe all those close wins are emblematic of great coaching and a winning culture, but what the metrics see is a 21-2 team that is about 10 shots away from a .500 record.
Expand ...
But they made all those 10 shots and are not at .500. They have one of the best records in the country and lead the Big East. At some point in the season, reality has to trump predictive.
 
SoFla Steve said:
I posted a longer version of this article on another thread, but I think it explains the situation accurately.

Bleacher Report
Kerry Miller
2/15/22

Their 10 Q1/Q2 wins have come by a combined 59 points, and the two losses were by a combined 50 points.

Providence fans want you to believe all those close wins are emblematic of great coaching and a winning culture, but what the metrics see is a 21-2 team that is about 10 shots away from a .500 record.
Expand ...
Providence fans want you to believe those wins were, in fact, wins. That is a sound position.

We are not trying to seed alternate-universe Providence with a .500 record.

Deprioritizing single-game outcomes is a bad idea.

The final minutes of a close regular season game are exciting based on the premise that the outcome matters.

If the NCAA decides that the outcome doesn’t matter - that a 1-point loss is really the same thing as a 1-point win - then those final minutes are meaningless. The regular season just becomes one long, continuous scoring exhibition for each team. However efficient the computers may find that, it’s boring as heck. There is a good reason every major sports league seeds teams by wins rather than “season-long scoring differential.”
 
CaptainRon said:
But they made all those 10 shots and are not at .500. They have one of the best records in the country and lead the Big East. At some point in the season, reality has to trump predictive.
Expand ...
Which is why they're rated highly in the things that look at past results. But that's not how math works in a predictive model
 
VAHuskyFan said:
Providence fans want you to believe those wins were, in fact, wins. That is a sound position.

We are not trying to seed alternate-universe Providence with a .500 record.

Deprioritizing single-game outcomes is a bad idea.

The final minutes of a close regular season game are exciting based on the premise that the outcome matters.

If the NCAA decides that the outcome doesn’t matter - that a 1-point loss is really the same thing as a 1-point win - then those final minutes are meaningless. The regular season just becomes one long, continuous scoring exhibition for each team. However efficient the computers may find that, it’s boring as heck. There is a good reason every major sports league seeds teams by wins rather than “season-long scoring differential.”
Expand ...
It strikes me that your argument is indisputable.
 
jeff99 said:
Close wins and close losses are treated the same. Winning doesn’t matter for the computers
Expand ...
No, it treats close wins as only slightly better than close losses, which they are. It treats blowout wins as much better and blowout losses as much worse.

The problem is, one of their losses is a 30-point blowout, and virtually all of their wins are squeakers. "Luck" factor through the roof.
 
.-.
You must log in or register to reply here.

Forum statistics

Threads
168,631
Messages
4,586,740
Members
10,497
Latest member
Orlando Fos


Top Bottom