NCAA Net Rankings

[auror]

So now the debate has come full circle, back to what I complained most about the RPI

Who has played the better chumps, that's how to determine who's the better. Oh yeah baby. We beat number 200 and you beat number 300, so we're better, even if you beat us by 10 head to head

I hope more of you are beginning to understand why many of us mock these metrics

You're reading way too far into a couple outliers this season. Last year had very few outliers.

 

[Mr. Wonderful]

But those games are still useful for the models and still have predictive power. The games against better teams are a better indicator, but there's just less of them. Games against worse teams have a weaker signal, but help the overall sample.

Here's a KenPom blog where he talks a bit about the weighting:

https://kenpom.com/blog/pomeroy-ratings-version-2.0/

This is where my disagreement with the methodology lies.

The predictive power of mismatches drops off much faster than he allows. There's endless reasons outside of the sphere of maximizing performance that a game ends with "only" a 25 point win versus a 50 point win.

There's another aspect of this that draws my ire...

We're talking about buy games here, primarily. Our buy games tend to be of the budget variety. That's why we play the worst of the worst when it comes to buy games. MVSU is cheap. Want better? Open your wallet and donate to the AD.

 

[auror]

This is where my disagreement with the methodology lies.

The predictive power of mismatches drops off much faster than he allows. There's endless reasons outside of the sphere of maximizing performance that a game ends with "only" a 25 point win versus a 50 point win.

Have you researched it?

 

[Mr. Wonderful]

Have you researched it?

There's no need.

Just getting the walk-ons playing time screws up the methodology beyond repair, lol...

 

[auror]

There's no need.

Just getting the walk-ons playing time screws up the methodology beyond repair, lol...

But that effect will hit most teams in those games at those margins. Along with the rubber band effect, etc. Since you're comparing against expectation as set by past results that also include those effects (and not necessarily comparing linearly), their impact will be smaller than you'd think.

 

[Mr. Wonderful]

But that effect will hit most teams in those games at those margins. Along with the rubber band effect, etc. Since you're comparing against expectation as set by past results that also include those effects (and not necessarily comparing linearly), their impact will be smaller than you'd think.

I disagree completely here. I don't think it's safe to assume all teams react the same way to mismatches. In fact, I contend the opposite, and playing walk-ons is just one confounding factor of many.

 

[auror]

I disagree completely here. I don't think it's safe to assume all teams react the same way to mismatches. In fact, I contend the opposite, and playing walk-ons is just one confounding factor of many.

You can say you disagree, but the data and research suggest these matchups still help in the aggregate (though as KenPom says when you get up to like #350 not very much).

 

[Mr. Wonderful]

You can say you disagree, but the data and research suggest these matchups still help in the aggregate (though as KenPom says when you get up to like #350 not very much).

We're splitting hairs now. We only disagree on scale.

 

[Scottobot]

NET ranking is an improvement over RPI, and having an unbiased data driven ranking system is a good thing to have in combination with people’s general opinions on team rankings.

When it come time to reward #1 seeds, winning conference is critical and national ranking is critical. If UConn wins the Big East and remains a top 4 ranked team they will get a 1 seed regardless of net ranking. Net ranking comes more into play for bubble teams.

 

[alexd]

PS: Our Women are Number 3 in the NCAA Net Rankings, so they're doing a much better job with them

Coach Hurley obviously needs to get some tips fron Geno.

 

[HuskyHawk]

This is the crux of it, but not just this game.

Torvik has a "Game Score" that evaluates the result depending on margin, opponent, and location. It's a bit like looking at how the game is seen by the model. Alabama has 6 results that it qualifies as 98, 99, or 100 in Game Score (think of them as "98th+ percentile beatdowns"). 100s are quite rare even for great teams, Alabama has 2: when they beat #220 South Alabama by 57 and when they beat #109 Liberty by 45 on a neutral. Balancing it out slightly, they've got 3 games scoring sub-70. But twice as many great games as bad ones.

We've had only 3 games rated 98, 99, or 100: Creighton, Gonzaga, and UNC. So in 3 of our highest profile games we've played the best. We've got 2 sub-70 games (Miss Valley St and Seton Hall). Pine Bluff was also a 78 game score. We have a ton of "good" results, but we don't have enough real ass kickings to get into the top 5.

Which is just stupid, because a good team will put the bench in and rest starters in those games. They really need to adjust this so that any win by more than say 16 is treated the same. That's a blowout. Whether it becomes 20, 24, 30 or 36 is mostly irrelevant. In many cases it could have been.

 

[MU in MKE]

Which is just stupid, because a good team will put the bench in and rest starters in those games. They really need to adjust this so that any win by more than say 16 is treated the same. That's a blowout. Whether it becomes 20, 24, 30 or 36 is mostly irrelevant. In many cases it could have been.

T-rank (barttorvik) accounts for "garbage time" in his metrics.

T-Rank Methodology Update

Adam Chorlton and Bart Torvik discuss Wisconsin sports, mostly the Badgers.

adamcwisports.blogspot.com

 

[CL82]

Joe Lunardi isn't great with hitting on the bracket, but I do think we're going to see stuff like this in the conversation as we get closer to the tournament:

"The Boilermakers have a nation's-best six Q1 wins, four of them in the top half of Q1, and scheduled only three Q4 wins. UConn has a whopping eight wins in Q4 -- half its season total -- and therein lies the difference"

He's not wrong, but he excludes how many Q1 wins we have had.

 

[Hunt for 7]

If we stay healthy the rest of the year we will be top ten in both offensive and defensive efficiency on KenPom. The 8 Q4 wins will diminish in negative value but they are probably going to keep us out of the top 3 in NET this year. Certainly some of this may adjust on its own as the teams we beat get better or worse but as we saw last year the committee does not use them as much as people think hence our ridiculously low 4 seed from last year when compared to where we were on KenPom and Net. Win the BE regular season with a record no worse than 16-4 and at least get to the final in the BE tourney and we are a one or two seed no matter what the analytics say. 17-3 and BE final we are a 1 seed.

 

[tzznandrew]

He's not wrong, but he excludes how many Q1 wins we have had.

I don't think he did; I just didn't quote it. But, regardless, Purdue has more.

 

[Hi-Low]

Texas is back to a Quad 1 win for the good guys. UConn now tied with Purdue for most Quad 1 wins at 7.

The NET rankings are still a bit baffling though...Auburn has Zero Q1 wins and yet they are ranked 7 in NET ahead of UConn at 8.

 

[Poe]

Texas is back to a Quad 1 win for the good guys. UConn now tied with Purdue for most Quad 1 wins at 7.

The NET rankings are still a bit baffling though...Auburn has Zero Q1 wins and yet they are ranked 7 in NET ahead of UConn at 8.

Auburn is 11-2 against Q1-3, and we’re 9-2, so I guess they’re better? Who the hell knows.

 

[uh-no]

Texas is back to a Quad 1 win for the good guys. UConn now tied with Purdue for most Quad 1 wins at 7.

The NET rankings are still a bit baffling though...Auburn has Zero Q1 wins and yet they are ranked 7 in NET ahead of UConn at 8.

quad wins are an output of NET, not an input.

 

[Hi-Low]

quad wins are an output of NET, not an input.

That's what She said

 

[MU in MKE]

Auburn is 11-2 against Q1-3, and we’re 9-2, so I guess they’re better? Who the hell knows.

I was curious about this too... but look at their Q2 games. They absolutely spanked a bunch of good quality teams. They've only played two Q1 games, it'll even out as they play more and their true colors show. Either they walk the walk or they don't.