MBB: Dork Polls 2018-19 Edition (starting w/ the NCAA's first release of NET)

[MChambers]

KenPom alert: The difference between #1 Duke and #2 Virginia in the rankings (31.36 vs 26.24 adjusted efficiency margin*) is greater than the difference between #2 Virginia and #13 Wisconsin (26.24 vs 21.73). Still some noise from preseason ratings but still- wow...


*Essentially, the expected margin of victory over an average D1 team if the game had 100 possessions: https://kenpom.com/blog/ratings-methodology-update/

Interestingly, T-Rank doesn't have such a large gap between Duke and #2, also Virginia. It's essentially the same difference as between Virginia and #4 MSU. Sagarin isn't updated yet, but even before yesterday's game it showed a similar gap as shown by Pomeroy.

[Troublemaker]

Some more criticism of NET, by the Washington Post's stats guy. I'm not a big fan of Greenberg's analysis and I don't see much new in this article:

https://www.washingtonpost.com/sport...=.17bba19ec01a

Neither am I. The author is either dumb or intellectually dishonest. All one has to do is click on the very links he provides to destroy his column.

There's not much new...except that another four weeks in, the NET is still screwed up compared to just about every other system. Is the NCAA NET right and everyone else wrong? Methinks not.

You need to read analysis like this more critically. Sometimes columnists have an agenda, an axe to grind, what-have-you. It's sad, but sometimes the media do not advance the truth.

I'm just so lazy, and I haven't had my coffee yet this morning. Can someone else pick apart Greenberg's column/analysis. I promise it's very easy to do. Once again, just click on his links.

[Troublemaker]

I will spork anyone who thoroughly picks apart that WaPo column.

[HereBeforeCoachK]

You need to read analysis like this more critically. Sometimes columnists have an agenda, an axe to grind, what-have-you. It's sad, but sometimes the media do not advance the truth.

I'm just so lazy, and I haven't had my coffee yet this morning. Can someone else pick apart Greenberg's column/analysis. I promise it's very easy to do. Once again, just click on his links.

You missed my point. I was not basing my opinion on the article....I even agreed that there was nothing new in the article....and links aside, simply taking the NET rankings now, there is some reason to be skeptical of the NET system.

FWIW, I'm a political media operative at a pretty high level...I'm fully aware of agenda driven reporting and columns. FULLY.

[JasonEvans]

Can we just table any conversation about the NET until January? It is abundantly clear that we need more data to judge it as it does not include any pre-season bias. In a few weeks we can start to look critically at it versus the ranking put forth by folks like BPI, Saragin, and Pom.

[Ian]

I will spork anyone who thoroughly picks apart that WaPo column.

How about I'll go with even with the small data it's already proven to be a huge improvement over the RPI.

Using his data and comparing the consensus 31 ranking to the NET ranking versus the RPI rankings:

Consensus	NET	RPI	Winner
1	3	10	NET
2	2	6	NET
3	5	1	TIE
4	1	20	NET
5	8	9	NET
7	6	3	NET
9	7	4	NET
10	4	25	NET

So just using the data he provided in his article, the NET beat RPI 7-0 with 1 tie in terms of which one is closers to the consensus rankings of the other 31 power ratings. Even with a small sample size, I can say with confidence that the NET is a huge improvement over the RPI, and I give the NCAA credit for taking a big step in the right direction.

[Troublemaker]

Some more criticism of NET, by the Washington Post's stats guy. I'm not a big fan of Greenberg's analysis and I don't see much new in this article:

https://www.washingtonpost.com/sport...=.17bba19ec01a

So, the problem with the WaPo stats guy and his column is the intellectual dishonesty. His central criticism of NET is that it strays far from the consensus Massey Ratings, which averages a bunch of computer rankings. But if you click on that link to Massey (last updated Dec 16th as of the time of this post), you'll see that almost all the individual computer rankings stray far from the consensus rankings. See the picture below that show the consensus rankings surrounded by individual computer rankings:




On the far right is NET, and just by the eyeball test, you can tell that it varies from the consensus about as much as any other rating system. By that measure, it's actually much better than the "SGR" (Singer) ranking system, which has UVA at #37.

This early in the season, you can always pick out one or two outliers from every computer ranking. On the far left is the well-respected Sagarin Predictor, which has Michigan St as the #2 team in the country. Even Sparty fans don't think they're the second-best team in the country. "TRK" (2/3rds of the way on the right) is the respected Pomeroy emulator T-Rank, and he has Kansas at #11.

Pomeroy himself is the 2nd column from the left ("POM"), and yes indeed his rankings align well with the consensus. BUT, as discussed previously, KenPom uses preseason expectations to anchor his rankings, which is why his look so reasonable. In another thread, DBR poster House P has attempted to estimate KenPom's rankings without the preseason anchor, and when he posted those, KenPom also starts to deviate from the consensus.

So, the WaPo stats guy basically slammed NET for something that is true of virtually all other computer ranking systems. All he had to do was click on his own link to Massey to see that.

Many people who read the WaPo column without doing so then reach the following conclusion...

There's not much new...except that another four weeks in, the NET is still screwed up compared to just about every other system. Is the NCAA NET right and everyone else wrong? Methinks not.

... which is just wrong, wrong, wrong. HBCK, my friend, I'm not ragging on you. I hope you understand that. It's a reasonable conclusion for you to reach if you don't follow the links out and just trust that WaPo's stats guy is being intellectually honest.

[-jk]

Wow: current Kenpom looks like a Duke-UVa final! (I wish!)

We're +4 on UVa, they're about +2 on Kansas - then things bunch up.

Gonna be a fun 1/19...

(Yeah, and it's still preloaded.)

-jk

[duke2x]

I'd agree if PR for the NCAA were high on my list of concerns, but I like watching rankings "take shape" as more data is accrued and wouldn't have minded if the NCAA had released NET in early November after *1* game had been played by each team. Even releasing it in late November as they did probably has been instructive about the power of sample size. Anyway, I hope the NCAA releases the formula behind NET at some point but am not very hopeful on that count.

NET is getting better but still has a mind of its own.

NET top 4 teams: Duke, UVA, MI, WI. NCSU is #6 and UNC is #20.
KenPom top 4 teams: Duke, UVA, Gonzaga, MI.
Torvik top 4 teams: Duke, UVA, MI, Gonzaga.

[BandAlum83]

KenPom alert: The difference between #1 Duke and #2 Virginia in the rankings (31.36 vs 26.24 adjusted efficiency margin*) is greater than the difference between #2 Virginia and #13 Wisconsin (26.24 vs 21.73). Still some noise from preseason ratings but still- wow...


*Essentially, the expected margin of victory over an average D1 team if the game had 100 possessions: https://kenpom.com/blog/ratings-methodology-update/

The AdjEm differential between KenPom #1 Duke and #2 UVA has closed significantly. It is down to a mere .73 (33.61/32.68)

Note how much higher the numbers are now versus when DukieTiger posted this original post on 12/19

[BandAlum83]

The AdjEm differential between KenPom #1 Duke and #2 UVA has closed significantly. It is down to a mere .73 (33.61/32.68)

Note how much higher the numbers are now versus when DukieTiger posted this original post on 12/19

Typo. .83 is the differential now.

[Mtn.Devil.91.92.01.10.15]

Typo. .83 is the differential now.

My math gives it a difference of 0.93 if those numbers are as stated.

[MChambers]

UVa has moved ahead of Duke at T-Rank.

[uh_no]

Note that the tightening of KP is mostly due to holding clemson to an absurd adjusted 66, moving their defensive efficiency from 86.4->84.7 after 1 game. Duke's numbers largely remained unchanged.

[jv001]

Note that the tightening of KP is mostly due to holding clemson to an absurd adjusted 66, moving their defensive efficiency from 86.4->84.7 after 1 game. Duke's numbers largely remained unchanged.

Thanks for that information. I guess things balance out over the season. In one of Virginia's recent games, they had something like a 30 point lead and Bennett took out his starters with about 3 or 4 minutes left. The really bad scrubs promptly let that lead dwindle down into the tens. So I guess that's pay back. GoDuke!

[BandAlum83]

My math gives it a difference of 0.93 if those numbers are as stated.

Late night/early morning math...I am better with numbers than this. Believe me! I have a very big brain and all the best words AND numbers!

[House G]

Wasn’t sure where to post this:

[House G]

Oops:
https://twitter.com/golfodds/status/...059411457?s=20

[Mtn.Devil.91.92.01.10.15]

Wasn’t sure where to post this:

Your indecisiveness was clear!

[MChambers]

I just discovered that on the team page for each team, T-Rank now has links titled "Similar Profiles" and "Similar Resumes". If you click on them, you'll get a list of past teams that had similar statistical profiles or similar resumes and how those teams did in the NCAAs. Bad news is that Duke's current closest match on Resume is UVa 2018!

Fun stuff! I could waste quite a bit of time on these comparisons. In fact, I just did.