Glitches in the KP/NET Rankings Matrix...

[JustbrewitMan]

I think in general, they do a decent job trying to put some objectivity into ranking/seeding teams for the tourney.

But in looking at KenPom and NET this AM, it's clear you can still game their systems.

Gonzaga is KP #9, NET #8.

They have absolutely no business being ranked that high in either. They're 2-6 in Q1, 5-2 in Q2.

They've lost twice to St. Mary's, lost to Oregon St. (KP#82) & Santa Clara (KP#56).

All the bracketologists have them as a 8/9 seed, which is interesting.


Pomeroy needs to stop swimming in subscription money and fine-tune his model

 

[JustbrewitMan]

Meanwhile, Memphis is at KP #46, NET #48. With Q1/Q2 records of 6-2, 4-1. Projected as a 7 seed.

 

[UConnTrumbull]

Yeah I don’t know what kind of voodoo magic Mark Few uses to game the NET. Impressive really.

 

[navery12]

I think in general, they do a decent job trying to put some objectivity into ranking/seeding teams for the tourney.

But in looking at KenPom and NET this AM, it's clear you can still game their systems.

Gonzaga is KP #9, NET #8.

They have absolutely no business being ranked that high in either. They're 2-6 in Q1, 5-2 in Q2.

They've lost twice to St. Mary's, lost to Oregon St. (KP#82) & Santa Clara (KP#56).

All the bracketologists have them as a 8/9 seed, which is interesting.


Pomeroy needs to stop swimming in subscription money and fine-tune his model

I agree they shouldn't be rated that highly, but it's also such an extreme outlier that I don't think it requires a change to the model. They're rated so highly because their losses have mostly been close with minimal negative impact, and their wins have been with massive point differentials.

They lost to Oregon State like you said, then in the rematch won by 38 in a game with a 10 point expected spread. The Baylor game was 3 points and they won by 38. They're a really weird team where they could theoretically go on a run but I'm also not convinced they're actually good

 

[JustbrewitMan]

I agree they shouldn't be rated that highly, but it's also such an extreme outlier that I don't think it requires a change to the model. They're rated so highly because their losses have mostly been close with minimal negative impact, and their wins have been with massive point differentials.

They lost to Oregon State like you said, then in the rematch won by 38 in a game with a 10 point expected spread. The Baylor game was 3 points and they won by 38. They're a really weird team where they could theoretically go on a run but I'm also not convinced they're actually good

I'm not able to stay up late enough to watch many of their games but from the scores of their games, it's clear that Few knows he can't let his foot off the gas vs. most of the WCC teams (that have horrible KP metrics) or their ratings will drop like a stone.

Luckily, most of the WCC teams all suck more at defense than offense, so the Zags are able to post these absurd 30-40pt MOV wins on the reg.

I mean, they're clearly a "good" to "very good" team, but no way in hell are they Top 10 KP/NET.

 

[auror]

I agree they shouldn't be rated that highly, but it's also such an extreme outlier that I don't think it requires a change to the model. They're rated so highly because their losses have mostly been close with minimal negative impact, and their wins have been with massive point differentials.

They lost to Oregon State like you said, then in the rematch won by 38 in a game with a 10 point expected spread. The Baylor game was 3 points and they won by 38. They're a really weird team where they could theoretically go on a run but I'm also not convinced they're actually good

Yeah and they lost to Santa Clara by 4 and then won the rematch by 19 at Santa Clara.

 

[connVA]

Yeah I don’t know what kind of voodoo magic Mark Few uses to game the NET. Impressive really.

The same voodoo magic that Kelvin Sampson was using to keep Houston perennially pinned in the top 3 in their pre-Big 12 days despite feasting on a steady diet of Tulane and USF.

 

[gsmooth]

I'm not able to stay up late enough to watch many of their games but from the scores of their games, it's clear that Few knows he can't let his foot off the gas vs. most of the WCC teams (that have horrible KP metrics) or their ratings will drop like a stone.

Luckily, most of the WCC teams all suck more at defense than offense, so the Zags are able to post these absurd 30-40pt MOV wins on the reg.

I mean, they're clearly a "good" to "very good" team, but no way in hell are they Top 10 KP/NET.

Admittedly, I don't know all the granular details of how the NET rankings are calculated or KP. Does NET calculate a 50 pt blowout win vs a sub 300 team more than 20 pt blowout win vs these same sub 300 teams? If so, IMO that has to change. Assuming that is still the case, maybe the model should have a gradual sliding scale between 20-25 pt victory over the cupcakes or middle of the pack mid majors and then anything above 25 pts MOV should all be the same so a 26 pt win counts as much as a 55 pt win over these sub standard teams.

 

[Kolumbo]

Concordantly

 

[gtcam]

The same voodoo magic that Kelvin Sampson was using to keep Houston perennially pinned in the top 3 in their pre-Big 12 days despite feasting on a steady diet of Tulane and USF.

The difference is that he was beating those AAC teams regularly while the rest of the conference, including UConn, were beating each other up.
Sampson consistently gets more out of his players than any other coach in D1.
I don't think other programs could transfer from the AAC to what people deem to be the best basketball conference these days without seemingly missing a beat.

 

[RuffRuff]

I agree they shouldn't be rated that highly, but it's also such an extreme outlier that I don't think it requires a change to the model. They're rated so highly because their losses have mostly been close with minimal negative impact, and their wins have been with massive point differentials.

They lost to Oregon State like you said, then in the rematch won by 38 in a game with a 10 point expected spread. The Baylor game was 3 points and they won by 38. They're a really weird team where they could theoretically go on a run but I'm also not convinced they're actually good

Analysts were lumping them as one of the top 4-5 teams in the first month. Some of it was about surprising results but it was also about the player quality.

If you look at the roster they should be a solid team given a veteran PG and solid 3 big rotation. Hickman while defensively terrible shoots at a super high clip. Battle looks the part. It feels like a solid roster build missing some chemistry, and a bit flimsy defensively. They’ll be an interesting watch in the tourney with Fews track record. I can see them pulling off an upset.

 

[6Nattys4Us]

If you look at KenPom, he has O and D ratings by points per possession (yours on O, what you give up on D).

Gonzaga is 9th in O and 36th in D which explains them ranked #9 vs all the conversations on who they played, won or lost by, etc.

We are 14th and 112th respectively because we score per possession pretty well, but give up 3s and dribble drives too much and give teams points per possession. Gonzaga is better on D in that regard.

All we need to do is get better at team D (tough ask at this point end of Feb) but we have been much better in March than the regular season and there's no reason to think we cannot improve.

 

[auror]

Admittedly, I don't know all the granular details of how the NET rankings are calculated or KP. Does NET calculate a 50 pt blowout win vs a sub 300 team more than 20 pt blowout win vs these same sub 300 teams? If so, IMO that has to change. Assuming that is still the case, maybe the model should have a gradual sliding scale between 20-25 pt victory over the cupcakes or middle of the pack mid majors and then anything above 25 pts MOV should all be the same so a 26 pt win counts as much as a 55 pt win over these sub standard teams.

Systems like KenPom and Torvik do treat games with wide "mismatches" differently, but the NET is less open about those things. In general, a 50 point win does tell you more than a 20 point win. It's very hard to win by 50 even if teams are playing full out for 40 minutes. There can be elements of "computer trickery" from some teams, but generally these are drowned out by the greater signal of more information/possessions. But as mentioned most of the systems weight these games less in the models to less the impact from those teams who are running up the score against a bad schedule. I would assume the NET does as well.