2023-2024 MBB computer projections thread

[cyclones500]

Tennessee lost at home. UNC lost on the road.

UNC has a better resume and beat Tennessee head-to-head.

Good point, especially UT being at home, forgot about that detail. That alone probably justifies the difference in drop.

NC does have better resume and the head-to-head, but that already was reflected in his ranking, I would assume. I was assessing more relative to context of his rankings. Even then, it's not like it's egregious, but since he's picky about how others decide their rankings, I sometimes nitpick his.

Parrish no longer does Poll Attacks. I kind of enjoyed that for a while. Eventually it became kind of redundant and I think even he admitted he tended to say some of the same voters come up over & over.

 

[CyPunch]

The problem with scheduling aggressively in the non-conference as a Big 12 team is if you stub your toe an go say 8-5, you're now looking at ending the season at 17 or 18 total wins even if you do relatively well in conference play. 18-13 and 17-14 win/loss records are frowned upon.

Won't be an Iowa State problem next year with Maui, @ Iowa, a Big East team not named DePaul.

 

[StevieISU23]

CBS has ISU at #6 (IE: a 2 Seed).
CBS's DipShizzle Jerry Palm (Big10 Homer), has ISU as a #5 seed........
Hmmmm?

 

[Gunnerclone]

I’ve been watching it play out on Twitter, and I think that oglesby is just an idiot, like a lot of people are. Two things can be true. The non conference for a lot of teams wasn’t great, but the big 12 is still the best conference.

Setting himself for a bet against himself in the ncaa tourney. Big 12 doesn’t win, big gets only 1 in FF, big 12 gets none in FF, big 12 has upsets in 1st round, he wins. All of those are more likely than the Big 12 winning. Douchebag internet clout chaser.

 

[MJ271]

Apparently there’s a narrative the Big 12 is overrated mainly due to scheduling a weak non-con. While I agree a lot of the Big 12 teams didn’t schedule aggressively I don’t think it means the league is overrated.










Watching the level of play game to game this month compared to other conferences in the Reg season shows that in my mind.

I think there also might be some confusion and conflation now between NET and some of the metric-based predictive systems like Kenpom and Barttorvik. My understanding is that Kenpom and Barttorvik effectively have a margin of victory caps that mean a 40-point win over a cupcake really isn't any more valuable than a 20-point win. However, we don't have any clear indication that NET's net efficiency component has any sort of adjustment, which is where a team can benefit from those extra-large blowouts.

Regardless, the non-conference records speak for themselves. That original poster pointed out in the replies that the Big 12 this year was 3-9 against the ACC, but he failed to point out that the Big 12 was 9-3 against the SEC and went 7-4 in the Big 12-Big East challenge.

 

[Sigmapolis]

I think there also might be some confusion and conflation now between NET and some of the metric-based predictive systems like Kenpom and Barttorvik. My understanding is that Kenpom and Barttorvik effectively have a margin of victory caps that mean a 40-point win over a cupcake really isn't any more valuable than a 20-point win. However, we don't have any clear indication that NET's net efficiency component has any sort of adjustment, which is where a team can benefit from those extra-large blowouts.

Regardless, the non-conference records speak for themselves. That original poster pointed out in the replies that the Big 12 this year was 3-9 against the ACC, but he failed to point out that the Big 12 was 9-3 against the SEC and went 7-4 in the Big 12-Big East challenge.

Torvik filters out the possessions occurring in "garbage time" in this manner...

https://adamcwisports.blogspot.com/2018/09/t-rank-methodology-update.html

We use "garbage time" colloquially, but the model requires a specific mathematical definition of it. Torvik uses the "safe lead" time/score differential methodology originally proposed by Bill James...

https://www.slate.com/articles/sports/sports_nut/2008/03/the_lead_is_safe.html

For instance, Iowa State had a "safe lead" with ten minutes left in the game in the first game of the season against UW-Green Bay. So nothing that happened after that goes into the model...

https://www.barttorvik.com/box.php?muid=Green+BayIowa+St.11-6&year=2024

I don't know what KenPom does. I'm not sure what NET does. I find it funny, though, that proprietary systems like Torvik and KenPom have more about their methodology published than does NET.

 

[MJ271]

Torvik filters out the possessions occurring in "garbage time" in this manner...

https://adamcwisports.blogspot.com/2018/09/t-rank-methodology-update.html

We use "garbage time" colloquially, but the model requires a specific mathematical definition of it. Torvik uses the "safe lead" time/score differential methodology originally proposed by Bill James...

https://www.slate.com/articles/sports/sports_nut/2008/03/the_lead_is_safe.html

For instance, Iowa State had a "safe lead" with ten minutes left in the game in the first game of the season against UW-Green Bay. So nothing that happened after that goes into the model...

https://www.barttorvik.com/box.php?muid=Green+BayIowa+St.11-6&year=2024

I don't know what KenPom does. I'm not sure what NET does. I find it funny, though, that proprietary systems like Torvik and KenPom have more about their methodology published than does NET.

On Kenpom, I didn't find anything 100% definitive. This article talks about his methodological changes that were adding in some kind of effective cap or at least downweighting especially large blowouts. Pomeroy has made some other methodological changes since then and didn't mention anything about related to it, so I assume that same cap or downweighting is still utilized.

Barttorvik's FAQ does also say this:

Notably, T-Rank has a wider "spread" between top and bottom teams, probably because Kenpom has a much more significant cap on margin of victory.

It's unclear when that page was updated, but that would make sense, with Barttorvik being more aggressive in rating a team like Iowa State than Kenpom has been.

I very much agree on your point about NET's lack of transparency. The NCAA isn't making money off of it; frankly, there's no reason they shouldn't just publish the complete formula.

 

[bosco]

This. They play one big guy, 7' 215 lb Warren Washington, almost 30 minutes per game, and basically 7 other guards.

Don't tell ZRF.

 

[NiceMarmot]

My two takes on this:

1. NCSOS is meaningless. I have no idea why anyone cares about it. It's 13 games out of 31 (or 11 out of 31 in the case of 20-game leagues like the Big Ten, ACC, Big East). The reason Oglesby is talking about NCSOS instead of overall SOS is because if you start including conference games, Big 12 schedules will start to look harder while ACC schedules will start to look easier. Per the NET, Iowa State's NCSOS is 325 while Duke's is 126. For overall SOS, ISU is 69, Duke is 88. So who has the harder schedule? What Oglesby is doing is pure confirmation bias.
2. The NET should release its methodology, but when people bash it and then use Kenpom/Torvik numbers, that logic doesn't track to me. NET/Kenpom/Torvik rankings tend to correlate pretty closely. It's not like NET is like BPI and has a ton of statistical outliers.

 

[Sigmapolis]

On Kenpom, I didn't find anything 100% definitive. This article talks about his methodological changes that were adding in some kind of effective cap or at least downweighting especially large blowouts. Pomeroy has made some other methodological changes since then and didn't mention anything about related to it, so I assume that same cap or downweighting is still utilized.

Barttorvik's FAQ does also say this:

It's unclear when that page was updated, but that would make sense, with Barttorvik being more aggressive in rating a team like Iowa State than Kenpom has been.

I very much agree on your point about NET's lack of transparency. The NCAA isn't making money off of it; frankly, there's no reason they shouldn't just publish the complete formula.

I know Sports Reference SRS filters caps margin of victory for college *football* at 14. I don't know if they do something similar for the NFL. KenPom being more aggressive about capping MoV relative to Torvik does make sense given Torvik likes teams with a lot of blowout wins like Iowa State slightly more.

I know why the NCAA doesn't want to release the documentation on the NET: (1.) immediately starts a lobbying war by teams and conferences to set up the model to favor what they're best at (e.g., low-majors would want # of wins, power conferences would want strength of schedule, etc.) and (2.) it would probably lead to attempts to "game the system" with scheduling rather than doing it "naturally" or "for fun."

It's not like the NET is just "for fun" or even just one of many ranking systems alongside KenPom and the like that show up on the team sheets for the committee. It determines the quadrants of wins and losses, which seem to be the most powerful factor for how the committee is going to treat you in March.

 

[MJ271]

1. The NET should release its methodology, but when people bash it and then use Kenpom/Torvik numbers, that logic doesn't track to me. NET/Kenpom/Torvik rankings tend to correlate pretty closely. It's not like NET is like BPI and has a ton of statistical outliers.

For me, it's an issue of what the goal is of the systems are and how they're used. Kenpom and Barttorvik are predictive and are trying to describe how good a team actually is. The NET is kind of trying to be not only predictive with its net efficiency component but also reflect resumes with its game results component. See Florida State and the CFP for how those aren't always the same thing.

Then I guess part of my issue isn't really the NCAA's fault, but it's that people just tend to overreact to the NET when it first comes out. Since it's so aggressive with teams (and I believe is only based on this season's results), we end up with more outliers early on that lead to more people doubting advanced analytics as a whole field. Sometimes those early outliers are justified, sometimes they aren't. Kenpom and Barttorvik are less aggressive early on since they ease into the current season's metrics, which I think is the better approach anyway, but also avoids those overreactions.

It's true that much of this doesn't matter in the end, because they all do mostly converge. But NET shapes a lot of discussion during the season in a way that I think is unhelpful.

 

[Sigmapolis]

For me, it's an issue of what the goal is of the systems and how they're used. Kenpom and Barttorvik are predictive and are trying to describe how good a team actually is. The NET is kind of trying to be not only predictive with its net efficiency component but also reflect resumes with its game results component. See Florida State and the CFP for how those aren't always the same thing.

Then I guess part of my issue isn't really the NCAA's fault, but it's that people just tend to overreact to the NET when it first comes out. Since it's so aggressive with teams (and I believe is only based on this season's results), we end up with more outliers early on that lead to more people doubting advanced analytics as a whole field. Sometimes those early outliers are justified, sometimes they aren't. Kenpom and Barttorvik are less aggressive early on since they ease into the current season's metrics, which I think is the better approach anyway, but also avoids those overreactions.

It's true that much of this doesn't matter in the end, because they all do mostly converge. But NET shapes a lot of discussion during the season in a way that I think is unhelpful.

They could/should take a page from the CFP and only start releasing rankings significantly late enough in the season that you have enough data to make the model defensible/look less wacky.

That would be... well, roughly now. Non-con is 99.9% over and roughly 1/3rd of conference seasons done.

We'd have plenty to talk about in the meantime with KP and BT and BPI until then.

 

[cymac2408]

Makes sense to climb since UNC and Tennessee each lost.

Why, though, did Parrish drop Tenn 3 slots for losing to (his) #24 team (So Carolina, which he also bumped up 9 spots), but NC only 1 for losing to GT (not in his ranking and I'm sure not even close to it)? Not like it was a blowout loss (4 pts, NC lost by 1).

South Carolina fans should be PO. They have an 18-3 record with wins over VT, Kentucky and at Tennessee.

 

[ebState]

CBS has ISU at #6 (IE: a 2 Seed).
CBS's DipShizzle Jerry Palm (Big10 Homer), has ISU as a #5 seed........
Hmmmm?

CBS 25 and 1 is just Gary Parrish. It's basically a single AP ballot being updated everyday. I think he does a generally good job, but he's not in a bunker with Jerry Palm comparing notes or anything. Also single bracketologists is kind of pointless. Better when averaged out.

 

[alarson]

I know Sports Reference SRS filters caps margin of victory for college *football* at 14. I don't know if they do something similar for the NFL. KenPom being more aggressive about capping MoV relative to Torvik does make sense given Torvik likes teams with a lot of blowout wins like Iowa State slightly more.

I know why the NCAA doesn't want to release the documentation on the NET: (1.) immediately starts a lobbying war by teams and conferences to set up the model to favor what they're best at (e.g., low-majors would want # of wins, power conferences would want strength of schedule, etc.) and (2.) it would probably lead to attempts to "game the system" with scheduling rather than doing it "naturally" or "for fun."

It's not like the NET is just "for fun" or even just one of many ranking systems alongside KenPom and the like that show up on the team sheets for the committee. It determines the quadrants of wins and losses, which seem to be the most powerful factor for how the committee is going to treat you in March.

Do these models look just at MOV or do they factor in possessions in this as well?

Like, it seems like a 10 point win from a team averaging 65 possessions per game might be a bit more controlling win than a team averaging 75

 

[isuclone89]

One bracketologists has us gong to Spokane

If they got to Spokane I would get to see them. I know it's to far for most people, but walking distance for me.

 

[ZRF]

I’ve been watching it play out on Twitter, and I think that oglesby is just an idiot, like a lot of people are. Two things can be true. The non conference for a lot of teams wasn’t great, but the big 12 is still the best conference.

He is an idiot. But he's also right in pointing out the Big 12 non-con (as a whole) was trash. We were the biggest offender.

With that said, the SOS ratings are biased when there are X amount of "bad teams" on the docket. Kansas' non-con was 60 something? GTFO of here with that nonsense. It's a lot harder to win games against elite competiton, on the road or at neutral sites, than it is to beat a slew of decent teams. Yet SOS constantly dings schedules, likes Kansas', that have more elite competition than anyone else, but also have more creampuffs than some other schedules.

His premise isn't without merit, but to the degree he harps on it without looking at the entire picture is ludicrous. He basically discredits himself.

 

[NoCreativity]

That Oglesby is a total moron.

The Big 12 scheduled teams like Kentucky, Uconn, Duke, North Carolina, Tennessee, Auburn, Dayton, Miami, San Diego State, Marquette, and Creighton.

That's not even counting some of the teams that were suppose to be good and just made the tourney last year like Texas A&M and Missouri.

 

[rosshm16]

I think the reason they don't publish the formula for NET is so teams don't know exactly how it's determined and try to game it.

Goodhart's law - Wikipedia

en.wikipedia.org

 

[Sigmapolis]

Do these models look just at MOV or do they factor in possessions in this as well?

Like, it seems like a 10 point win from a team averaging 65 possessions per game might be a bit more controlling win than a team averaging 75

They're all adjusted for tempo.

"MoV" is more of a per possession concept than a per game concept.

Iowa State defeated Kansas 79-75 in 40 minutes of play with 70 possessions.

That's how we read it in the real world. The model reads it as something like...

ISU scored 1.1286 points per possession and gave up 1.0714 points per possession.

Kansas did the opposite.

So Iowa State won by 0.0571 points per possession when adjusting the score for tempo.

Doing this helps teams that play slow and appropriately downgrades teams that play fast.