2023 Team Outlook

[PointWarrior]

UConn

Who is finishing behind us in this scenario? Creighton, nova, Xavier or UConn ?

[Pakuni]

That's good.  He's far more accurate than the Joe Beer Guts on Scoop or 99% of most college basketball analysts.  It's why coaches use him

I don't say this often, but panda is correct here. KenPom is good once there's some data, but his preseason predictions are no better - and no less obvious - than the AP Poll (or Joe Beer Guts).

[panda]

UCONN is fairly easily the worst of this group. I think marquette fans hold them in a higher regard because they are a poor matchup for marquette. I think uconn will be better than MU this year, but there are alot of scenarios where they won't be.

Going out in a limb, but I really think UConn is a dark horse to win the league this year.

We all know Sanogo as one of the best, if not the best returning player in conference. I'm a big fan of Andre Jackson as a do it all guy ready to take a huge step this year. Diarra and Alleyne are both below the radar but big time additions to the rotation. These are both proven high major players that won't have trouble with a change of scenery.

The biggest addition that's not really talked about is the Tristan Newton addition. Without him, I'm not remotely as high on UConn as I am now. Two way player who vastly improved his overall offensive game last year. If he fully embraces the distribution role, he'll be a step up from Cole. Bigger more physical guard who is just as good a scorer and a better passer than Cole.

[TAMU, Knower of Ball]

I don't say this often, but panda is correct here. KenPom is good once there's some data, but his preseason predictions are no better - and no less obvious - than the AP Poll (or Joe Beer Guts).

KenPom preseason is just an educated guess, but it is a more educated guess than the ones made by the AP Poll and Joey Beer Guts

[Pakuni]

KenPom preseason is just an educated guess, but it is a more educated guess than the ones made by the AP Poll and Joey Beer Guts

Is it? Based on what? It doesn't seem to be any better at predicting the future.
KenPom's preseason top 25 vs final top 25:
Correct teams in top 10: 5
Correct teams in top 25: 17

AP's preseason top 25 vs final top 25:
Correct in top 10: 7
Correct in top 25: 15

It's just one year - because I have no time or inclination to do multiple years - but it doesn't seem as if Ken's preseason algorithm is any more accurate than the collective wisdom of the Omaha World Herald, Raleigh News & Observer or Cleveland Plain Dealer. Heck, when it came to the top 10, Ken's algorithm did no better than a coin flip at predicting what it would say at season's end.

[TAMU, Knower of Ball]

Is it? Based on what? It doesn't seem to be any better at predicting the future.
KenPom's preseason top 25 vs final top 25:
Correct teams in top 10: 5
Correct teams in top 25: 17

AP's preseason top 25 vs final top 25:
Correct in top 10: 7
Correct in top 25: 15

It's just one year - because I have no time or inclination to do multiple years - but it doesn't seem as if Ken's preseason algorithm is any more accurate than the collective wisdom of the Omaha World Herald, Raleigh News & Observer or Cleveland Plain Dealer. Heck, when it came to the top 10, Ken's algorithm did no better than a coin flip at predicting what it would say at season's end.

It is because it is based on unbiased data instead of biased opinions.

Comparing preseason to final poll is not an accurate indicator. No one can predict things like injuries, transfers, etc.

[Pakuni]

It is because it is based on unbiased data instead of biased opinions.

Comparing preseason to final poll is not an accurate indicator. No one can predict things like injuries, transfers, etc.

If comparing his preseason rankings to his final rankings isn't an accurate indicator of the predictive power of his algorithm, what is? How do you judge?

Are the predictions of AP voters not subject to the same vagaries of injuries, transfers, etc. as KenPom?

Which of the teams in KenPom's preseason top 10 that didn't live up to that billing underachieved as a result of injuries, transfers, etc.?

[TAMU, Knower of Ball]

If comparing his preseason rankings to his final rankings isn't an accurate indicator of the predictive power of his algorithm, what is? How do you judge?

Are the predictions of AP voters not subject to the same vagaries of injuries, transfers, etc. as KenPom?

Which of the teams in KenPom's preseason top 10 that didn't live up to that billing underachieved as a result of injuries, transfers, etc.?

You judge by method. Which method uses data, which method uses feelings.

They are. Off the court factors are random and can't be predicted by any method. They could benefit one method one year and one method another. Theoretically it would balance out if analyzed over multiple seasons.

I don't know, feel free to look and get back to me.

[The Equalizer]

You judge by method. Which method uses data, which method uses feelings.

They are. Off the court factors are random and can't be predicted by any method. They could benefit one method one year and one method another. Theoretically it would balance out if analyzed over multiple seasons.

I don't know, feel free to look and get back to me.

You're using circular reasoning: saying a data-driven approach is better because it's data-driven, not because it produces better results.   

And then let's consider that you're giving credit to KenPom for being based on data, even though for the pre-season projections he doesn't have any actual performance data for the 2022-23 season.   

And if he's simply applying an algorithm on last year's stats, he's unable to incorporate qualitative observations that would cause a player to deviate from that algorithm. Justin and Ighodaro both overperformed the data-driven expectations because his freshmen stats were impacted by an injury. Kolek and Morsell outperformed data-driven projections because they were asked to take on different roles on a new team. 

Sportswriters and others incorporate those observations into their "feelings" in ways that a data-driven algorithm cannot. 

[brewcity77]

Who is finishing behind us in this scenario? Creighton, nova, Xavier or UConn ?

I get why Creighton is the favorite, but they have a lot of screaming flaws that gives me pause to consider them a top-25 team, much less the top-10 that seems to be the foolhardy consensus.

• McDermott has had a top-25 defense just twice in his 21 year career. Not sure he'll replicate that while losing his best defensive on/off performer in Hawkins.
  
• Baylor Scheierman was a beast against the no-defense Summit, but his career Tier A+B numbers (which is every night in the Big East) are terrible. He's also a really bad defender himself.
  
• McDermott's best offenses have been buoyed by 3PFG%. Last year's 30.8% from deep was the worst of his career, and they lost their two best three point shooters in Hawkins and O'Connell. Scheierman might seem like the answer, but despite his 46.2% from 3 last year, he is 26.1% for his career from deep against A+B competition and was just 23.3% last year.

They return plenty, and I understand why they're the favorite, but right now I think Creighton, UConn, Xavier, Villanova, Marquette, and maybe Providence are on pretty even footing in that there are reasons to be optimistic about competing for the league title but just as many reasons to be skeptical. Honestly, if Creighton loses in overtime to SDSU, no one is talking about them. I think Kaluma has great breakout potential, but they feel like one of those classically overhyped teams that is more like a 20-30 ranked team than someone with legitimate national championship aspirations.

[Pakuni]

You're using circular reasoning: saying a data-driven approach is better because it's data-driven, not because it produces better results.   

And then let's consider that you're giving credit to KenPom for being based on data, even though for the pre-season projections he doesn't have any actual performance data for the 2022-23 season.   

And if he's simply applying an algorithm on last year's stats, he's unable to incorporate qualitative observations that would cause a player to deviate from that algorithm. Justin and Ighodaro both overperformed the data-driven expectations because his freshmen stats were impacted by an injury. Kolek and Morsell outperformed data-driven projections because they were asked to take on different roles on a new team. 

Sportswriters and others incorporate those observations into their "feelings" in ways that a data-driven algorithm cannot.

Yup, this is exactly it, and it's why last year's KenPom prediction for MU - which we were told over and over should be the basis for our expectations - was totally worthless. Virtually everything about last year's MU team was new. New coaches. New system. New players who'd never played together. Players taking new and often more significant roles.
Why would anyone reasonably expect the previous year's data - with players on different teams in different roles with different systems and different teammates - would be an accurate indicator of what would happen at Marquette? Because, clearly Ken's algorithm missed the mark by a wide margin.

[TAMU, Knower of Ball]

Yup, this is exactly it, and it's why last year's KenPom prediction for MU - which we were told over and over should be the basis for our expectations

This is your problem here. No one said this last year or any year. This is a strawman that some like to bring out in the same flavor as "why don't we let KenPom crown the champion" and "why do we even need to play the games". KenPom is one tool of many out there that can help give an idea of what a reasonable projection for a team is. That's it. No one is saying it is gospel. Of course there are factors that can't be accurately accounted for by any algorithm. It's just meant to give an unbiased ballpark projection. If you want to make an argument why you think KenPom (or other services) is over or underestimating a specific team, great that makes sense. But insisting Marquette will do better because Marquette is my favorite team isn't a good argument.

Why would anyone reasonably expect the previous year's data - with players on different teams in different roles with different systems and different teammates - would be an accurate indicator of what would happen at Marquette? Because, clearly Ken's algorithm missed the mark by a wide margin.

Wide margin? It projected 87, we ended 56. Out of 359 teams, that's not a very wide margin. And you are right, there a lot of factors that apply to last years team that made us a lot harder to project than most teams so some variability was expected. But you may also recall that other methods, including Joey Beer Gut, were also projecting us lower than we ended up.

[brewcity77]

Yup, this is exactly it, and it's why last year's KenPom prediction for MU - which we were told over and over should be the basis for our expectations - was totally worthless. Virtually everything about last year's MU team was new. New coaches. New system. New players who'd never played together. Players taking new and often more significant roles.
Why would anyone reasonably expect the previous year's data - with players on different teams in different roles with different systems and different teammates - would be an accurate indicator of what would happen at Marquette? Because, clearly Ken's algorithm missed the mark by a wide margin.

I'm a bit confused by this. Who told us over and over that kenpom should be the basis for expectations last year?

Also, is it any surprise an algorithm would perform worse when there's mass turnover? I get that the AP Poll to start the year is typically a pretty good indicator, but that's likely largely because the teams that are good tend to continue to be good.

I guess my question to the last bit would be "who got us right?" Pretty much every major pundit I recall had Marquette somewhere in the 9-11 range in the league. If Pomeroy's model deserves criticism for having us there, so be it, but where is the better indicator that accurately pegged us 5-6?

The prediction business is tricky. If someone wants to do a comprehensive review of how kenpom, Sagarin, T-Rank, 3MW, BPI, EvanMiya, Conference Coaches, and all the other people who try to prognosticate in October do when the first Monday in April concludes, I'm sure we'd all love to see it. I'm equally sure that whatever model ended up doing the best would have plenty of years when it did poorly as well.

I think the advantage to a model like kenpom or T-Rank, or a more thorough review like 3MW or the conference preseason rankings aggregated is that they look at everyone. The AP Poll gives you 25 teams. That's it. And no, the "others receiving votes" don't count as 26-40 (or whatever) because the poll can only measure to 25. When you have 350+ teams, that's pretty useless. Even when you are just trying to peg 68 NCAA teams it's not very helpful. And while the data out will only be as good as the data in and can't account for anything, I definitely think a lot more goes into the data of 350+ teams that a model like that spits out than the pollsters that have a vague working knowledge in the preseason of maybe 50-75 teams. Because if you're not putting as much effort into 76-350+, you really aren't doing much at all.

[MU82]

You judge by method. Which method uses data, which method uses feelings.

So it's your contention that AP poll voters use "feelings" rather than data? They just get up in the morning and say, "Hmmm, I think I'll put Kentucky No. 7 today based on nothing but my feelings"?

I am NOT saying the voters are infallible or wonderful or even better than KP. I am saying that most of them probably use KP (and other data services) in helping them decide how to vote and that they don't just slap together a top-25 based on feelings. Do their own personal observations influence their picks? Certainly yes for some (or even most), and I happen to look at that as a plus; but I'll admit that I'm not neutral on the subject.

Of course, rankings in August or October or even December don't really "matter" at all. They're just for conversations like the one we're having.

[TAMU, Knower of Ball]

So it's your contention that AP poll voters use "feelings" rather than data? They just get up in the morning and say, "Hmmm, I think I'll put Kentucky No. 7 today based on nothing but my feelings"?

I am NOT saying the voters are infallible or wonderful or even better than KP. I am saying that most of them probably use KP (and other data services) in helping them decide how to vote and that they don't just slap together a top-25 based on feelings. Do their own personal observations influence their picks? Certainly yes for some (or even most), and I happen to look at that as a plus; but I'll admit that I'm not neutral on the subject.

Of course, rankings in August or October or even December don't really "matter" at all. They're just for conversations like the one we're having.

That's fair, I shouldn't have said that. AP writers use a variety of methods ranging from detailed study of every team to feelings depending on the contributor. Brew's point is a good one though, projecting the top 25 is easy compared to projecting #s 100-320

[Pakuni]

But insisting Marquette will do better because Marquette is my favorite team isn't a good argument.

What's that you were saying about straw men?
I don't wish to review a year's worth of posts on Scoop, but if you don't recall how many people  pointed to the KenPom preseason rankings during debates about expectations last season, then I'll just say you and I remember things differently.

Wide margin? It projected 87, we ended 56. Out of 359 teams, that's not a very wide margin. And you are right, there a lot of factors that apply to last years team that made us a lot harder to project than most teams so some variability was expected. But you may also recall that other methods, including Joey Beer Gut, were also projecting us lower than we ended up.

Yeah, a wide margin because 200+ of those 359 teams don't count. I'm not giving kudos to Ken or anyone else for accurately predicting that MU would have a better team than McNeese State and Houston Baptist. A 31-spot swing amongst its peer group - which is significantly smaller than 359 - is a large margin.

[lawdog77]

It is because it is based on unbiased data i
.

My only comment to this exercise is that the "data" may be unbiased, but the weight he gives to each data point is subjective .

[TAMU, Knower of Ball]

What's that you were saying about straw men?
I don't wish to review a year's worth of posts on Scoop, but if you don't recall how many people  pointed to the KenPom preseason rankings during debates about expectations last season, then I'll just say you and I remember things differently.

Pointing to KenPom during a conversation is not the same thing as telling someone that should be the baseline for your expectations. Again it is one tool of many.

And I'm not building a straw man. There were plenty of posters who insisted that it doesn't matter who's on the roster, their expectations for Marquette don't change, they just expect Marquette to do well. There were more who didn't give any explanation for their expectations just insisted that Marquette was going to do well.

Yeah, a wide margin because 200+ of those 359 teams don't count. I'm not giving kudos to Ken or anyone else for accurately predicting that MU would have a better team than McNeese State and Houston Baptist. A 31-spot swing amongst its peer group - which is significantly smaller than 359 - is a large margin.

Of course they count. KP's algorithm doesn't give you a bonus for being a high major or a perceived better team. It is just based on the numbers and all teams have an equal opportunity to put up numbers that will lead to higher rankings. And again, that's the biggest difference between the AP Poll and something like KP. Picking out the top 25 teams is relatively easy because there are a few programs that are on the end of the bell curve and obviously stand out from the rest. Projecting the middle of the bell curve is much more difficult.

[wadesworld]

Wide margin? It projected 87, we ended 56. Out of 359 teams, that's not a very wide margin. And you are right, there a lot of factors that apply to last years team that made us a lot harder to project than most teams so some variability was expected. But you may also recall that other methods, including Joey Beer Gut, were also projecting us lower than we ended up.

I mean, the absolute worst power conference team finished 219 last year.  I can just blindly guess that in 2025 Marquette will finish somewhere between 20th and 85th in KenPom.  The realistic range is much, much smaller than 359 spots for Marquette.  Given that, I'd say off by 31 spots is a pretty wide margin.

KenPom is fine.  The problem with it is that, while the backers here are claiming nobody says it's Gospel, there are A LOT of people who treat it like Gospel.  It's a nice tool.  But it's a flawed tool.  It had Wisconsin way overranked for the better part of a decade.  It has had Gonzaga in the top 2 in 5 of the past 6 seasons.  They've been that in one or two seasons.  Certainly not 5 of the past 6.

[brewcity77]

My only comment to this exercise is that the "data" may be unbiased, but the weight he gives to each data point is subjective .

Sure, but those data point weightings are equal across the 350+ teams, so I don't see how that would create a bias.

[TAMU, Knower of Ball]

And again, that's the biggest difference between the AP Poll and something like KP. Picking out the top 25 teams is relatively easy because there are a few programs that are on the end of the bell curve and obviously stand out from the rest. Projecting the middle of the bell curve is much more difficult.

To illustrate this point, the difference between #56 Marquette and the team that actually finished #87 (Northwestern) was 3.68 Adj eDM (the actual stat that KP calculates and then ranks). That is smaller than the difference between #1 Gonzaga and #4 Baylor (3.91).

[wadesworld]

To illustrate this point, the difference between #56 Marquette and the team that actually finished #87 (Northwestern) was 3.68 Adj eDM (the actual stat that KP calculates and then ranks). That is smaller than the difference between #1 Gonzaga and #4 Baylor (3.91).

Which goes to show how flawed it is.  Gonzaga was...not that good this year, relative to teams Few has had recently.

[brewcity77]

It has had Gonzaga in the top 2 in 5 of the past 6 seasons.  They've been that in one or two seasons.  Certainly not 5 of the past 6.

Hmm...so I assume you are just as biased against the AP Poll? It has had Gonzaga in the top 2 in 4 of the past 6 seasons. In one of the two outliers, they were 4th.

Gonzaga was absolutely top-2 in 2017, 2020, and 2021. Both the vast preponderance of polls and NCAA results bear that out. I don't think there's any argument against that.

In 2022, they started the year #1, fell when they dropped their first game, then climbed back to #2 in Week 10 and remained in the top-2 the rest of the season. So while I guess you can dispute 2022 because they lost in the Sweet 16, the AP Poll consensus was that they were one of the top two teams in the country for the vast majority of the season.

2019 is the other year you can dispute. Their only losses before falling to St. Mary's in the WCC Final were to AP Final Poll #3 UNC and #6 Tennessee. They beat the Duke team that finished #1. So they were maybe third or fourth, with UVA and UNC the other contenders for top level status.

Honestly, saying Gonzaga has been #1 or #2 in "one or two seasons" is FAR more hyperbolic than saying they were #1 in five of the last six. I'd say it's pretty clearly at least four of six with a decent case for five.

[lawdog77]

Sure, but those data point weightings are equal across the 350+ teams, so I don't see how that would create a bias.

Depends on what data points he weighs heavier.

[panda]

I get why Creighton is the favorite, but they have a lot of screaming flaws that gives me pause to consider them a top-25 team, much less the top-10 that seems to be the foolhardy consensus.

• McDermott has had a top-25 defense just twice in his 21 year career. Not sure he'll replicate that while losing his best defensive on/off performer in Hawkins.
  
• Baylor Scheierman was a beast against the no-defense Summit, but his career Tier A+B numbers (which is every night in the Big East) are terrible. He's also a really bad defender himself.
  
• McDermott's best offenses have been buoyed by 3PFG%. Last year's 30.8% from deep was the worst of his career, and they lost their two best three point shooters in Hawkins and O'Connell. Scheierman might seem like the answer, but despite his 46.2% from 3 last year, he is 26.1% for his career from deep against A+B competition and was just 23.3% last year.

They return plenty, and I understand why they're the favorite, but right now I think Creighton, UConn, Xavier, Villanova, Marquette, and maybe Providence are on pretty even footing in that there are reasons to be optimistic about competing for the league title but just as many reasons to be skeptical. Honestly, if Creighton loses in overtime to SDSU, no one is talking about them. I think Kaluma has great breakout potential, but they feel like one of those classically overhyped teams that is more like a 20-30 ranked team than someone with legitimate national championship aspirations.

Totally agree on Creighton. Schierman is the lovable low major tourney success type player that garners lots of pre season hype, but certainly isn't a lock to play at the same level.

Xavier is a more even money pick for me to win the league. They played so well in the NIT post Steele and Sean miller is a massive upgrade on the sidelines. It will be interesting to see if he can figure out a way (or wants) to play Nunge and Freemantle together for long stretches, look out. Souley Boum is an interesting piece. Admittedly only watched him play twice last year, but he's a bucket. He'll be a very dangerous piece if he can accept being third fiddle. He's never met a shot he didn't like so the situation could definitely go either way.