So....What are we ranked on Monday - 2/24/2020

[Newsdreams]

Season's over. We've already got a bad win. Maybe the new metric will be kinder to us for scheduling so poorly.

FIRE WOJO!

[Its DJOver]

I understand how it works but couldn't give you the exact formulas different systems use off the top of my head. You are welcome to do the research and find out.

And while it is a range, the difference between Haani and Markus is not as different as you might think. Haani had a terrible freshman year with a statline that was buoyed by playing an atrocious non-conference schedule with subpar teammates. So when he became a sophomore his numbers got worse because he played a tougher schedule and had better teammates around him taking opportunities away from him to bolster his stats. Markus had an elite freshman year with a statline that was limited by a deep team and strong schedule. So when he became a sophomore and had worse talent around him and played an easier schedule his stat line grew. Also keep in mind that we're talking development, meaning what % a player gets better by. Markus improved a lot more than Haanie, but the % he got better between freshman and sophomore year was closer to what Haanie was (but still better IMHO).

Here's yet another variable that makes this impossible to accurately predict; level of competition.  Other than his career high 24 against Jackson State, every time Haani broke the 20 point marker his Freshman year was against high majors.  Sophomore year he only broke that barrier twice against low majors, and also combined to score 9 points the last 6 games.  Sam never broke the 20 point barrier his Freshman year, although he did have 19 twice.  His Sophomore year, he broke it 8 times including a pair of 30 point performances.  So when the schedule got tougher for Haani, the level of production dropped off, when it got tougher for Sam, production picked up.  I don't care what the efficiency rating are, any metric that sees going from four 20 point performances to 2, as equal to going from 0 to 8, with two 30 point performances, is majorly flawed.

[Its DJOver]

As for your previous point of all the different predictive models using different criteria, that doesn't make any of them "wrong." You just have to understand what each model is measuring so you properly understand the results. There are always going to be outliers on both ends but that doesn't mean that most don't fall in the center of the bell curve.

That's the thing though, we know that BPI has us too high because of experience, we know that KenPom will have UW too high because of pace.  We don't know how they view missed time for injury, degree of that injury, potential loss of production due to playing through an injury, loss of production due to move up in competition.  I'm not going to blindly trust something that doesn't have an explanation.  BPI, KenPom have one, "normal" player development does not. 

[TAMU Eagle]

Here's yet another variable that makes this impossible to accurately predict; level of competition.  Other than his career high 24 against Jackson State, every time Haani broke the 20 point marker his Freshman year was against high majors.  Sophomore year he only broke that barrier twice against low majors, and also combined to score 9 points the last 6 games.  Sam never broke the 20 point barrier his Freshman year, although he did have 19 twice.  His Sophomore year, he broke it 8 times including a pair of 30 point performances.  So when the schedule got tougher for Haani, the level of production dropped off, when it got tougher for Sam, production picked up.  I don't care what the efficiency rating are, any metric that sees going from four 20 point performances to 2, as equal to going from 0 to 8, with two 30 point performances, is majorly flawed.

You are way too focused on points and using odd benchmarks. What's the significant of 20 points in a game? Also, your 20 points a game stat needs context. Haanie reached 20 points three times against high majors as a freshman. One was a very good game against a good Xavier team. One was against DePaul and not the recently rejuvinated DePaul, bad DePaul. One was against Providence and required 2 overtimes and 16 shots to do it.

Also, it is very easy to capture level of competition.

That's the thing though, we know that BPI has us too high because of experience, we know that KenPom will have UW too high because of pace.  We don't know how they view missed time for injury, degree of that injury, potential loss of production due to playing through an injury, loss of production due to move up in competition.  I'm not going to blindly trust something that doesn't have an explanation.  BPI, KenPom have one, "normal" player development does not. 

I'm trying not to read it this way, but this sounds like you are saying you don't understand how these models work and because you don't understand them you don't trust them. If you are trying to say something else please correct me. A lot of these you can find the nuts and bolts for you to research and learn more about though some are behind a paywall.

If your argument is that predictive models like KenPom are not 100% accurate than you have no argument from me. They are nothing more than very educated guesses. Still have to play the games.

[Its DJOver]

No insult perceived as I'm sure no insult was implied (actually think this is a really good discussion).  I'll try to explain my reasoning differently.

While I don't completely understand them, I know that number gurus like Kenpom use hard data.  Offensive efficiency and defensive efficiency are calculated based on data that comes off the court.  They plug all their data in, and out pops a number, that compared to other teams numbers create a ranking.  I know that there a lot more nuts and bolts to it than that, but it all comes from numbers pulled off of the court, which are available to everyone.

If you were try to calculate "normal" player development, you would need a lot of data that doesn't come off the court.  How much was Sam's hip bothering him at the end of the year?  You'd have to ask Sam.  When did he return to 100%?  Again, talk to Sam.  Has pain permanently effected Greg's shot (hope not)?  How much off-season work was Ike able to do with his bad back?  You'd have to be in the weight room.  How do you compare Sam's hip injury to Greg's thumb injury, to Ike back injury?  You would need to be within the program to get an answer other than "coach speak". 

I have no doubt that statisticians, and mathematicians can come up with a formula, that I certainly wound't understand, to "attempt" to account for all of this.  Although this would involve a lot of guesswork for a few reasons.  Every injury is different.  Even the same injury to the same player.  You also wouldn't have the raw data that you can't get from a box score, you'd have to get it straight from the horses mouth, which doesn't happen.  The more guesswork that goes into the formula, the less accurate that the formula will end up being.  Kenpom doesn't use much (any?) guessing, he just plugs and chugs.  Any player development calculation would have to use exponentially more guesswork, which would lead to inaccuracies.  That's why I would be extremely hesitant to trust any player development model.

[TAMU Eagle]

As far as I am aware, no model considers how an injury will impact a player, they look retroactively at how an injury did impact a player. And development projections all come from hard data that comes from the court. I'm not aware of any "off the court" data that gets used. For example, all the data I alluded to in a past post about Frank Kaminsky being a breakout candidate all came from his performance on the court. Now he did exceed expectations, but all the data that pegged him as a breakout candidate was all readily available. There is no secret sauce, just different formulas.

[Its DJOver]

You would still need an absurd amount of off-court data to get anything close to an accurate prediction.  Kaminsky's shot total from his entire Sophomore season was less than 100, yet I would bet he took at least 100 shots per day in the off-season before his breakout Junior year.  Where were they from, what was his clip, what coach was he working with.  100 shots with Nelson (he's the shooting guy right?) isn't equal to 100 shots with Stan.  How many hours of film did he watch?  You won't be able to get that data, and without it, any model would be far less accurate, IMO of course.

I'm sure people are smart enough to come up with formulas, but they would not have the data for an accurate model.

[TAMU Eagle]

You would still need an absurd amount of off-court data to get anything close to an accurate prediction.  Kaminsky's shot total from his entire Sophomore season was less than 100, yet I would bet he took at least 100 shots per day in the off-season before his breakout Junior year.  Where were they from, what was his clip, what coach was he working with.  100 shots with Nelson (he's the shooting guy right?) isn't equal to 100 shots with Stan.  How many hours of film did he watch?  You won't be able to get that data, and without it, any model would be far less accurate, IMO of course.

I'm sure people are smart enough to come up with formulas, but they would not have the data for an accurate model.

Not really no. Again, these systems have already proved their accuracy by...well...being accurate. While some programs are better at developing talent than others the differences between programs are usually negligible. Some models do account for this but that's based on how players in that program have developed in the past....which is based on on the court numbers.

I'll repeat again, you can claim these systems aren't accurate. But the major predictive ones have shown year after year that they are. They aren't pinpoint accurate, but then tend to put teams into the right ballparks.

[Its DJOver]

I guess we'll just agree to disagree.  I just can't say that a year-to-year projection can be accurate when it ignores April through October. 

[TAMU Eagle]

I guess we'll just agree to disagree.  I just can't say that a year-to-year projection can be accurate when it ignores April through October.

It doesn't. They have gathered data from years and years to show how much a typical player will improve based on a variety of different factors from April to October. That's why predictive models love Wisconsin this season. They don't add much in terms of freshmen/transfers but should add a ton via offseason improvement from all their returning players.

Again, you say it can't be accurate....yet the major ones have been...so are they just getting lucky?

[Its DJOver]

It doesn't. They have gathered data from years and years to show how much a typical player will improve based on a variety of different factors from April to October. That's why predictive models love Wisconsin this season. They don't add much in terms of freshmen/transfers but should add a ton via offseason improvement from all their returning players.

Again, you say it can't be accurate....yet the major ones have been...so are they just getting lucky?

So they know that Illikainen is the new Evan Anderson? 

This is another flaw in this system.  I'm sure that they predict that Iverson will be an improved 3 point shooter, and he could hit one against Eastwestern Armpit State, and their model will be "proved" correct because he technically improved his three point shooting percentage, even if he never hits another one all year.

Bolded sounds like you saying they're proved correct because they say they've been proved correct.  If this is the same model that said Haani's Freshman to Sophomore growth is in the same category as Sam's and that "proves them correct", then all they have to do to be proved correct is make a blanket statement that every returning player will get at least a little better.  Are they wrong?  No, but it doesn't tell us anything.

[TAMU Eagle]

These models don't know anything. They are predictive. These are just very educated guesses.

You are looking very micro. Look macro. It can't tell you with much accuracy that this specific player will improve in this specific way (though there are often clues in the data that could help make these type of predictions). These models are meant to give a Birdseye view of the entire landscape of college basketball. The more data you add, the more accurate it becomes.

So no,  you shouldn't use these models to say Khalil Inverson will develop a three point shot. You should use it to say Wisconsin will be a top 50 team this season.

And no I'm not saying they have been proven correct because they say they have been proven correct. You can go back and look at their preseason projections and compare them to their postseason results in past years and you will find they have been accurate. Not pinpoint accurate,  not perfect,  but most of their projections end up forming a nice and perty bell curve

[Its DJOver]

You're looking too macro by saying that UW was good in the past, so they'll be good in the future.  Look more micro, they have 3 major players coming off major Injuries and one that is already out indefinitely.  Just because Nakovil developed does not guarantee that Reuvers will. Just because Leuer developed does not guarantee that Thomas will.  4 injury question mark's, and 2 scholarship players that were behind the walk ons last year.  Not a great start. Could the be top 50? Sure, but they will need players to overachieved their "proven predictive model".

[TAMU Eagle]

You're looking too macro by saying that UW was good in the past, so they'll be good in the future.  Look more micro, they have 3 major players coming off major Injuries and one that is already out indefinitely.  Just because Nakovil developed does not guarantee that Reuvers will. Just because Leuer developed does not guarantee that Thomas will.  4 injury question mark's, and 2 scholarship players that were behind the walk ons last year.  Not a great start. Could the be top 50? Sure, but they will need players to overachieved their "proven predictive model".

Nakovil, Leuer, etc have nothing to do with this conversation. No one is saying because these players developed these players will. The formulas are based on data from every team across multiple seasons. Not one team. Wisconsin doesn't get a bonus compared to other teams when it comes to player development. How Wisconsin has done in seasons prior to 17-18 is irrelevant to this discussion.

You make some pretty out there assumptions saying that they will need players to overachieve in order to make top 50. If every one of their players develop at a normal pace, they will be a top 50 team easily. Without breaking a sweat.  Now, that doesn't mean all of their players will. Some will exceed, some will fall short. They could be a weird team where everyone exceeds or everyone falls short and they would become an outlier. The most sound assumption is that as a team they will develop at a normal pace.

You use words like guarantee a lot. If you are looking for a guarantee, you are not going to find it anywhere. These are just projections. They are not going to be guarantee level accurate. The goal is results on a bell curve. Most results right in the middle, with some outliers on both ends.

Maybe Wisconsin ends up being one of those outliers this season. Maybe even for the reasons you list. Certainly possible. I doubt the injury to Aleem Ford is included in most models given that we don't know the severity yet. But those listing them as a top 50 team having statistically sound evidence to back up their projection. There is no Wisconsin bias. This isn't "the Badgers are usually good so I'm ranking them higher". It's simply the fact that they were a top 75ish team last season that returns everyone, gets 2 key players back from injury, and will go from being one of the least experienced teams in Division 1 to a team in the middle of the pack for experience. Projecting a 25 point jump in the rankings is completely reasonable for a team with that profile, especially when you consider that virtually every team in front of them lost a lot more production than they did.

[Its DJOver]

Nakovil, Leuer, etc have nothing to do with this conversation. No one is saying because these players developed these players will. The formulas are based on data from every team across multiple seasons. Not one team. Wisconsin doesn't get a bonus compared to other teams when it comes to player development. How Wisconsin has done in seasons prior to 17-18 is irrelevant to this discussion.

You make some pretty out there assumptions saying that they will need players to overachieve in order to make top 50. If every one of their players develop at a normal pace, they will be a top 50 team easily. Without breaking a sweat.  Now, that doesn't mean all of their players will. Some will exceed, some will fall short. They could be a weird team where everyone exceeds or everyone falls short and they would become an outlier. The most sound assumption is that as a team they will develop at a normal pace.

You use words like guarantee a lot. If you are looking for a guarantee, you are not going to find it anywhere. These are just projections. They are not going to be guarantee level accurate. The goal is results on a bell curve. Most results right in the middle, with some outliers on both ends.

Maybe Wisconsin ends up being one of those outliers this season. Maybe even for the reasons you list. Certainly possible. I doubt the injury to Aleem Ford is included in most models given that we don't know the severity yet. But those listing them as a top 50 team having statistically sound evidence to back up their projection. There is no Wisconsin bias. This isn't "the Badgers are usually good so I'm ranking them higher". It's simply the fact that they were a top 75ish team last season that returns everyone, gets 2 key players back from injury, and will go from being one of the least experienced teams in Division 1 to a team in the middle of the pack for experience. Projecting a 25 point jump in the rankings is completely reasonable for a team with that profile, especially when you consider that virtually every team in front of them lost a lot more production than they did.

Man we're just going in circle, because I'm about to say that this isn't a real thing.  You'll say that there are formulas.  I won't think the formulas will have enough data.  You'll say that they were right in the past, so they'll be right in the future.  I'll counter with the "accuracy" of those correct predictions isn't all that accurate.  Then you'll justify a player regressing in usage, eFG%, and scoring, by saying that our schedule got tougher.  The only thing I'll say about UW is that IMO they have a lot of question marks, and if a majority of those question marks don't turn out positive, this season could have similarities to last.

[TAMU Eagle]

Man we're just going in circle, because I'm about to say that this isn't a real thing.  You'll say that there are formulas.  I won't think the formulas will have enough data.  You'll say that they were right in the past, so they'll be right in the future.  I'll counter with the "accuracy" of those correct predictions isn't all that accurate.  Then you'll justify a player regressing in usage, eFG%, and scoring, by saying that our schedule got tougher.  The only thing I'll say about UW is that IMO they have a lot of question marks, and if a majority of those question marks don't turn out positive, this season could have similarities to last.

You have the order of things mixed up in there but I digress.

You say there is no such thing. Of course there is. This is an oversimplification but you take all the data from one season, compare it to all the data from the following season, calculate how much each individual player improved, then average all those together and ipso facto you have your normal pace of development for that season. You add the same data from other seasons and you will start being able to project how players are going to develop. Not a specific player, but on average how much you can expect a player to develop. Some will be over the average, some will be under. Some will be WAY over and some will be WAY under. In the end, it will form a pretty neat bell curve.

Let me repeat that this is a gross oversimplification. I'm not sure how you can say there isn't a normal pace of development. Unless your arguing that most players either are way over average or way below average. Or that it will vary wildly from year to year. One year a ton of people will above average, the next year a ton will be below average.

[Its DJOver]

Weird that "normal" and "average" are now the same word.  Statistically speaking, yes, of course there is an average.  In this case, however, "normal" would be closely associated with "mode", or most frequent, and since you would have a very large data pool, it would be safe to assume that you could do a little rounding, to calculate mode.

[Jay Bee]

Here's yet another variable that makes this impossible to accurately predict; level of competition.  Other than his career high 24 against Jackson State, every time Haani broke the 20 point marker his Freshman year was against high majors.  So when the schedule got tougher for Haani, the level of production dropped off, when it got tougher for Sam, production picked up.  I don't care what the efficiency rating are, any metric that sees going from four 20 point performances to 2, as equal to going from 0 to 8, with two 30 point performances, is majorly flawed.

No. Your metrics are ridiculous. Do we need to get you on a 72 hour hold?

Haani as a frosh against KenPom Tier A&B: 86.5 ORtg, 9.9% DR%
Haanif as a soph against KenPom Tier A&B: 93.0 ORtg, 14.0% DR%

[Its DJOver]

No. Your metrics are ridiculous. Do we need to get you on a 72 hour hold?

Haani as a frosh against KenPom Tier A&B: 86.5 ORtg, 9.9% DR%
Haanif as a soph against KenPom Tier A&B: 93.0 ORtg, 14.0% DR%

I thought eFG% ruled all???

[Jay Bee]

I thought eFG% ruled all???

It does when it comes to a team winning and losing basketball games. u no undastand

#FTsNoMatta

[Its DJOver]

It does when it comes to a team winning and losing basketball games. u no undastand

#FTsNoMatta

I prefer players that help us win basketball games.  Players with higher eFG% do that better than those with lower eFG%.

[Jay Bee]

I prefer players that help us win basketball games.  Players with higher eFG% do that better than those with lower eFG%.

False!

[jesmu84]

Nm. Wrong thread

[Its DJOver]

False!

So then eFG% mo matta??? You're contradicting yourself here bud. 

[Jay Bee]

So then eFG% mo matta??? You're contradicting yourself here bud.

Not at all. You're not understanding the difference between individual stats and team stats.