Is Ennis the bet frosh in college basketball?

[Cusefannotindc]

Isn't not perfect another way of you admitting that it is not so accurate?

No, and that is just a stupid statement.

 

[jordoo]

No, and that is just a stupid statement.

Why my argument (one that you perused by the way not me I simply made a statement about my opinion on WAR in bball) is that you are using a stat that isn't really quantifiable to begin with, so saying it isn't perfect strengthens my argument at its base.

 

[Cusefannotindc]

I agree they can be useful. I like player efficiency ratings but anyone who tries to say they are a good predictor of wins and losses is absurd, because there are so many factors and so many other players to be considered. And, again, most of the things that factor into how effective a player is cannot be measured by any stat. Playing good positional defense doesn't appear on a box score, for instance. Making a smart pass that doesn't lead to a basket but leads to a pass that leads to a basket never shows up. And so forth.

Teams that make smart passes will typically continue to make smart passes and it will be reflected in their Offensive Efficiency rates.

 

[jekelish]

Teams that make smart passes will typically continue to make smart passes and it will be reflected in their Offensive Efficiency rates.

Now you're talking about TEAMS, and not individual players. And smart passes are not measured aside from not turning it over, and getting assists. You can have plenty of smart passes and a guy decides to take it to the hoop and no one gets credit for the assist. That's not measured anywhere, because there's no assist, and there's no turnover.

Again, just to reiterate, the vast majority of things that happen on a basketball court have no unit of measurement. They can't really be factored in. Tell me how you quantify good defensive positioning or a well set screen? And again, this is individual player stuff, since that's what you began this whole thing talking about.

 

[Orangezoo]

The stronger point about WAR here is not that it is imperfect but that it is not some kind of near miss prediction model that has no equal. Its just another hobbyist's way to fool around with with statistics to gain a small advantage. Therefore its limited and over valued in the argument for it in the thread.

 

[Cusefannotindc]

Why my argument (one that you perused by the way not me I simply made a statement about my opinion on WAR in bball) is that you are using a stat that isn't really quantifiable to begin with, so saying it isn't perfect strengthens my argument at its base.

First off, it is Wins Produced, which measures something different than WAR does in baseball. Wins can be quantified, in many ways. In no way does saying a particular model is 100% accurate bolster your argument. If you can successfully pick the winner of 65% of games using your eyes, and I can predict 75% using a model, is my model not better because it doesn't predict 100%?

 

[Orangezoo]

Teams that make smart passes will typically continue to make smart passes and it will be reflected in their Offensive Efficiency rates.

Uggh... there is a reason we separate qualitative from quantitative.

 

[Knicks411]

The accuracy comments are a little puzzling to me.

No stat is perfect. Shooting % would have told you that Trevor Cooney is a 45% 3 point shooter or whatever it was before yesterday. Since he shot 2-11 or whatever it was instead, does that mean shooting% is an inaccurate stat? If you expect any stat to be 100% accurate, then you should be throwing everything out.

Again, just to reiterate, the vast majority of things that happen on a basketball court have no unit of measurement. They can't really be factored in. Tell me how you quantify good defensive positioning or a well set screen? And again, this is individual player stuff, since that's what you began this whole thing talking about.

Yeah, I wanted to comment on this as well, because it seems like we have some people talking about using some of these stats as predictors. I dont think anyone is using individual player stats to predict the outcomes of games, it's more on a team level. (Though like i said, we seem to be going all over the place so I don't know)

On a team level, I find the efficiency based stats to be very useful and pretty predictive. 100% predictive? Of course not, nothing is. On an individual level, there is a ton of noise in all the data and you need to be aware of all of the things that don't show up in the stats when you are evaluating guys. And for the record, a lot of the time, for guys like us who are just fans and have lives going on outside of this, that's kind of impossible. But you look at some of the stuff the NBA can track with synergy and the sportvu cameras, and it's pretty amazing

 

[Orangezoo]

First off, it is Wins Produced, which measures something different than WAR does in baseball. Wins can be quantified, in many ways. In no way does saying a particular model is 100% accurate bolster your argument. If you can successfully pick the winner of 65% of games using your eyes, and I can predict 75% using a model, is my model not better because it doesn't predict 100%?

So you are saying the delta is 10 pct? BS.

 

[jordoo]

First off, it is Wins Produced, which measures something different than WAR does in baseball. Wins can be quantified, in many ways. In no way does saying a particular model is 100% accurate bolster your argument. If you can successfully pick the winner of 65% of games using your eyes, and I can predict 75% using a model, is my model not better because it doesn't predict 100%?

Lets start where players do not produce wins in bball. We are not talking about picking the winners were talking about how Wins Produced (if that's what you are calling it) is basically a BS stat for college basketball because its not produced in a way that makes it an actual stat. I think you have posted enough in this thread that I get you and we don't need to go any further. People can feel that something you think is very cool is silly and it doesn't make them any less intelligent than you.

 

[jekelish]

The accuracy comments are a little puzzling to me.

No stat is perfect. Shooting % would have told you that Trevor Cooney is a 45% 3 point shooter or whatever it was before yesterday. Since he shot 2-11 or whatever it was instead, does that mean shooting% is an inaccurate stat? If you expect any stat to be 100% accurate, then you should be throwing everything out.



Yeah, I wanted to comment on this as well, because it seems like we have some people talking about using some of these stats as predictors. I dont think anyone is using individual player stats to predict the outcomes of games, it's more on a team level. (Though like i said, we seem to be going all over the place so I don't know)

On a team level, I find the efficiency based stats to be very useful and pretty predictive. 100% predictive? Of course not, nothing is. On an individual level, there is a ton of noise in all the data and you need to be aware of all of the things that don't show up in the stats when you are evaluating guys. And for the record, a lot of the time, for guys like us who are just fans and have lives going on outside of this, that's kind of impossible. But you look at some of the stuff the NBA can track with synergy and the sportvu cameras, and it's pretty amazing

Team efficiency? Yes. I think that's a damn good indicator of whether a team is going to win when you compare it to the opponent's efficiency numbers...as a team. But trying to say that you can use INDIVIDUAL efficiency numbers to predict whether a team will win, in a true team sport like basketball, seems absurd to me.

And let's remember, that's where this whole thing started, when it was argued that Ennis is the best freshman based on his individual "Wins Produced" numbers. Individual efficiency versus team efficiency are so different it's just ridiculous and not worth even talking about the two together in a real team sport.

 

[Cusefannotindc]

Now you're talking about TEAMS, and not individual players. And smart passes are not measured aside from not turning it over, and getting assists. You can have plenty of smart passes and a guy decides to take it to the hoop and no one gets credit for the assist. That's not measured anywhere, because there's no assist, and there's no turnover.

Again, just to reiterate, the vast majority of things that happen on a basketball court have no unit of measurement. They can't really be factored in. Tell me how you quantify good defensive positioning or a well set screen? And again, this is individual player stuff, since that's what you began this whole thing talking about.

Look I have never argued that either of these are exclusive of one another, but saying that modeling doesn't matter at all is just wrong.

To the Freshmen argument, on eye test alone it is very difficult to tell who is the best player, they play different positions, on different teams with different styles. But a rebound is a rebound, 2 points is two points, etc. looking at the box score and creating a measure of all these gives another way to determine who is "the best".

 

[Orangezoo]

The accuracy comments are a little puzzling to me.

No stat is perfect. Shooting % would have told you that Trevor Cooney is a 45% 3 point shooter or whatever it was before yesterday. Since he shot 2-11 or whatever it was instead, does that mean shooting% is an inaccurate stat? If you expect any stat to be 100% accurate, then you should be throwing everything out.



Yeah, I wanted to comment on this as well, because it seems like we have some people talking about using some of these stats as predictors. I dont think anyone is using individual player stats to predict the outcomes of games, it's more on a team level. (Though like i said, we seem to be going all over the place so I don't know)

On a team level, I find the efficiency based stats to be very useful and pretty predictive. 100% predictive? Of course not, nothing is. On an individual level, there is a ton of noise in all the data and you need to be aware of all of the things that don't show up in the stats when you are evaluating guys. And for the record, a lot of the time, for guys like us who are just fans and have lives going on outside of this, that's kind of impossible. But you look at some of the stuff the NBA can track with synergy and the sportvu cameras, and it's pretty amazing

I think its a bit deceiving. These models are as simple as looking at a few numbers of a few teams. We can all do a little homework without scoring efficiency to get greater insight. The problem with these more complex models is that matchups are immeasurable due to qualitative factors. Also no one bets 100 pct on a model. They might use it but its too inexact to be verbatim.

 

[Knicks411]

Team efficiency? Yes. I think that's a damn good indicator of whether a team is going to win when you compare it to the opponent's efficiency numbers...as a team. But trying to say that you can use INDIVIDUAL efficiency numbers to predict whether a team will win, in a true team sport like basketball, seems absurd to me.

And let's remember, that's where this whole thing started, when it was argued that Ennis is the best freshman based on his individual "Wins Produced" numbers. Individual efficiency versus team efficiency are so different it's just ridiculous and not worth even talking about the two together in a real team sport.

Yeah agreed, that's what I meant; on the team level, not as much on the individual level.

 

[jekelish]

Look I have never argued that either of these are exclusive of one another, but saying that modeling doesn't matter at all is just wrong.

To the Freshmen argument, on eye test alone it is very difficult to tell who is the best player, they play different positions, on different teams with different styles. But a rebound is a rebound, 2 points is two points, etc. looking at the box score and creating a measure of all these gives another way to determine who is "the best".

But that's the thing, a box score doesn't tell the whole story. Anyone can look at a box score and take a wild guess about who the best player was. Say someone scores 20 points. What if eight of those points came in garbage time with his team down 30 and the opponent stopped playing defense, allowing him four easy layups? His FG% goes up and so does his point total and overall offensive efficiency. Box scores don't tell the whole story.

That's why you need to watch the whole game and why the eye test is more important. Numbers are far from everything. That's why I think these sort of things are a secondary measurement at best, in determining player value.

 

[Knicks411]

The problem with these more complex models is that matchups are immeasurable due to qualitative factors.

I agree with that, to an extent, but I dont think that is a problem with the model; that's a problem with predicting the results of matchups that have any number of factors that are difficult and/or impossible to measure. It would be the same problem with one of us predicting the outcome of a game.

That's why you need to watch the whole game and why the eye test is more important. Numbers are far from everything. That's why I think these sort of things are a secondary measurement at best, in determining player value.

The eye test is certainly important, but part of the problem I have with it is that it's really hard, especially in college hoops. There are 351 teams, they all play roughly 30 games a year. No one (Marsh, maybe?) is going to be able to watch any significant % of those games.

So if an SU fan is gonna say the eye test tells than Ennis is a better player thn Aaron Gordon or Julius Ranfle or whoever (which he very well might be); the fact of the matter is that while they probably have seen 95% of the minutes Ennis played, they have seen what, 10 or 15% of the minutes the other guys have played?

 

[Orangezoo]

Its not modeling its analysis. Modeling means we can predict the rest of a season not the next game.

 

[Orangezoo]

I agree with that, to an extent, but I dont think that is a problem with the model; that's a problem with predicting the results of matchups that have any number of factors that are difficult and/or impossible to measure. It would be the same problem with one of us predicting the outcome of a game.

Exactly.

 

[Cusefannotindc]

But that's the thing, a box score doesn't tell the whole story. Anyone can look at a box score and take a wild guess about who the best player was. Say someone scores 20 points. What if eight of those points came in garbage time with his team down 30 and the opponent stopped playing defense, allowing him four easy layups? His FG% goes up and so does his point total and overall offensive efficiency. Box scores don't tell the whole story.

That's why you need to watch the whole game and why the eye test is more important. Numbers are far from everything. That's why I think these sort of things are a secondary measurement at best, in determining player value.

Those type of stats will regress to the mean.

 

[Cusefannotindc]

So you are saying the delta is 10 pct? BS.

It was an example. But in 2011-2012, the Vegas favorite won 70%, Pomeroy's favorites won 77%.

 

[Orangezoo]

It was an example. But in 2011-2012, the Vegas favorite won 70%, Pomeroy's favorites won 77%.

And thats not the eye test vs a model. The reality is numbers and the eye test are used together. It depends on the choice made by the individual. In a vacuum we all know nothing about sports and use a model vs knowing lots and using none. The models and analysis used are created or pre existent. Unless you can make a point that someone who uses a model by itself vs someone who uses some basic statistics to compare two teams has a significant delta then you have no argument that models are more effective and a good predictor.

The thing that makes sports what they are is unpredictability. There are hundreds of methodologies to predicting outcomes. Most are good and include using the same figures. Creating a formula that combines figures is an arbitrary way to try to make it a more exact science. It really does not do any such thing other than provide many the perception that it does. If modeling hit 90 pct you have a point. But since resides within a range any non modeler could hit there is just no argument its something revolutionary or better.

 

[Orangezoo]

Oh and r val can approach -1 or 1 fyi

 

[Cusefannotindc]

And thats not the eye test vs a model. The reality is numbers and the eye test are used together. It depends on the choice made by the individual. In a vacuum we all know nothing about sports and use a model vs knowing lots and using none. The models and analysis used are created or pre existent. Unless you can make a point that someone who uses a model by itself vs someone who uses some basic statistics to compare two teams has a significant delta then you have no argument that models are more effective and a good predictor.

The thing that makes sports what they are is unpredictability. There are hundreds of methodologies to predicting outcomes. Most are good and include using the same figures. Creating a formula that combines figures is an arbitrary way to try to make it a more exact science. It really does not do any such thing other than provide many the perception that it does. If modeling hit 90 pct you have a point. But since resides within a range any non modeler could hit there is just no argument its something revolutionary or better.

I disagree with your argument that only large differences matter. But that likely has to do with our differences in modeling backgrounds. I am used to building models for small but significant changes, where they matter a great deal.

 

[Cusefannotindc]

Oh and r val can approach -1 or 1 fyi

But not R^2

 

[Orangezoo]

I disagree with your argument that only large differences matter. But that likely has to do with our differences in modeling backgrounds. I am used to building models for small but significant changes, where they matter a great deal.

I didnt say large differences. I said differences where you can clearly define you are comparing one set of circumstances vs another set and they are not the same. In this case the delta of vegas vs pomeroy likely contains very similar analysis and there is no way to isolate one methodology vs the other. Therefore it is too autonomous of a decision.