Jack’s shooting % - Is it a problem?

[LTS]

21 hours ago, BullBuchanan said:

Sorry if being accurate is too "political" for you. In science, constraints and controls matter, so yes, when you calculate predictive outcomes you have to make sure you're taking all appropriate data into account so that you're just not seeing a correlation = causation scenario.

Re: predicting the future. Do you understand how math and probability works? I'm guessing not.  Predictions are never 100% for a given scenario at a given time. That's not how math works. Probability states that over an extended or even infinite amount of time the algorithm will be true.Predicting the future happens all the time. How do you think doctors know what cancer treatments to provide, or how scientists will predict what path a hurricane will take, how a poker player determines the likelihood of a given play being successful? They use hard data and probability. 

When you have a 0.0001% chance to win a lottery and you win it twice, does that make the math false? No, of course not, the math never changes. What you just experienced is called variance, commonly called luck. Could you experience positive variance or negative variance over the course of a career or even a life? Sure. The math doesn't care how old you are, it just cares what is true. Your positive variance is likely paid for by someone else's negative variance. You can beat a 95% percent terminal illness, lose an election you were favored by 80%, it doesn't matter in the grand scheme. If you bet against the odds, you are invariably going to lose if you live long enough.


Will goalie equipment affect the model? Of course. Will it change the game to a degree that a significant portion of the population will experience a disproportionate rise in their efficiency vs other equally or superior skilled opponents? Unlikely. They're all playing int he same scenario. It's effectively a control. Sure, some may benefit more than others, but it would be reasonable to assume that the majority of the population would be affected more or less equally with potential modifiers based on skill gap. Maybe Crosby benefits slightly more than Skinner, by virtue of already being better.

Don't try to preach about things you clearly don't even have the most rudimentary understanding of.

Whatever scientific method you used to deduce my level of education was flawed. Enjoy your arrogance though. I am sure it will serve you well.

Just curious though, before I go...

If I there's a chance that I could experience a positive or negative variance over the course of my entire life, then how is it possible that I am invariably going to lose if I live long enough?

I don't want to assume that while you were filling your head with ways to put people down that you failed to study logic, but it would appear that you don't understand how it works.

 

[JJFIVEOH]

Maybe somebody mentioned this and I missed it. But Eichel is tied for 4th in the league in assists with Malkin. Sorry, just not going to complain about his performance. Instead of trying to be the savior like he's done in years last, he's matured and is now just doing what it takes to win. Perhaps he's taking the long shot purposely so Skinner and P-Ville can jump on the rebounds. 

[Wyldnwoody44]

We are what 4th in the league and I'm catching up on the boards and we still have some serious sour-pussing going on over a simple conversation, we are winning more than we ever have in the past decade. I don't care if eichel is passing to Dahlin and then to P ville and Skinner and not getting any assists as long as we keep win steaks like this one going, it's all that really matters. I'm a science guy too, but man, some of the stuff is just over the top. Just watch the games and smile, and drink 

[LGR4GM]

9 hours ago, LTS said:

Whatever scientific method you used to deduce my level of education was flawed. Enjoy your arrogance though. I am sure it will serve you well.

Just curious though, before I go...

If I there's a chance that I could experience a positive or negative variance over the course of my entire life, then how is it possible that I am invariably going to lose if I live long enough?

I don't want to assume that while you were filling your head with ways to put people down that you failed to study logic, but it would appear that you don't understand how it works.

 

It isn't about losing per say. Picture a Bell curve. Now in a standard bell you expect to see your variables fall within 95% of that curve, so the -2.5% and+2.5% are outliers. Picture a curve and cut off each end. Those you exclude in this. So what you do is find the mean, where the middle of that curve is. Then you basically figure out how far +/- away from that curve the 95% of your data could fall. So the - don't necessarily mean losing but just falling below the average in that curve. 

Let's say your career sh% is that middle of the curve. You plot the 5-7 season long sh% and then you can calculate the variance and a confidence interval. The confidence interval tells you a range that 95% of the time that sh% should fall. This is a super shortened version and I'm sure I said something wrong but don't think of it as losing but just falling in that range as opposed to hitting your average. 

[bobs0108]

3 hours ago, LGR4GM said:

It isn't about losing per say. Picture a Bell curve. Now in a standard bell you expect to see your variables fall within 95% of that curve, so the -2.5% and+2.5% are outliers. Picture a curve and cut off each end. Those you exclude in this. So what you do is find the mean, where the middle of that curve is. Then you basically figure out how far +/- away from that curve the 95% of your data could fall. So the - don't necessarily mean losing but just falling below the average in that curve. 

Let's say your career sh% is that middle of the curve. You plot the 5-7 season long sh% and then you can calculate the variance and a confidence interval. The confidence interval tells you a range that 95% of the time that sh% should fall. This is a super shortened version and I'm sure I said something wrong but don't think of it as losing but just falling in that range as opposed to hitting your average. 

If you take out this year from his shooting percentage he has a mean of 10.575 (average of averages) and a standard deviation of 2.29769. If you run a normal cdf with a shooting percentage of 20% or higher the odds are .00002 so that’s the right part of the bell curve. So his 20% is a severe outlier from his average. With that we only have 8 data points so it’s not really statistically valid argument do to lack of data points. 

[LTS]

4 hours ago, LGR4GM said:

It isn't about losing per say. Picture a Bell curve. Now in a standard bell you expect to see your variables fall within 95% of that curve, so the -2.5% and+2.5% are outliers. Picture a curve and cut off each end. Those you exclude in this. So what you do is find the mean, where the middle of that curve is. Then you basically figure out how far +/- away from that curve the 95% of your data could fall. So the - don't necessarily mean losing but just falling below the average in that curve. 

Let's say your career sh% is that middle of the curve. You plot the 5-7 season long sh% and then you can calculate the variance and a confidence interval. The confidence interval tells you a range that 95% of the time that sh% should fall. This is a super shortened version and I'm sure I said something wrong but don't think of it as losing but just falling in that range as opposed to hitting your average. 

I completely understand it.  Believe me.

My point, in my response to Joe Stats, PhD, was that he contradicted himself.  He claimed it was possible to live within the variance (be an outlier, anomaly, exception, so on) and also that you will invariably lose.  Those are mutually exclusive of each other.  If there's any chance of living within the variance then there's no way it can be invariable.

Analyzing data though an analysis of past results used to predict likely future outcomes.  However, when historical data is created it is fixed by the influences of the point in time it was created. As such, significant shifts in influences can relegate data to being useless for future predictions because the environment has changed. You can attempt to control for it but you can only do so if the data that you had was properly controlled.

My point about Gretzky was that until he came along there was really no one who scored like he did.  So, if you were using analytics he would have continued to be an outlier.  The idea that two players, when teamed up, might also work in the same way is not insane. It might not be probable, but it certainly is possible.

When someone says that they WILL regress to the norm, that's not accurate. They may LIKELY regress back but we all know that they might not or at least not within a time frame that is pertinent to discussing the team today.  For example, if the Eichel-Skinner combo continues to put up numbers outside of the expected Botterill will be forced to act before any regression occurs.  He may choose to accept that the MOST likely outcome will be that those two will not be the same again once Skinner is signed.  Skinner's agent, I'm sure, is going to take the opposite view.  Even then, they might be great next year, or the year after, and the year after that.  They could continue to be great and eventually someone will find the data point that explains it.  Until then, they'll appear like outliers.

What it really boils down to, in a nutshell, is that you can use data models to predict the likely outcomes and if you restrict yourself to those likely outcomes it's basically taking the safe path. The risk is always in trying to find the exception or at least get an exception long enough to make a difference. 

[BullBuchanan]

15 hours ago, LTS said:

Whatever scientific method you used to deduce my level of education was flawed. Enjoy your arrogance though. I am sure it will serve you well.

Just curious though, before I go...

If I there's a chance that I could experience a positive or negative variance over the course of my entire life, then how is it possible that I am invariably going to lose if I live long enough?

I don't want to assume that while you were filling your head with ways to put people down that you failed to study logic, but it would appear that you don't understand how it works.

 

Seems we have to go back to the ABCs of this. I don't know your education level, and I don't really care. You don't have to be educated to understand that if you're on the wrong side of probability that it cannot last forever. If you quit or die before a correction can take place it doesn't mean that it wouldn't have happened.

If you want to argue on the model being flawed, that's fine. It probably is, because hockey analytics are in their infancy. However to try to say that a player experiencing an extreme statistical outlier is not extremely likely to regress towards his mean, is beyond illogical. You can ask John Scott about his NHL All-star performance or any other one of hundreds of athletes over the last hundred years that had surreal seasons that they never ever came close to replicating again. 

It's cool - you made up your mind before any evidence presented itself and today opinion is as good or better than facts.

[BullBuchanan]

On 11/21/2018 at 1:01 PM, LTS said:

If analytics were as big back in the day... when do you think people would have predicted that Gretzky or Lemieux were never going to regress back to the mean?  Some players set new norms. I know those are extremes, but the underlying point is that sometimes some factors come together that cause an outlier condition to occur.

So, while you say it absolutely WILL happen, that's not necessarily true.  It is more that it's probable it will happen. However, there needs to be an allowance for what happens when certain factors combine and potentially break away from the statistical norm.

This is why arguing metrics is pointless.  It's good for comparisons and projections but it does not predict the future.  The shrinking of goaltender equipment could account for an increase in SH%.  If it does, arguing the metrics from 3 years ago is pointless because the game has changed.  Next year a change in the glove could increase SH%.  Any of these factors will play into it.

Even the same player in the same situation will trend one way or another in relation to changes on the ice.  So you can say it WILL happen and I think that's where people take umbrage.  You say it as an absolute.  It's not. 

 

Lucky for you, that data is available. You know what's cool about them? They prove the point that players regress and rise to their mean. Kinda the opposite of what you were hoping for, so sorry about that.

https://www.hockey-reference.com/players/g/gretzwa01.html

https://www.hockey-reference.com/players/l/lemiema01.html

Gretzky's is especially beautiful, because you can see the impacts of age and experience in the data. He has 3 mean periods throughout his career. He starts his prime in only his second year, and then proceeds to carry much of that forward for the next 7 seven years. The one year he shot 14.9% during that span and hen rose right back up to his mean. After that stretch he begins a 6 year period where he enters the backside of his prime seeing all of his numbers shift a bit downward. at 31 he hits his cliff and begins his sharp but steady decline downward for the next 10 years until he retires. by the time these guys hit 26/27, their best years were already behind them. They each had good seasons afterward, but they were on a clear trend down.

Edited November 23, 2018 by BullBuchanan

[nfreeman]

On 11/22/2018 at 2:47 AM, BullBuchanan said:

Sorry if being accurate is too "political" for you. In science, constraints and controls matter, so yes, when you calculate predictive outcomes you have to make sure you're taking all appropriate data into account so that you're just not seeing a correlation = causation scenario.

Re: predicting the future. Do you understand how math and probability works? I'm guessing not.  Predictions are never 100% for a given scenario at a given time. That's not how math works. Probability states that over an extended or even infinite amount of time the algorithm will be true.Predicting the future happens all the time. How do you think doctors know what cancer treatments to provide, or how scientists will predict what path a hurricane will take, how a poker player determines the likelihood of a given play being successful? They use hard data and probability. 

When you have a 0.0001% chance to win a lottery and you win it twice, does that make the math false? No, of course not, the math never changes. What you just experienced is called variance, commonly called luck. Could you experience positive variance or negative variance over the course of a career or even a life? Sure. The math doesn't care how old you are, it just cares what is true. Your positive variance is likely paid for by someone else's negative variance. You can beat a 95% percent terminal illness, lose an election you were favored by 80%, it doesn't matter in the grand scheme. If you bet against the odds, you are invariably going to lose if you live long enough.


Will goalie equipment affect the model? Of course. Will it change the game to a degree that a significant portion of the population will experience a disproportionate rise in their efficiency vs other equally or superior skilled opponents? Unlikely. They're all playing int he same scenario. It's effectively a control. Sure, some may benefit more than others, but it would be reasonable to assume that the majority of the population would be affected more or less equally with potential modifiers based on skill gap. Maybe Crosby benefits slightly more than Skinner, by virtue of already being better.

Don't try to preach about things you clearly don't even have the most rudimentary understanding of.

Dude — this and several of your other posts in this thread are highly obnoxious.  This is not how we communicate here.