"I'm still going over Desjardins' post, but it would seem to me that the Time On Ice-based system of measure quality of competition got it right"
This is classic bad reasoning. If you're really trying to be scientific, you don't find data to support the conclusion you want and disregard the other data that contradicts what you want. That should clue Staples in immediately that the calculation has a tremendous amount of noise in it. Sadly, it does not.
Like I said, Corsi data is useless without the proper context. All Corsi shows is the relation of shots for and against when you're on the ice. Any fool should be able to understand the notion that when you're starting more often than not in front of your own net, you're more likely to be on the ice for a shot against than a shot for. Even the best faceoff men lose 40% or more of their draws, so it's not uncommon for a shot against to occur without anybody being at fault defensively. This is especially likely for defensemen who have little to no impact on the outcome of a faceoff.
Like QoC, that shows there is a lot of noise involved, which means it's not a terribly useful measurement of anything to any significant level of confidence.
Quote:
Kulka (Mirtle is bad for this as well) is just showing that people who don't understand the numbers they're using shouldn't try to use them to prove or disprove anything.
That's like saying that people who don't understand biorhythms shouldn't use them to prove or disprove anything.
Like I said, Corsi data is useless without the proper context. All Corsi shows is the relation of shots for and against when you're on the ice. Any fool should be able to understand the notion that when you're starting more often than not in front of your own net, you're more likely to be on the ice for a shot against than a shot for. Even the best faceoff men lose 40% or more of their draws, so it's not uncommon for a shot against to occur without anybody being at fault defensively. This is especially likely for defensemen who have little to no impact on the outcome of a faceoff.
Kulka (Mirtle is bad for this as well) is just showing that people who don't understand the numbers they're using shouldn't try to use them to prove or disprove anything.
The thing about these stats is they are meaningful, I my head at any one that says they are useless. the fualt in them is overrating their value, they mean enough, but any one stat is not going to tell the whole story. THe think about Jaybo is what stat exaclty shows he is good, 42 points? his +/-? All the stats add up against him. 6.6 mill for that. I am glad he is not a oiler.
"At any rate, I generated Quality of Competition four different ways for the Edmonton Oilers. Rankings among regular players are below:"
What follows is a table containing 4 wildly different calculations for each player. Proof positive that this QoC number is more noise than signal.
Reminds me of the 3 different Feng Shui experts giving 3 wildly different and conflicting opinions on that episode of Bullshiat with Penn and Teller.
Like Feng Shui, I guess QoC is whatever you want it to be.
Your reasoning is even more flawed than what you're trying to dispute. So because people have different ideas of how to determine QoC, none of them can be accurate?
The thing about these stats is they are meaningful, I my head at any one that says they are useless. the fualt in them is overrating their value, they mean enough, but any one stat is not going to tell the whole story. THe think about Jaybo is what stat exaclty shows he is good, 42 points? his +/-? All the stats add up against him. 6.6 mill for that. I am glad he is not a oiler.
ya, the oilers really dodged that bullet.
Lowebellini must be neo, with all these bullet time contract dodges.
What a slick GM, you can't get in trouble for making the wrong move when you don't make any at all.
From his comments, he is glad he is not an Oiler also.
Your reasoning is even more flawed than what you're trying to dispute. So because people have different ideas of how to determine QoC, none of them can be accurate?
It's possible one or none of them are accurate. Kind of like choosing a religion to follow.
But, since they are all open to "context", which is as subjective as all get out, is it really all that important?
"Bouwmeester is the defenceman who had the most ice time on a bad team in the NHL. Perhaps that explains his inclusion on this list, but it is alarming that he didn’t do enough to drive the Florida offence to have a better Corsi Number. Is that worth $6.6 million a year from the Calgary Flames? I am skeptical."
Jaybo is overrated.
Jaybo was considered the prize UFA available in the offseason by GM's, hockey experts and fans but not by Jim Corsi.
The thing about these stats is they are meaningful, I my head at any one that says they are useless. the fualt in them is overrating their value, they mean enough, but any one stat is not going to tell the whole story. THe think about Jaybo is what stat exaclty shows he is good, 42 points? his +/-? All the stats add up against him. 6.6 mill for that. I am glad he is not a oiler.
I'm not saying it doesn't have any value, but Corsi numbers alone don't do anything to prove that a player is good or bad. +/- is basically the Corsi number after you factor in goalie SV% (or team SH%, but there shouldn't be much difference). Both are end points, but without the starting point, there isn't much to be learned from it.
If you have 2 guys running 10K, and one guy finishes at 2:30pm and the other guy finishes at noon, you don't know who was faster unless you know when each guy started. It's the same idea with Corsi, because it only gives you the finish times.
Jaybo was considered the prize UFA available in the offseason by GM's, hockey experts and fans but not by Jim Corsi.
One of them is over rated to be sure.
Yes, and Olli Jokinen was considered the "prize" at the trade deadline last year.
Now, poor Joke is at best considered forgettable. Even Flash forgot to mention him in the Horcoff thread, when he was busy telling us all about the great centers in the NW division.
It's possible one or none of them are accurate. Kind of like choosing a religion to follow.
But, since they are all open to "context", which is as subjective as all get out, is it really all that important?
Yes it's possible that none of them are accurate, but the fact that there are multiple methods/theories with different resultes alone doesn't do anything to prove they're all wrong either. It only proves that they can't all be completely right.
Your reasoning is even more flawed than what you're trying to dispute. So because people have different ideas of how to determine QoC, none of them can be accurate?
You just answered your own question. Obviously, all of those different calculations are coming up with different things -- you can't put them all under a label called "QoC" and then pick and choose whatever one you think looks like it works out best.
"I'm still going over Desjardins' post, but it would seem to me that the Time On Ice-based system of measure quality of competition got it right"
This is classic bad reasoning. If you're really trying to be scientific, you don't find data to support the conclusion you want and disregard the other data that contradicts what you want. That should clue Staples in immediately that the calculation has a tremendous amount of noise in it. Sadly, it does not.
It really is a pseudo-science.
I don't think he's finding data to support what he wants...in fact, Staples doesn't believe in stats based on plus/minus. I think he's saying that the TOI system matches what his eyes have told him, whereas the other contradicts what he's seen. If you're trying to find a numerical model, you need to have something to compare it to in order to define its validity. In this case, you compare against what you've observed...not perfect, since observations can be inherently biased, but the best available.
I'd agree...the last two QoC measures seem way out of wack compared to what I observed. No way Strudwick and Cogs were matched against the Thornton and Iginlas of the world. That was Souray and Horcoff more often than not. I think the TOI do a good job, though I can only judge that against my own observations.
Also, of course this is pseudo-science. What would you expect? True experimental control of all independent variables? It's a sport, not a physics lab.
Yes it's possible that none of them are accurate, but the fact that there are multiple methods/theories with different resultes alone doesn't do anything to prove they're all wrong either. It only proves that they can't all be completely right.
You're thinking about this from the wrong end of the slide rule.
Either some new way of looking at the world is helpful (with some decent level of confidence), or it's not.
If "QoC" can be calculated in four wildly different ways with wildly different results with low correlation, then obviously it's an exercise in deception through numbers. It means that QoC is whatever you want it to be, meaning it's not scientific, meaning it's not useful as "proof" of anything.
Jaybo was considered the prize UFA available in the offseason by GM's, hockey experts and fans but not by Jim Corsi.
One of them is over rated to be sure.
Those were just raw Corsi numbers. Hence, a player who spends a lot of time on the ice (check) against good players (check) on a poor team (check) will be more likely to have a poor Corsi. Corsi numbers are useful, but like everything they require context. Those who support Corsi tend to acknowledge that and often provide that context (noting what type of opposition the player played, whether he was used in a defensive (own zone) role, etc).
I'd certainly agree that raw Corsi numbers are relatively useless without correcting for TOI and noting the level of competition, etc.
Someone call Flash Walken...I DON'T CARE IF HE'S WATCHING G0-BOT RERUNS IN HIS PARENTS BASEMENT, WE NEED HELP.
I will respond with any kind of nonsense in kind, as I think I have demonstrated in this forum. To say that the oilers are lucky that they don't have a player of Bouwmeesters ability is laughable. Bouwmeester is a elite level player signed for long term at a below market value contract who is under 25 years old.
Any team in the league would have benefited from that signing. He is a player you move pieces to get if you can get him.
Quote:
Originally Posted by Red Deer Rebel
Yes, and Olli Jokinen was considered the "prize" at the trade deadline last year.
Now, poor Joke is at best considered forgettable. Even Flash forgot to mention him in the Horcoff thread, when he was busy telling us all about the great centers in the NW division.
Jokinen had 15 points in 18 games on a brand new team and almost single handedly won two playoff games for the Flames.
Since the lockout, he has averaged goals, assist and point totals that Horcoff has never hit in his single best season.
For Lombardi and a 1st rounder a year from now, that is a great trade.
If "QoC" can be calculated in four wildly different ways with wildly different results with low correlation, then obviously it's an exercise in deception through numbers. It means that QoC is whatever you want it to be, meaning it's not scientific, meaning it's not useful as "proof" of anything.
No, it might just mean that only one of the equations is right.
The key is to determine which, if any, that is. As I said above, the best we can do right now is to compare against what we've seen. By that measure, the TOI/plus-minus QoC measures seem to rank players in the right order. With that validity established (if you believe the observations), you can then make a reasonable assumption that it will rank players on other teams (who you haven't watched) relatively well and it becomes a useful tool.
I don't think he's finding data to support what he wants...in fact, Staples doesn't believe in stats based on plus/minus. I think he's saying that the TOI system matches what his eyes have told him, whereas the other contradicts what he's seen. If you're trying to find a numerical model, you need to have something to compare it to in order to define its validity. In this case, you compare against what you've observed...not perfect, since observations can be inherently biased, but the best available.
The proper way to do it would be to say "Well, my eyes tell me X Y and Z, before I look at the data and the calculations. After I look at the data and the calculations, do any of them consistently tell me X Y and Z? If not, then ignore them. If they do, further study is required".
He's doing it backwards.
Quote:
Also, of course this is pseudo-science. What would you expect? True experimental control of all independent variables? It's a sport, not a physics lab.
The problem with Corsi etc. is that the people promoting them -- snake oil salesmen -- aren't holding them up as a philosophy, but rather as some kind of "advanced statistic", some kind of amazing new way to prove things about hockey. There is never your kind of disclaimer in their statements. Heck, there's so much noise involved here that I'm not even sure one can claim it as a philosophy without doing oneself a disservice.
Like QoC, that shows there is a lot of noise involved, which means it's not a terribly useful measurement of anything to any significant level of confidence.
Just because a stat doesn't prove outright one thing or another, it doesn't mean it isn't a useful tool to get your answer.
Quote:
That's like saying that people who don't understand biorhythms shouldn't use them to prove or disprove anything.
No, it isn't. Kulka shouldn't use Corsi to prove that certain players are good or bad, because that's not what Corsi does. It can be used to help you get to that answer, but you can't use it on its own to prove anything other than how many shots for/against a player was on the ice for.
No, it might just mean that only one of the equations is right.
It means nothing of the sort.
If there are 4 ways to compute QoC, why not 5? If 5, why not 6?
Why can't all of them be wrong?
More importantly, there can't be two different ways to compute different numbers but called the same thing. You can't say "there are four ways to compute QoC". It's fishing.
It's an exercise to parameterize completely subjective rationale and make it look objective and scientific. In short, a form of ignorance at best, deception at worst.
Quote:
The key is to determine which, if any, that is. As I said above, the best we can do right now is to compare against what we've seen. By that measure, the TOI/plus-minus QoC measures seem to rank players in the right order. With that validity established (if you believe the observations), you can then make a reasonable assumption that it will rank players on other teams (who you haven't watched) relatively well and it becomes a useful tool.
It's not a useful tool if no one can agree on how to calculate it, and when it's calculated, how to interpret the numbers. Not to mention the whole Texas Sharpshooter fallacy where people do the calculations first, then determine which one was the One True Method after.
You're thinking about this from the wrong end of the slide rule.
Either some new way of looking at the world is helpful (with some decent level of confidence), or it's not.
If "QoC" can be calculated in four wildly different ways with wildly different results with low correlation, then obviously it's an exercise in deception through numbers. It means that QoC is whatever you want it to be, meaning it's not scientific, meaning it's not useful as "proof" of anything.
If QoC can be calculated in 4 wildly different ways with wildly different results, then all 4 methods can't give you an accurate QoC. That's what I'm trying to say. And just because all 4 don't give you accurate QoC, it doesn't mean none of them do.
To use another example, to find the speed of something, you use the formula of distance/time. If someone comes by and says that you sould do it by dividing distance by weight, and another guy comes by and says it's distance/pi. Having 3 people with 3 different theories on how to determine speed doesn't make "speed=Distance/time" wrong, and it certainly doesn't mean that speed is whatever you want it to be.
I will respond with any kind of nonsense in kind, as I think I have demonstrated in this forum. To say that the oilers are lucky that they don't have a player of Bouwmeesters ability is laughable. Bouwmeester is a elite level player signed for long term at a below market value contract who is under 25 years old.
Any team in the league would have benefited from that signing. He is a player you move pieces to get if you can get him.
Jokinen had 15 points in 18 games on a brand new team and almost single handedly won two playoff games for the Flames.
Since the lockout, he has averaged goals, assist and point totals that Horcoff has never hit in his single best season.
For Lombardi and a 1st rounder a year from now, that is a great trade.
If QoC can be calculated in 4 wildly different ways with wildly different results, then all 4 methods can't give you an accurate QoC. That's what I'm trying to say. And just because all 4 don't give you accurate QoC, it doesn't mean none of them do.
You fail to see the fundamental problem with this "advanced statistic".
Let me ask you a question. How do you compute the (unadvanced) statistic of average? Is there more than one way? How do you compute the (unadvanced) statistic of standard deviation? Is there more than one way?
The answer is, of course, no to both.
More generally, if I give you algorithm A and algorithm B, both which produce different results for the same input I, how can you say both are calculating the same thing? By definition, they are not.
Quote:
To use another example, to find the speed of something, you use the formula of distance/time. If someone comes by and says that you sould do it by dividing distance by weight, and another guy comes by and says it's distance/pi. Having 3 people with 3 different theories on how to determine speed doesn't make "speed=Distance/time" wrong, and it certainly doesn't mean that speed is whatever you want it to be.
You don't understand mathematics is all I will say about that.
my head at any one that says they are useless. the fualt in them is overrating their value, they mean enough, but any one stat is not going to tell the whole story. THe think about Jaybo is what stat exaclty shows he is good, 42 points? his +/-? All the stats add up against him. 6.6 mill for that. I am glad he is not a oiler.
