Coming Soon

[Darkstone42]

Since it's free agency and there's a lot of talk about monetary value for players, I've decided to create a formula that can be used to determine approximately what each player should be worth on the open market. I'm plotting every NHL player's ppg vs. his salary and fitting a linear trendline to it, and then getting an equation from that line. Since this factors in every NHL player, I feel fairly confident it should describe the worth of a player who has a certain ppg with average defense, agitation ability, physicality, and intangibles, since it does factor in literally every NHL player, and thus should provide at least something close to that average.

Anyway, I just wanted you to know that I'm doing this so you can all get excited about it. =P But it's going to take a little while to take 800 ppg stats and look up every player's salary, too. Expect it in about a week or so. Until then, I'll take feedback on the idea.

=)

EDIT: I forgot to mention that this is just forwards. I'll do something for defensemen and goalies later.

 

[dare2be]

EDIT: I forgot to mention that this is just forwards. I'll do something for defensemen and goalies later.

That was going to be my only comment.

 

[Bizzle McDizzle]

what's your plan for differentiating defensman who are not expected to have a high ppg, yet are worth a lot based on how they play defense?
I assume you are leaving goalies out then?

 

[Bizzle McDizzle]

i type too slow.

nevermnd

 

[Forty_Sixand2]

Shouldn't you leave the regression modeling to me? I am the epidemiologist.

 

[Darkstone42]

Shouldn't you leave the regression modeling to me? I am the epidemiologist.

Haha, you can feel free to make adjustments to it after I finish it if you choose to do so. Us number nerds need to stick together. =P

And I got the idea from my work with kinetics. So there. =P

 

[Forty_Sixand2]

(NERD ALERT: PLEASE STOP READING NOW IF YOU NOT KNOW WHAT THE TERM "ANALYSIS OF VARIANCE" MEANS)

What program are you using to run your models. I would assume some sort of stepwise regression is ideal as there is no real limit to how many variables you can use. I think that You can use continuous variable as well as none of the predictor data should be off the chart. However you can add in some categorical indexes for defensive responsibility and the such. Maybe use a dummy variable to add a bit of value to centers over wingers. If you use SAS or export your data file to me I could add tweaks if u want. However I am sure you do not need much help. You know your shit.

 

[Darkstone42]

(NERD ALERT: PLEASE STOP READING NOW IF YOU NOT KNOW WHAT THE TERM "ANALYSIS OF VARIANCE" MEANS)

What program are you using to run your models. I would assume some sort of stepwise regression is ideal as there is no real limit to how many variables you can use. I think that You can use continuous variable as well as none of the predictor data should be off the chart. However you can add in some categorical indexes for defensive responsibility and the such. Maybe use a dummy variable to add a bit of value to centers over wingers. If you use SAS or export your data file to me I could add tweaks if u want. However I am sure you do not need much help. You know your shit.

I'm using Excel for now, because I'm just using a linear model. Until I can figure out how to assign monetary values to defensive play and centers' value over wings and so on, I won't be expanding it. Perhaps you could help me there. I should have Kaleidagraph 4.0 by the week's end, and it would support an expansion to the formula.

My ultimate goal is to develop a numerical value to assign for player value. I'm hoping that I can use this salary system I've devising now, which is simple and linear, and start branching off of it, assigning higher salaries for centers than wings and for better defensive of more physical players, and from that monetary amount, derive a number value for defensive capability. Of course, it'll be very difficult to compile such a stat during the season, but perhaps coupling these numbers with observations could ultimately lead to a new defensive stat.

It might be a pipe dream, but I figure it's worth a shot.

 

[Automattic]

(NERD ALERT: PLEASE STOP READING NOW IF YOU NOT KNOW WHAT THE TERM "ANALYSIS OF VARIANCE" MEANS)

What program are you using to run your models. I would assume some sort of stepwise regression is ideal as there is no real limit to how many variables you can use. I think that You can use continuous variable as well as none of the predictor data should be off the chart. However you can add in some categorical indexes for defensive responsibility and the such. Maybe use a dummy variable to add a bit of value to centers over wingers. If you use SAS or export your data file to me I could add tweaks if u want. However I am sure you do not need much help. You know your shit.

I read it anyway, just to piss you off. Maybe us dumb animals wanna learn something too!

 

[puckhead]

food for thought:
- is cap hit more effective than salary? might not be given the few Hossa-style deals, but for the most part it should be a more smoothed out measure of remuneration.

 

[Darkstone42]

food for thought:
- is cap hit more effective than salary? might not be given the few Hossa-style deals, but for the most part it should be a more smoothed out measure of remuneration.

Oh, yeah, I'm using cap hits. I said salary by mistake. Not only does the cap hit provide a smoother measure, as you suggest, they're also a heck of a lot easier to find.

I've gotten the first 80 players done, by the way, and right now the scatter is enormous. It should be interesting to see how it plays out as I approach the 520 or so forwards on the list I'm viewing.

 

[Darkstone42]

A quick update on my progress:

I'm about 1/3 through the league, and I have over 150 data points. The scatter is starting to diminish just because of the size of the sample. By the end, I expect to see a line with essentially even distribution around it, and then I'll present the equation.

 

[Jimmy_the_Tongue]

Since it's free agency and there's a lot of talk about monetary value for players, I've decided to create a formula that can be used to determine approximately what each player should be worth on the open market. I'm plotting every NHL player's ppg vs. his salary and fitting a linear trendline to it, and then getting an equation from that line. Since this factors in every NHL player, I feel fairly confident it should describe the worth of a player who has a certain ppg with average defense, agitation ability, physicality, and intangibles, since it does factor in literally every NHL player, and thus should provide at least something close to that average.

Anyway, I just wanted you to know that I'm doing this so you can all get excited about it. =P But it's going to take a little while to take 800 ppg stats and look up every player's salary, too. Expect it in about a week or so. Until then, I'll take feedback on the idea.

=)

EDIT: I forgot to mention that this is just forwards. I'll do something for defensemen and goalies later.

I'm glad that someone else got started on this first....I was jusr about to create a theorem thingy with a whatchamacallit it that would describe the difference in variables and other thing a majjigies....but , alas I didn't have enough jiggawatts to fire the damn thing up...

 

[puckhead]

A quick update on my progress:

I'm about 1/3 through the league, and I have over 150 data points. The scatter is starting to diminish just because of the size of the sample. By the end, I expect to see a line with essentially even distribution around it, and then I'll present the equation.

I'm actually getting pumped about this. holy slow news day.

x = a + b*y
will a = ~$550,000 (league min)?

thanks nerd!

 

[dash]

I'm glad that someone else got started on this first....I was jusr about to create a theorem thingy with a whatchamacallit it that would describe the difference in variables and other thing a majjigies....but , alas I didn't have enough jiggawatts to fire the damn thing up...

I could send you some whoosits if you think it would help...

 

[Darkstone42]

I'm glad that someone else got started on this first....I was jusr about to create a theorem thingy with a whatchamacallit it that would describe the difference in variables and other thing a majjigies....but , alas I didn't have enough jiggawatts to fire the damn thing up...

Well, here's your problem. You forgot the do-hickey!

 

[Darkstone42]

I'm actually getting pumped about this. holy slow news day.

x = a + b*y
will a = ~$550,000 (league min)?

thanks nerd!

It depends on what the fit is. It could just be a discretionary decision to say that the league minimum is the lowest contract you can offer. Right now, it looks like the intercept is negative, which clearly makes no sense, but if I do force the line through a point, it could significantly affect the higher salary players, which would not be desirable.

Of course, I still have 2/3 of the league to go through, so the fit might end up being forced through something around $550,000 anyway once all the data is in.

 

[Darkstone42]

Also, so far of all the teams I've looked at, the Rangers have had the most bad contracts based on the trendline. But I think that's to be expected...

 

[kenyg54]

Dang. The thread title got me all excited

 

[dare2be]

Dang. The thread title got me all excited

What, and the content didn't?