Market Value Mathematics and Made up Statistics

[gregkash]

July 1st came and everyone kept going blah blah over paid for blah blah, and then the Sabres signed Leino for what I originally thought was a ridiculous amount of money until I thought, well, let's compare Brad Richards cap hit vs Production to Ville Leinos using a simple x/y = a/n formula. Figuring out that Leinos contract was dead on I started to feel better. I liked doing this, so I decided to make up my own statistic and focus on defensemen.

 

Basically, what the statistic does is give the actual market value of a player based upon the deal that James Wisniewski signed. Please note, I'm not saying that these players should have contracts based on this, I'm just saying, based upon what Wisniewski signed for, this is their market value.

 

Here's what the statistic is, it puts the most emphasis on points scored and plus/minus since those two values seem to be characteristics of the highest paid defensemen.

 

So basically here's the formula:

(5 * Points) + (.5 * Hits) + (.5 * Blocks) + (5 * Plus/Minus) + ((.5 * Avg TOI) * (.5 * GP))

 

I realize many of the parenthesis are not needed but it helps me visualize it better, so whatever.

 

That gives you a number, and then using x/y = a/n you can compare what value that person played like if they were a UFA.

 

For example, Wisniewski scored a 729.75 on the formula.. and he has a cap hit of 5.5 Mil..

Tyler Myers scored a 737.5 on his score. So based on what Wisniewski made, Myers Market Value is 5.5Mil.

 

I'm not saying he's going to sign a contract with a cap hit for 5.5 mil, I'm saying based on the market, last year he played like a 5.5 Mil Defenseman.

 

Many factors work into contracts and contract numbers, Age, Potential, Leadership, RFA vs UFA so there's never going to be a way to accurately predict the complete value of a player. But this does give you some sense of what value teams are getting from their players..

 

I listed every defenseman with at least 60 Games played by team..

 

The Top 3 defensemen in the formula are: Zdeno Chara (1033) Kris Letang (955) Shea Weber (954)

 

I can categorically say that James Wisniewski is overpaid because virtually everyone in the league is playing better than their contract.

 

Not everyone though. According to the Formula, the most overpaid players are: Brian Campbell, Sergei Gonchar, Jay Bouwmeester, Mike Komisarek

 

Players who are most outperforming their contract: John Carlson, Tyler Myers, P.K. Subban (no coincidence they're all on entry level contracts)

 

Anyway, I think it's pretty interesting, nothing to base huge assumptions on, but it's neat to see.

 

I'll include the excel sheet if anyone wants to look it over, maybe make some suggestions to improve it.

 

https://spreadsheets.google.com/spreadsheet/ccc?key=0AnK2TksuSrG7dEpXekhiQnFtS09CUU01LWJ5bVJSMHc&hl=en_US

[SabresFan526]

You could run a linear regression or exponential regression against multiple variables and then determine what a player ought to make against a regression analysis of this year's UFA class. Need to adjust for any multi-collinearity, but it's probably doable. I did a similar exercise like this in my Sports Business Class. The R² will likely be low because the difficulty will be in determining cap hits for guys who are defensive defenseman and their relative value versus an offensive defenseman versus the relative value of a total overall defenseman. Similarly, you'll get into a problem as it relates to forwards who are 3rd line guys versus top line guys especially if one of your variables for the value of a center becomes faceoff percentage.

 

You can try to make it scientific, but some of this is just gut feel by the GMs I imagine. Either way, it would be a fun exercise that I'm sure numerous people on here have tried. I'm looking at you Carp.

[Sterling Archer]

It seems like everyone is worth at least $3mil dollars.

[gregkash]

You could run a linear regression or exponential regression against multiple variables and then determine what a player ought to make against a regression analysis of this year's UFA class. Need to adjust for any multi-collinearity, but it's probably doable. I did a similar exercise like this in my Sports Business Class. The R² will likely be low because the difficulty will be in determining cap hits for guys who are defensive defenseman and their relative value versus an offensive defenseman versus the relative value of a total overall defenseman. Similarly, you'll get into a problem as it relates to forwards who are 3rd line guys versus top line guys especially if one of your variables for the value of a center becomes faceoff percentage.

 

You can try to make it scientific, but some of this is just gut feel by the GMs I imagine. Either way, it would be a fun exercise that I'm sure numerous people on here have tried. I'm looking at you Carp.

 

I understood 7 words in that post.

[gregkash]

It seems like everyone is worth at least $3mil dollars.

 

Not worth 3 million, Market Worth. Basically Adjusting for the standard that is James Wisniewski, they played like they were worth 3 million. THink of it like inflation.

[deluca67]

Not worth 3 million, Market Worth. Basically Adjusting for the standard that is James Wisniewski, they played like they were worth 3 million. THink of it like inflation.

I think you have sealed the fate of all scouts in the NHL.

 

Forget "Video Scouting" here comes "Spreadsheet Scouting".

 

I am joking with you. Nice job! :thumbsup:

[gregkash]

I think you have sealed the fate of all scouts in the NHL.

 

Forget "Video Scouting" here comes "Spreadsheet Scouting".

 

I am joking with you. Nice job! :thumbsup:

 

it's kind of a combination of rotisserie league / money ball approach to valuing players. I doubt anything like this will ever be used by anyone important.

 

Might actually be very valuable to Fantasy Players, especially those who play in leagues where they have to compile a team with a salary cap. knowing that John Carlson gives you more value than a bigger name can be a real advantage.

[North Buffalo]

You could run a linear regression or exponential regression against multiple variables and then determine what a player ought to make against a regression analysis of this year's UFA class. Need to adjust for any multi-collinearity, but it's probably doable. I did a similar exercise like this in my Sports Business Class. The R² will likely be low because the difficulty will be in determining cap hits for guys who are defensive defenseman and their relative value versus an offensive defenseman versus the relative value of a total overall defenseman. Similarly, you'll get into a problem as it relates to forwards who are 3rd line guys versus top line guys especially if one of your variables for the value of a center becomes faceoff percentage.

 

You can try to make it scientific, but some of this is just gut feel by the GMs I imagine. Either way, it would be a fun exercise that I'm sure numerous people on here have tried. I'm looking at you Carp.

 

A couple of the key varialbes include factoring in indifference curves for a particular year v. supply and demand and the old imeasurable social phycho component of that players value in a given city.

[gregkash]

A couple of the key varialbes include factoring in indifference curves for a particular year v. supply and demand and the old imeasurable social phycho component of that players value in a given city.

 

yup, alright, added that into the equation.. same link, you'll notice all the numbers are the same, but rest assured, it's in there.

[SabresFan526]

I understood 7 words in that post.

Regression Analysis is a common tool used for Statistical Analysis and Predictive Modeling. It's based on the ability to forecast against multiple variables and the notion that a dependent variable (in this case salary cap value) can be determined by one or more variables used to project that value (e.g. goals, assists, +/-, blocked shots, PIMs, etc.). Here's some more information on Regression Analysis.

 

In the example here, you are trying to forecast the "relative salary cap value" of a player against multiple independent variables. It can be used to help forecast the value of a player based on statistical performance in their upcoming free agency year. What you have done is not that far off, but it's a bit manual whereas regression analysis is an Excel Add-In and can be done for you without all the time and effort you put in to develop the scoring. Additionally, regression analysis will likely give you a better sense for relative value as your base case is James Wisniewski whereas with Regression, you can figure out the relative value of a particular player based on the results of everyone else in your sample, e.g. James Wisniewski is not your base case but an entire list of forwards and defenseman for a particular UFA class.

 

There are other tools for statistical analysis, but regression is a good one for what it is you are trying to do by projecting the relative cap value of an individual player against everyone else in the league. If you have or get the opportunity to take a statistics or modeling class, I encourage you to do so as it is a great tool for this kind of sports analysis and is fun.

[Taro T]

Regression Analysis is a common tool used for Statistical Analysis and Predictive Modeling. It's based on the ability to forecast against multiple variables and the notion that a dependent variable (in this case salary cap value) can be determined by one or more variables used to project that value (e.g. goals, assists, +/-, blocked shots, PIMs, etc.). Here's some more information on Regression Analysis.

 

In the example here, you are trying to forecast the "relative salary cap value" of a player against multiple independent variables. It can be used to help forecast the value of a player based on statistical performance in their upcoming free agency year. What you have done is not that far off, but it's a bit manual whereas regression analysis is an Excel Add-In and can be done for you without all the time and effort you put in to develop the scoring. Additionally, regression analysis will likely give you a better sense for relative value as your base case is James Wisniewski whereas with Regression, you can figure out the relative value of a particular player based on the results of everyone else in your sample, e.g. James Wisniewski is not your base case but an entire list of forwards and defenseman for a particular UFA class.

 

There are other tools for statistical analysis, but regression is a good one for what it is you are trying to do by projecting the relative cap value of an individual player against everyone else in the league. If you have or get the opportunity to take a statistics or modeling class, I encourage you to do so as it is a great tool for this kind of sports analysis and is fun.

Actually, it wouldn't be a bad idea for you to take a stats course, as I'm sure you could get a ton of material for your 'day' job watching the interactions of the losers (such as Carp, myself, & apparently SF526) that get it and the losers that don't get it. (And at the end of the day, you're probably gonna have more fun hangin' and having a beer w/ the losers that don't get it, than the ones that do; I know I do. ;) And they'll probably give you some more good material. :thumbsup: )

[SabresFan526]

Actually, it wouldn't be a bad idea for you to take a stats course, as I'm sure you could get a ton of material for your 'day' job watching the interactions of the losers (such as Carp, myself, & apparently SF526) that get it and the losers that don't get it. (And at the end of the day, you're probably gonna have more fun hangin' and having a beer w/ the losers that don't get it, than the ones that do; I know I do. ;) And they'll probably give you some more good material. :thumbsup: )

Too true! Sadly, I spent 2 years of my life just living Regression Analysis as every single quantitative and statistical analysis we had to do in MBA School was all based on Regression. Having said that, I would say it was hands down the most important concept I learned in MBA school. All the other stuff is putting BS on a Powerpoint with pretty slides and public speaking and educated debate/opinions based on case facts. Sadly, it's not much different than my day job. Regression Analysis and the 3 day Sports Business Seminar Class I took were the two best things I got out of MBA school. Although, I'd still recommend it to anyone that's interested as MBA school is fun, but the real hard value I got out of it was really around regression. The rest was kind of fluff.

 

Anyway, there's a lesson in there somewhere kids. Not sure what it is.

[gregkash]

I got a 4 on my AP stats class 11 years ago.

[Taro T]

I got a 4 on my AP stats class 11 years ago.

Good for you, Count Rugen. ;)

 

Hope you and everyone else had a good 4th. :beer:

[Braedon]

Too true! Sadly, I spent 2 years of my life just living Regression Analysis as every single quantitative and statistical analysis we had to do in MBA School was all based on Regression. Having said that, I would say it was hands down the most important concept I learned in MBA school. All the other stuff is putting BS on a Powerpoint with pretty slides and public speaking and educated debate/opinions based on case facts. Sadly, it's not much different than my day job. Regression Analysis and the 3 day Sports Business Seminar Class I took were the two best things I got out of MBA school. Although, I'd still recommend it to anyone that's interested as MBA school is fun, but the real hard value I got out of it was really around regression. The rest was kind of fluff.

 

Anyway, there's a lesson in there somewhere kids. Not sure what it is.

 

 

That's my life right now. Although I prefer variance modeling for running interactions, our lab runs most statistical analysis in regression modeling, and usually hierarchical. All about the standardized Betas. Funny thing is I can't seem to get enough stats, it just so damn sexy.

[carpandean]

MBA School

He-he, never heard anyone call it that before.

 

losers (such as Carp, myself, & apparently SF526)

:(

 

 

 

:thumbsup:

[Ghost of Dwight Drane]

it's kind of a combination of rotisserie league / money ball approach to valuing players. I doubt anything like this will ever be used by anyone important.

 

I think you may be surprised. Darcy is king of the spreadsheets so I would not be shocked to hear they run some screening numbers at the very least to try and find value guys.

 

In horseracing, there is a handicapping tool that uses 20 or so variables to try and create a standard and compare performances. It is sort of crazy because I've seen people watch a race with their own eyes and not make a big deal about what they just saw...but then the number comes out and all of a sudden they are beating down the door saying they want first dibs at buying the horse.

[Braedon]

I think you may be surprised. Darcy is king of the spreadsheets so I would not be shocked to hear they run some screening numbers at the very least to try and find value guys.

 

In horseracing, there is a handicapping tool that uses 20 or so variables to try and create a standard and compare performances. It is sort of crazy because I've seen people watch a race with their own eyes and not make a big deal about what they just saw...but then the number comes out and all of a sudden they are beating down the door saying they want first dibs at buying the horse.

 

 

Talk about over-fitting a model.........

[Ghost of Dwight Drane]

Talk about over-fitting a model.........

 

Most of it is needed and 7 or 8 are key with another batch almost constant, but the crazy thing is some numbers are adjusted subjectively so it can easily be wrong if someone does something out of line. I guess you wouldn't call it a pure statistical conclusion, but the idea of a set of numbers driving performance achievments and transactions does happen.