FA Salary crunching

[silvermike]

So, in my copious free time, I went down the list of forwards signed since free agency began, to get a feel for the market rate. This doesn't account for defensive skill, leadership, or anything like that: I just accounted for cap dollars per point scored. Rather than take just the most recent year, I took the middle one of the last three years for UFAs, and the best of the three for RFAs, figuring that would generally account for expecations. And, drumroll please:

 

2009 free agents are paid $75,000 per point. That's a lot of money, and would peg Stafford's contract at $3.3 million for next season. He may accept less than that as long as he doesn't have to give up any UFA years, however, and I'm not sure if lumping RFAs in with UFAs was a good idea. That said, it's proportionate to what David Booth got in Florida - a third more points and a third more dollars.

 

Connolly would have earned $3M under this system, meaning we paid a severe premium for potential. It is, however, in line with Steve Sullivan's deal in Nashville, who is also suffering with injury problems.

 

Looking at notable signings this offseason so far, Cammalleri came in at exactly 75k per point, and the Sedins a hair cheaper. Hossa got paid $73,000.

 

Havlat is making close $87k per point, and Gaborik's injury plagued seasons give him the highest premium on a big money deal: $131k per point. Of course, those injuries really hurt him in this system, and if the docs think he's healthy, the Rags will be fine with his deal. His $7.5M per season would earn 100 points a year, which is an optimistic but not impossible goal for him now that he's off the Wild.

 

In any case, it's a rough system, but it gives you a ballpark of player value based off of scoring alone. We'll see how it holds up this summer.

[wonderbread]

So, in my copious free time, I went down the list of forwards signed since free agency began, to get a feel for the market rate. This doesn't account for defensive skill, leadership, or anything like that: I just accounted for cap dollars per point scored. Rather than take just the most recent year, I took the middle one of the last three years for UFAs, and the best of the three for RFAs, figuring that would generally account for expecations. And, drumroll please:

 

2009 free agents are paid $75,000 per point. That's a lot of money, and would peg Stafford's contract at $3.3 million for next season. He may accept less than that as long as he doesn't have to give up any UFA years, however, and I'm not sure if lumping RFAs in with UFAs was a good idea. That said, it's proportionate to what David Booth got in Florida - a third more points and a third more dollars.

 

Connolly would have earned $3M under this system, meaning we paid a severe premium for potential. It is, however, in line with Steve Sullivan's deal in Nashville, who is also suffering with injury problems.

 

Looking at notable signings this offseason so far, Cammalleri came in at exactly 75k per point, and the Sedins a hair cheaper. Hossa got paid $73,000.

 

Havlat is making close $87k per point, and Gaborik's injury plagued seasons give him the highest premium on a big money deal: $131k per point. Of course, those injuries really hurt him in this system, and if the docs think he's healthy, the Rags will be fine with his deal. His $7.5M per season would earn 100 points a year, which is an optimistic but not impossible goal for him now that he's off the Wild.

 

In any case, it's a rough system, but it gives you a ballpark of player value based off of scoring alone. We'll see how it holds up this summer.

 

wow thats some solid work! not very good news for for DR.

[silvermike]

I should note that I haven't tracked all the earlier signings, but looking over a few signings from the middle of the year, it seems like most of them paid less than that 75 grand rate.

[carpandean]

Nice job. I would suggest a few things:

 

1) The effect of points on salary is nonlinear. That is, if you look at the effect of going from a role player who scores 10 points to a third/second liner who scores 50, you are looking at $500 ($50k/point) --> $3 million ($60k/point), while going to a 90 point top-line forward might cost $7 million or more ($78k per point.) This is a result of supply and demand. There are relatively few 90 point scores, while every GM wants them. 50 point scorers are more readily available.

 

2) Goals cost more than assists. A pure goal scorer like Vanek will generally be paid as much or more for 40 goals, even with relatively few assists, than someone who scores 20 goals and 60 assists.

 

3) Players like Connolly and Gaborik aren't well measured by $/point; they are measured by $/Points Per Game (PPG) and some, possibly incorrect, belief about their ability to stay healthy in the future and maintain those PPG paces. In fact, one could more accurately use $/PPG for all players (assuming some minimum number of games played) and then apply a discount based on injury potential.

 

 

#1 is good for keeping Stafford's salary lower. Of course, then there is "potential" ...

[shrader]

How exactly how are getting total money for points, just looking at the average money over the life of the contract? One thing that will cause some trouble with that is a contract like Hossa's where the money drops off at the end and the player won't ever actually see that part of the contract. Those years are going to pull your money/point figures down a bit.

[silvermike]

How exactly how are getting total money for points, just looking at the average money over the life of the contract? One thing that will cause some trouble with that is a contract like Hossa's where the money drops off at the end and the player won't ever actually see that part of the contract. Those years are going to pull your money/point figures down a bit.

 

That's a good point; I'm going to revise it slightly. I had been just using the cap number, but since a lot of those years are sort of strange, I'm going to average the cap number with next year's salary figure. Does that sound more reasonable?

[wjag]

As a back of the envelope calculation for future conversations, I think we have a winner. We'll see what happens to Stafford. Your number isn't out of line with my "what's he worth" thinking... If nothing else, it gives us a basis for discussion of worth: $= Points * 75K, +/-.. The +/- for the true goal scorer or game changer, aka Heatley, Malkin, OV, Zetterburg, types...

 

And can I say, WOW.. 75K a point... Nice....

[silvermike]

How exactly how are getting total money for points, just looking at the average money over the life of the contract? One thing that will cause some trouble with that is a contract like Hossa's where the money drops off at the end and the player won't ever actually see that part of the contract. Those years are going to pull your money/point figures down a bit.

 

Okay, that actually didn't change much - Hossa's the only one with a really variable contract. He gets much more expensive (about $98k), and the total average goes up about $3,000: $78k per point.

[silvermike]

Nice job. I would suggest a few things:

 

1) The effect of points on salary is nonlinear. That is, if you look at the effect of going from a role player who scores 10 points to a third/second liner who scores 50, you are looking at $500 ($50k/point) --> $3 million ($60k/point), while going to a 90 point top-line forward might cost $7 million or more ($78k per point.) This is a result of supply and demand. There are relatively few 90 point scores, while every GM wants them. 50 point scorers are more readily available.

 

2) Goals cost more than assists. A pure goal scorer like Vanek will generally be paid as much or more for 40 goals, even with relatively few assists, than someone who scores 20 goals and 60 assists.

 

3) Players like Connolly and Gaborik aren't well measured by $/point; they are measured by $/Points Per Game (PPG) and some, possibly incorrect, belief about their ability to stay healthy in the future and maintain those PPG paces. In fact, one could more accurately use $/PPG for all players (assuming some minimum number of games played) and then apply a discount based on injury potential.

#1 is good for keeping Stafford's salary lower. Of course, then there is "potential" ...

Good thoughts:

 

1.) I'm going to need more data before I can really break that out - there have only been about 30 UFA signings this offseason, so I haven't been able to narrow it down too much. That $75k number seems to be strongest at the top - it's what the Sedins earn, and Cammalleri too. They're guys in their prime, so that might be the top-player standard. That middle chunk: Knuble, Recchi, Grabovski, Lehtinen, etc. seem to settle in somewhere in the $45k-50k range. I'd say Stafford may be in there - if you give him the $50k rate, that cuts his expected pay down to $2.25M That seems more fair. The cutoff seems lower than you might think, though - around 55 points.

 

2.)Surprisingly, this doesn't seem to be true - I entered in goals for all these UFAs, and it does give Hossa a big premium, but other guys seem to fall around the same range. We may just need more examples here, since Hossa is the only real sniper on the list.

 

3.) Yeah, it gets complicated. On a full healthy season, Gaborik is being paid for about 100 points. That seems about what he's capable of. He's better than a point-a-game player. Connolly's getting paid for a 60 point season, which is well below what he could do in a healthy season. He just hasn't had one in a while.

[Lanny]

2009 free agents are paid $75,000 per point. That's a lot of money, and would peg Stafford's contract at $3.3 million for next season. He may accept less than that as long as he doesn't have to give up any UFA years, however, and I'm not sure if lumping RFAs in with UFAs was a good idea. That said, it's proportionate to what David Booth got in Florida - a third more points and a third more dollars.

 

Am I incorrect in thinking that unless Stafford or Sekera receive offer sheets, due to their age and not being arbitration eligible, they are required to accept their qualifying offer at the 10% raise. Giving Stafford a number of 1.08 and Sekera .75.

 

Under your $75,000 per point system, MacArther will get $2.3 mil next season, see you later Clarke.

[silvermike]

Am I incorrect in thinking that unless Stafford or Sekera receive offer sheets, due to their age and not being arbitration eligible, they are required to accept their qualifying offer at the 10% raise. Giving Stafford a number of 1.08 and Sekera .75.

 

Under your $75,000 per point system, MacArther will get $2.3 mil next season, see you later Clarke.

 

After the conversation above, I'm looking at more of a $50k/point for guys under 50 points. That'd put him at a more reasonable $1.55M. A premium starts to show up around there. I think a lot of that premium comes from being the kind of player who is likely in line for an offer sheet.

[carpandean]

As you mention, lumping RFAs in with UFAs also isn't necessarily correct. Being a RFA, in essence, reduces competition for that player. Other teams are free to make offer sheets, but they have to pay in compensatory picks if they do (assuming the original team doesn't match.) The increase in cost reduces their willingness to pay. The original team does not have this additional cost, so they will likely get that player for less. Of course, that "potential" factor would be highest for those players, so it may be a wash (though, I suspect, not quite.)

[R_Dudley]

I have to admit I find this all very interesting and compelling facts and reading.

 

I just realized we have a Mensa wannbe sabre's chapter on SS...... ;)

[shrader]

Am I incorrect in thinking that unless Stafford or Sekera receive offer sheets, due to their age and not being arbitration eligible, they are required to accept their qualifying offer at the 10% raise. Giving Stafford a number of 1.08 and Sekera .75.

 

Under your $75,000 per point system, MacArther will get $2.3 mil next season, see you later Clarke.

 

They're never forced to accept the offer. I'm too lazy to look it up, but there is a date where the offer expires, which is generally why you'll see certain guys cave and eventually accept it. Stafford has too much leverage for this to be an issue, but all the rest don't.

[Stoner]

I have to admit I find this all very interesting and compelling facts and reading.

 

I just realized we have a Mensa wannbe sabre's chapter on SS...... ;)

 

Yeah, but the old saying applies. A little knowledge is a dangerous thing.

[R_Dudley]

Yeah, but the old saying applies. A little knowledge is a dangerous thing.

 

I knew you could find a way to bring it back to me... :rolleyes:

[carpandean]

I just realized we have a Mensa wannbe sabre's chapter on SS...... ;)

I meet the qualifications, but in my line of work everyone does, so nobody cares if you actually are a member.

 

Of course, I've also been called the dumbest smart person that someone knows a few times. :blush:

[darksabre]

Yeah, but the old saying applies. A little knowledge is a dangerous thing.

 

and we must remember, "It's a very deadly weapon to know what you're doing." - William Murderface

[silvermike]

As you mention, lumping RFAs in with UFAs also isn't necessarily correct. Being a RFA, in essence, reduces competition for that player. Other teams are free to make offer sheets, but they have to pay in compensatory picks if they do (assuming the original team doesn't match.) The increase in cost reduces their willingness to pay. The original team does not have this additional cost, so they will likely get that player for less. Of course, that "potential" factor would be highest for those players, so it may be a wash (though, I suspect, not quite.)

 

I'm guessing it's not a wash, but that there is too much variability in RFA situations to really get anything to accurate. I think the rule of thumb to go with (and with leadership, grit, and defensive ability not measured, this is never going to be better than a rough estimate) is about $75,000/point for a guy scoring over 60 points in a season, and about $60,000 per under that.

 

For a generic RFA, you'd expect to pay less than his market rate, unless you're asking him to give up UFA years. How much is really tough to say, but if you're paying at- or above-market rate, you're getting ripped off (take note, Darcy and Stafford)

[SabresFan526]

Nice job. I would suggest a few things:

 

1) The effect of points on salary is nonlinear. That is, if you look at the effect of going from a role player who scores 10 points to a third/second liner who scores 50, you are looking at $500 ($50k/point) --> $3 million ($60k/point), while going to a 90 point top-line forward might cost $7 million or more ($78k per point.) This is a result of supply and demand. There are relatively few 90 point scores, while every GM wants them. 50 point scorers are more readily available.

 

2) Goals cost more than assists. A pure goal scorer like Vanek will generally be paid as much or more for 40 goals, even with relatively few assists, than someone who scores 20 goals and 60 assists.

 

3) Players like Connolly and Gaborik aren't well measured by $/point; they are measured by $/Points Per Game (PPG) and some, possibly incorrect, belief about their ability to stay healthy in the future and maintain those PPG paces. In fact, one could more accurately use $/PPG for all players (assuming some minimum number of games played) and then apply a discount based on injury potential.

#1 is good for keeping Stafford's salary lower. Of course, then there is "potential" ...

Carp, why would not linear regression work in this case to forecast what salaries might be for other players? I'm thinking the dependent variable is what I would call adjusted cap hit. Actual cap hit will not work for players like a Hossa or Zetterberg with huge long term contracts where the cap hit lightens out because of what I call the "buy out years." I figure adjusted cap hit would be the cap hit for the first X years of the contract before the "buy out years" come into play. The independent variables are subject to change based on the position i.e. different for centers vs. wings vs. defenseman vs. goalies, obviously.

 

For wingers, I would basically go with games played, even strength goals, even strength assists, power play points, +/-, shorthanded points, and potentially average time on ice as my independent variables. ATOI can help adjust for what line a player is on i.e. first line vs. second line player, etc. However, if I were to hypothesize, I would imagine ATOI for forwards is not statistically significant whereas for defenseman it likely would be.

 

For centers, I'd likely use the same independent variables but include faceoff percentage as well.

 

For defenseman, I'd simply go with games played, even strength points, Powerplay Points, +/-, blocked shots, and ATOI as my independent variables against the adjusted cap hit as my dependent variable.

 

From there you figure out which variables are statistically significant and see what gives you the best R squared. Obviously the model won't be perfect, but I'd shoot for an R squared in 45-55% as a good ballpark to see how effective it is at determining salary. The other thing you could do is create another variable for subjective criteria like "intangibles" and somehow figure out a score per player based on leadership, historical performance, injury prone, age, etc. Potentially age itself could be an independent variable that you use in the actual model. I don't know.

 

Just throwing some thoughts out there, but I would imagine linear regression might be a good way to figure out a player's adjusted cap hit. It's by no means perfect, but I think it could certainly help. I took a business of sports class this last semester and the professor really felt that linear regression could be effectively used to predict salaries of players. His examples were primarily in basketball and baseball, but I don't see why it wouldn't work in hockey if you took the right variables. Interested in your thoughts on this.

[carpandean]

Carp, why would not linear regression work in this case to forecast what salaries might be for other players?

Certainly, one could create such a model. However, given the relatively small sample size and large number of potential independent variables (not just those listed, but things like market conditions, supply within the FA market for a given position, etc., etc.), as well as interaction effects among the independent variables (for example, a higher physicality score would be worth more on an 80 point player vs. a 30 point player), it could be difficult to create a model that truly estimates these effects well. Additionally, the traditional assumptions (for Gauss-Markov; e.g., homoskedasticity) would likely be violated, requiring more advanced models. However, as with most cases, you could likely create one that would be detested by true statisticians, but would work well enough for our purposed. :thumbsup:

[silvermike]

Certainly, one could create such a model. However, given the relatively small sample size and large number of potential independent variables (not just those listed, but things like market conditions, supply within the FA market for a given position, etc., etc.), as well as interaction effects among the independent variables (for example, a higher physicality score would be worth more on an 80 point player vs. a 30 point player), it could be difficult to create a model that truly estimates these effects well. Additionally, the traditional assumptions (for Gauss-Markov; e.g., homoskedasticity) would likely be violated, requiring more advanced models. However, as with most cases, you could likely create one that would be detested by true statisticians, but would work well enough for our purposed. :thumbsup:

 

I wish I remembered more from stats class - I was good up to the r-squared, but I had to make sure that homoskedasticity is a real thing and you weren't just messing with us.

 

This all makes me wonder how detailed the analysis is that GMs and agents do. Because if you were a GM, you have to jobs to do - both figure out the market prices of all of these different players, then see how all the different factors contribute to winning games. Anywhere you could find a discrepancy could give you a pretty big advantage. Maybe I'm too cynical, but my hunch is that most of these guys build teams based on feel and gut instinct rather than anything more formal.

[carpandean]

This all makes me wonder how detailed the analysis is that GMs and agents do. Because if you were a GM, you have to jobs to do - both figure out the market prices of all of these different players, then see how all the different factors contribute to winning games. Anywhere you could find a discrepancy could give you a pretty big advantage. Maybe I'm too cynical, but my hunch is that most of these guys build teams based on feel and gut instinct rather than anything more formal.

Really, I would guess that their method isn't quite as scientific and I'm not sure that it needs to be. They probably look at the figures for recent contracts for each type of players (e.g., 70-90 point finesse center) to get an idea of where the market is, who is available that meets their needs and how much that need is worth to the team composition (a certain player might have more value to one team versus another, meaning they might be willing to pay a higher premium.) Then, they start a bit below that and negotiate up. What gets really fun is trades, where you are not just estimating the value of one player, but the value of the packages from each side.

 

but I had to make sure that homoskedasticity is a real thing

Really, it's heteroskedasticity that you need to verify is real, because if it's not, then the necessary assumption is always valid. :thumbsup:

[Stoner]

Nohomoskedasticity is when an otherwise straight professional athlete gets partial wood when a friend grabs his sausage.

[carpandean]

Nohomoskedasticity is when an otherwise straight professional athlete gets partial wood when a friend grabs his sausage.

I knew that when homoskedasticity was mentioned it would only be a matter of time before someone went there. Nice tap in, Miro.