The more that I surf the web looking for the newest in Advanced Statistics in Hockey, the more I realize how much I really like the Puck Prospectus website and the Behind the Net website. The people at those sites are creative and just plain brilliant. The latest and greatest was offered up by Robert Vollman on Puck Prospectus and it is called Goals Versus Salary (GVS). In case you wish to take a gander at it for yourself, here is the link: www.puckprospectus.com/article.php?articleid=236
What GVS attempts to do in a nutshell is compare a player’s on-ice contributions (measured in Goals Versus Threshold – GVT) in relation to the player’s actual salary minus the approximate salary of a replacement level player. Baseball Prospectus has a similar stat in use for baseball called MORP (Money Over Replacement Player) that compares a player’s VORP (Value Over Replacement Player) with his actual salary. This has the potential to be the most useful of the advanced statistics that I have found. Using this simple mathematical formula it becomes possible to not only see if a player is performing up to his commensurate salary but you can also determine what a fair salary would be for any player given his on-ice contributions.
The math of this formula is very simple assuming you already have access to the player’s GVT rating. For the purpose of this example I will use Marian Gaborik, currently of the Rangers and formerly of the Minnesota Wild:
Step 1. Take the Player’s actual salary ($7.5 million) and subtract from it the approximate cost of a replacement player or minor league/minimum salary player ($500,000). This leaves us with a figure of $7 million.
Step 2. Multiply the remainder by 3 - 7 * 3 = 21 (an explanation of why the number three is used will be offered further on in this post). Marian Gaborik’s GVT should have been +21 last season but unfortunately for him and for the Wild, it was only +5.6.
Step 3. Subtract the actual GVT (+5.6) by the GVT rating that he should have posted based on his salary (+21). This leaves you with -15.4 which was Marian Gaborik’s GVS rating for last season.
Now, a quick explanation of why we multiplied the remainder from Step 1 one by 3 as we did in Step 2. If we had a full roster (20 players) of replacement level players who all made the approximate league minimum salary ($500,000) our team payroll would be $10 million. The average payroll last season was approximately $50 million which leaves us with a difference of $40 million in payroll allocated for above replacement level production. This is the same thing we did in Step 1 of the GVS formula. The average GVT rating per team last season was around +120. If we divide the average team GVT rating (+120) by the $40 million left over in the above replacement level production payroll, we are left with 3 goals for every $1 million. That’s why we use three in that step.
Back to Step 3 for a moment. GVT is based in part using statistics more likely to be accrued while actually playing; like goals, assists and such. Thus if a player is injured and misses games then he is likely going to finish with a lower GVT and GVS rating than if he had played the whole year. This is the case with Gaborik who only appeared in 17 contests last season. Had he played a full season, he would have scored a +27 GVT rating given his pace through 17 games. That would have changed his GVS rating to a +6.
I actually don’t have a problem with the GVS stat reflecting poorly on players who miss time due to injuries or what have you. A player who is not on the ice is not producing and therefore not earning his salary. The purpose of this stat is to determine a player’s value in relation to his on-ice performance. This seems to work perfectly toward that end.
Let’s look at the 10 worst GVS ratings for last season to see what they may tell us:
Player Salary (in millions) GVT GVS
Mats Sundin $8.6 1.6 -22.7
Daniel Briere $8.0 4.3 -18.2
Wade Redden $8.0 4.4 -18.1
Joe Sakic $6.0 0.6 -15.9
Dany Heatley $10.0 13.0 -15.5
Marian Gaborik $7.5 5.6 -15.4
Brad Richards $7.8 6.6 -15.3
Sergei Zubov $5.4 -.04 -15.0
Mike Fisher $6.0 2.1 -14.4
Scott Gomez $8.0 8.2 -14.3
I’d have to agree that every player listed on the chart above severely underperformed expectations either because of injury or plain old poor performance. It’s also scary to realize the connection a lot of these players have to the New York Rangers. Two (Redden and Gomez suited up for the Rangers last season. Two more (Redden and Gaborik) will skate for the team this year. Mats Sundin was hotly pursued by Glen Sather last season before he ultimately signed with Vancouver. Richards and Heatley have both been rumored to be trade targets of Sather at various points this off-season. That means that of the 10 players with the worst GVS ratings last season, 6 have either been Rangers or were targeted by Glen Sather to be Rangers at some point. I’m thinking that I need to email this post to Glen Sather so maybe he can use this to help him properly gauge a player’s real value before next summer’s UFA spending spree.
One suggestion I have for improving this stat would be to weigh the player’s contributions versus his salary cap hit rather than his actual salary number. In the cases of Dany Heatley and Mike Fisher it would make a difference. Heatley earned $10 million in salary but because the contract is for a total of 6 years and $45 million, his annual cap hit is only $7.5. The same applies to Fisher who was paid $6 million but whose contract is for 5 years and $21 million or a $4.2 million annual cap hit. Using the cap numbers instead of the actual salaries would mean this:
Player Salary GVT GVS
Heatley $7.5 13.0 -8.0
Fisher $4.2 2.1 -9.0
Heatley’s GVS improved by 7.5 points while Fisher’s improved by 5.4.
Another thing to consider is it is natural that younger players with lower salaries and ascending statistical production are more likely to have better GVS ratings than older players whose salaries are higher but whose skill levels are in decline. Too many GM’s still reward veteran players with expensive contracts that pay them exorbitant salaries well into the declining phases of their careers. This stat unfortunately doesn’t predict a player’s future performance and should therefore only be used by GM’s who are seeking to establish the true worth of a player already in their prime or just entering it. That way they aren’t paying a player on a long term contract for what they were worth 3 or 4 years ago instead of what they would be worth at that time. The object is to pay less money for more production as often as possible.
Here is a list of the top 10 finishers from last season in the GVS category; let’s see if the chart backs up my statements about youth being served with the GSV stat.
Player Team Salary GVT GVS
Evgeni Malkin Pittsburgh 0.98 23.4 +22.0
Zach Parise New Jersey 2.50 24.2 +18.2
David Krejci Boston 0.83 18.3 +17.3
Nicklas Backstrom Washington 0.85 17.4 +16.4
Phil Kessel Boston 0.85 15.3 +14.3
Johan Franzen Detroit 1.15 15.2 +13.3
Devin Setoguchi San Jose 0.85 13.9 +12.9
Loui Eriksson Dallas 1.50 15.5 +12.5
Anton Babchuk Carolina 1.00 13.7 +12.2
David Booth Florida 0.68 12.3 +11.8
Scrolling down this list you’ll find that 7 of the 10 players listed just completed either their 2nd or 3rd NHL season and thus were still playing on their entry level contracts. The other 3 just finished with their 4th season and none have been eligible to cash in on the big UFA bucks. This chart emphasizes exactly the point I made in the previous paragraph that younger players with some NHL experience are likely to have higher GSV ratings than older players with declining production.
I find this particular stat to be extremely useful in determining which teams are getting more bang for their buck; especially when competing in an environment dominated by the salary cap and the limitations it puts on a team’s budget. It is critical that GM’s (yes, this means you Glen Sather) do a better job of analyzing the real worth of players and managing their assets more like an everyday business does. The days of traditional player analysis is at an end.
The more research that I do the more I see that Statistical Analysis in hockey is making great strides almost daily. GVS is yet another example of the brilliance of some of the minds working to modernize the way in which hockey teams are run and hockey players are evaluated.