The Puck Stops Here

Several times when I have stated that I support Nicklas Lidstrom of the Detroit Red Wings as MVP, faux rumors makes a comment about his MVP selections (for example this post) that I usually disagree with. I thought it would be best to explain how I select an MVP.

From the standpoint of sabermetrics and hockey, the principle is simple. The MVP is the player who wins the most games for his team. The problem is measuring those wins. I reject outright any of the crazy arguments that usually come up that player X is not so valuable because he has all star calibre teammates or player Y is not so valuable because his team missed the playoffs (presumably due to a lack of all star teammates - which was player X's problem). The only thing that matters is how many wins (win shares - to borrow a Bill James baseball term) a player earns.

It is simplest to try to measure win shares for goalies - but nevertheless problematic. In principle one can measure quite accurately how many goals a goalie prevents (or fails to prevent) versus the average goalie in the NHL over a season. This is the principle of the goals saved method. The method can be further advanced by taking intro account shot quality - if we believe these numbers to be reliable. Once we know the number of goals prevented by a given goalie, we can turn this into wins, using the "Pythagorean formula" where winning percentages (after adjusting for shootouts and regulation ties) scale as the square of the percentage of goals a team scores (ie. a team with 200 goals for and 180 goals against should have a 200² / ( 200² + 180² ) = 55.2%. The problems with this are that it neglects anything else a goalie does other than stop shots (and for example puckhandling ability has a value) and it references things relative to average, where sometimes average is not the most meaningful quantity. Nevertheless, it is possible to get a rough value for how many wins a goalie has contributed to his team.

For a purely offensive player it is also possible to gage his offensive contributions. This can be done using a goals created method where the goals and assists a player gets are turned into a number of goals created. This is at best an approximate number, but it gives a ballpark value that can be compared with a goalie's value. Of course, it assumes a player does nothing except score, which is clearly false. Also, it can wrongly attribute goals, especially in the case where a player benefits by being a linemate of a very good sniper or a very good setup man (or where the player is that sniper or setup man and his linemates are not so great). There is no attempt to take into account defensive play at all.

Defensive play is very hard to accurately measure. One can look at a +/- rating or a team's goals against when a given player is on the ice. These numbers are somewhat "polluted" because they also depend on the quality of teammates, although one can attempt to normalize for them. One can measure things like blocked shots or hits, but these do not tell the whole story. A player who plays his position well and prevents a shot does not get credited with a blocked shot or a hit or any other meaningful number - and these numbers are often unreliable as what constitutes a hit can vary from scorer to scorer. For the most part, to identify good defence nothing beats watch a player play. A good fan can see who is playing well defensively and has some numbers to back up their intuition, but the numbers are not so reliable.

What does this all mean when it comes to picking an MVP for this season? One can get a rough idea of how valuable offensive players are relative to goalies. Right now Alexander Ovechkin of the Washington Capitals has the most goals created (about 56) and Martin Brodeur of the New Jersey Devils has the most goals saved (with about 22). One might argue (correctly) that Brodeur's number is an underestimate due to the fact he is compared to the average NHL goalie, but the point is clear that offensive players this season have contributed more toward winning than goalies (this is not always the case some years). Therefore, it would be a mistake to call a goalie the MVP this season.

Nevertheless, we haven't answered the question of who is MVP. We have concluded that Ovechkin's offense has contributed more wins to his team than any goalie's saves. That is about as much "science" as can go into picking an MVP. Now for the "art" portion of the job. I think it is clear that Alexander Ovechkin has been the best offensive player in the NHL this season. I think it's not much of a stretch to call him the best forward this season as the only forward nearing his offensive numbers is Evgeni Malkin of Pittsburgh and there is really no reasonable argument to make that Malkin contributes more defensively then Ovechkin does (Ovechkin actually has a better +/- rating on a worse team). Any other forward is far enough back offensively that they are unlikely to make up the difference from defensive considerations. But what about defenders? Nicklas Lidstrom leads the NHL's defencemen in scoring. He has about 39 goals created. Puck moving defencemen often have their value underrated by such measures because a lot of their value comes in carrying the puck out of their own zone to start a rush and this does not necessarily lead to goal scoring. Lidstrom is also widely recognized as one of the best defensive players in hockey (if not the best). He has the best +/- rating of any player in the league (which is supporting evidence - but also shows he plays on a good team). I think it is quite reasonable to believe Nicklas Lidstrom has contributed more win shares to his team than Alexander Ovechkin. His defence should make up the difference of 17 goals created. That is why I pick Nicklas Lidstrom as MVP right now.

Hockey sabermetrics is not as well established a science as it is in baseball - and it probably never will be. It is not so easy to develop a good statistical record the way it is in baseball. Nevertheless, it is a guideline for how one picks an MVP. It allows up to make some rough judgments to help us pick an MVP and remove some candidates from contention. I think a good case ca be made that Nicklas Lidstrom has probably got the most win shares this season and thus should be MVP - at least as of this point.