As a self-proclaimed advocate of Statistical Analysis and its potential in the world of hockey I must now defend the practice after a recent blog authored by former Minnesota Wild Assistant GM Tom Lynn was published on The Hockey News on August 28th. For those of you who may be interested in reading the whole post, here is the link: http://thehockeynews.com/...only-take-you-so-far.html
While not an outright condemnation of the use of statistical analysis, Lynn does seem to compose his blog with a condescending tone towards the “professionals, professors and students, each purporting to have developed a mathematical measurement of player performance that would ensure success on the ice.” In truth, I agree with much of what Mr. Lynn has to say in his blog; to a certain extent. “The key to understanding the usefulness of statistics is context.” Lynn argues. He then cites an argument used during the arbitration hearing of former Avalanche forward Antti Laaksonen: “I heard an argument in a salary arbitration case years ago that Antti Laaksonen, then one of the better defensive wingers in the game, had only one less even-strength goal than Joe Sakic in 2000-2001, so the only thing standing between Laaksonen and superstardom was Sakic’s five minutes in power play time per game” Of course, as Lynn contends, there is no way that Laaksonen ever should have been compared to Sakic for purposes of considering value and salary.
Another scenario cited by Lynn as an example of stats being misleading because of the context with which the stats were presented was the trade of the late Sergei Zholtok from Montreal to Edmonton for a 9th round draft choice. After coming off of a 26 goal season, it would have seemed that Zholtok should have had more value than a lowly 9th round draft choice but as Lynn states in his post, “On closer examination, however, one found that the Canadiens’ top two centers, Saku Koivu and Trevor Linden, had been injured much of the 1999-2000 season and Zholtok had received all of the offensive center ice time during that stretch, giving him the opportunity for a career year.” Lynn’s position makes sense; context is critical when using statistics to assess the values of players.
I do think that Mr. Lynn is over-simplifying statistical analysis. In both of the examples mentioned above the only stat being referenced is goals scored; Laaksonen’s number of even-strength goals relative to that of Joe Sakic’s and Sergei Zholtok’s goals scored in a given year. The whole purpose of modern analysis is to develop numbers and stats that encompass more than just goals or points. In previous points I have identified a number of modern stats that attempt to incorporate as many facets of a player’s on-ice contributions as possible. These new stats try to account for factors such as when goals are scored, the strength of competition that a player faces in given situations, etc. Essentially they are doing their best to make context a part of the formula.
Mr. Lynn also mentions that the NHL already distributes a plethora of statistical information to the teams, although, “the approaches tended to be hindsight oriented, meaning they were designed to explain events that happened in the past, rather than what is going to happen in the future. Put another way, it is much easier to develop a scientific explanation for how a tornado developed than it would be to accurately predict the next one”. This statement is dead on. Trying to use a mathematical formula to predict a player’s future performance is next to impossible. Again, though, I think that Mr. Lynn is at least partially missing the boat. Advanced statistics aren’t designed to project a player’s future contributions as much as they are designed to gauge past contributions in order to establish a trend. The key is in evaluating trends involving a specific player and combining that with traditional methods of player evaluation.
The most practical way to utilize advanced statistics, or metrics, is the same way as what any management team of any business should be looking at; through the parameters of trends. For example, if a player has scored at least 30 goals in 4 out of the previous 5 seasons and none of the underlying factors change (i.e. player is not reaching an age where a decline in production should be anticipated or the player has not changed teams, etc.) then isn’t it reasonable to conclude that this player is likely to score at least 30 goals again the next season? This is an example of utilizing statistical trends to project future production while incorporating context. It’s the same as any company projecting anticipated sales based on previous years’ numbers for the same period.
In actuality, despite the perceived sleight of statistical analysis, Mr. Lynn has actually reinforced my own position on its use in the NHL. Mr. Lynn never says there isn’t a place for advanced statistics in the NHL; rather he seems to argue that it is important to not discount the other elements of player evaluation. That’s a position that I can agree with.