clock menu more-arrow no yes mobile

Filed under:

Money Puck

The Montreal Gazette reports that a handful of teams including San Jose Sharks, LA Kings and probably the Ottawa Senators are using quantitative analysis to gain a competitive edge in the NHL. (I was happy to see fellow southern blogger and devotee of analysis The Forechecker cited in the piece as well.) Personally I see the use of statistics as inevitable since number crunching has proven to supply advantages to certain baseball teams and such analysis has now spread to the NBA and NFL.

Hockey is a game with a lot of randomness. The bounce of the "monkey puck" is unpredictable at times. But over the long haul there are trends that emerge from all of the noise in the data. Teams to sift through the data to find these trends will gain an advantage over their competitors and they increase their odds of winning, it is as simple as that, evolve or be left behind.

A small number of teams are paying a group headed by former Blackhawks GM Mike Smith for this sort of analysis. What I find ironic is that when you look at the league the Edmonton bloggers do some very high end quantitative work and put it all out there for free. And yet the Oilers make one bad decision after another in terms of signing players (i.e. Souray's recent contract). That is one team that could really benefit from in an in-house guy providing them with Return on Investment numbers for available free agents.

In the Thrashers case the team continues to bet millions of dollars each summer that they will defy the odds and guess correctly which players will avoid the negative effects of aging. Each year they spend and spend on players who are over 33 (Holik, Rucchin, Zhitnik, Kozlov, Todd White). The vast majority of NHLers suffer a drop off in both their health and effectiveness after age 33. Hall of Fame level players have longer careers but Father Time catches up with everyone (except Chris Chelios) eventually. The Thrashers keep betting against the house by signing or trading for players over the age of 33.

What I find interesting is that it has taken so long for statistical analysis to be utilized in professional sports. Think about it, what multi-billion dollar industry doesn't employ some form of statistical analysis to inform their decision making process? Yet the NHL is one of the last places where it is acceptable for managers to make multi-million dollar decisions with incomplete data gathered through an idiosyncratic process. If I were a millionaire owner I would ask this question: Are NHL GMs simply more brilliant than mangers in other money making enterprises or are they failing to take advantage of a resource that could make my franchise more successful in the long run?

Everyone has opinions about players and NHL GMs and scouts form these opinions by watching a great many hockey games. The scouting process is absolutely crucial it provides you with information that cannot always be gained from looking at numbers. But at the end of the day a NHL GM (or his scouts) only watch a relatively small percentage of the total minutes played by a guy he just signed to a multi-year multi-million dollar deal. Every GM scouts players or uses scouts--but here is the $64,000 question: What if that player performed better or worse when your scout was NOT watching? How will you know if what you observed was atypical?

Let me give a very specific example of what I am talking about. After Don Waddell traded for Steve McCarthy at the deadline two years ago he was asked about what he saw in this player. Waddell replied something like this "McCarthy was a guy we scouted and liked, but he was not available when we picked because another club drafted him. I remember seeing him play in juniors. His team was not as talented as the opposition that night but McCarthy assisted on all three goals they scored and he had a fight. I thought to myself 'this guy is a player' and we've been interested in acquiring him every since. He is more of an offensive guy who has misused by previous clubs."

The problem with the human brain is that it is very inclined to remember the exceptionial cases--the one game where a guy was fantastic--and forget about all the other games (i.e. data) that suggest otherwise. That is where statistics can be useful because my spreadsheet doesn't forget about all the other games where that player did not perform well. The spreadsheet tells me what is typical for this player and how he compares to other players.

I read an article a few years ago where the author noted how GMs tend to trade for players who have performed strongly against their own clubs. This is exactly the sort of biased decision making that can lead to a big-mistake-contract or a big-mistake-trade. Just because Player X had the game of his life when you saw him doesn't mean that that performance is something you should expect or count on happening again and again and yet this a completely acceptable way of making decisions in profession sports.

Can you imagine McDonald's making a decision about the location of a future store without consulting some numbers on traffic, disposable income, proximity to competitors, etc? What if Wendy's was run by people who placed their stores based on their own seat-of-the-pants opinion about a location while McDonald's chose their store locations after looking at data and models that estimate how that location would likely perform. Which company would you buy stock in? I know I'd take the one that took advantage of all the available information.