A recurring theme in this space is the mistakes NHL general mangers make in signing or acquiring goaltenders. Steve Mason, Michael Leighton and Antti Niemi are recent examples, although the past is littered with similar missteps: Cristobal Huet in Chicago, Roman Turek in Calgary, Pascal Leclaire in Columbus, Andrew Raycroft in Toronto, etc. The penchant for paying a lot for a puck stopper has fallen in the post-lock-out NHL with the saturation of cheap, capable goalies in the market, but it certainly hasn’t been eliminated.

There are several reasons this tendency persists. First is the fact that NHL GM’s justifiably have difficulty evaluating goaltending talent. The combination of team effects, variance, special teams frequency as well as idiosyncrasies of the player and the position make rationally predicting future goalie performance an unenviable task. For instance, Tim Thomas, this season’s SV% leader, was considered an immovable contract last year when his save rate fell to a mediocre .919. For less accomplished ‘tenders’, the up-and-down of their SV% from year-to-year can make them bounce from scapegoat to savior; from capable starter to marginal back-up, depending on the season. Miikka Kiprusoff, who is still perceived as one of the best in the business, has ranged from excellent (.928 ES SV% in 2009-10) to poor (.907 ES SV% in 2008-09) inside of the last three years. His results this season fall between the two-extremes (.918 ES SV%).

So there is a lot of volubility in goalie results – even amongst those near the top-end of the talent pool. As a result, sometimes even seemingly good bets are going to look foolish in the future as a matter of course.

Another contributor to overpaying goaltenders is a bias called probability neglect or “probability blindness” as described by Dan Gardner is in his book Risk:

Toss in a strong emotion and people can become…probability blind. The feeling simply sweeps the numbers away. In a survey, Paul Slovic asked people if they agreed or disagreed with that a one-in-10 million lifetime risk of getting cancer from exposure to a chemical was too small to worry about. That’s an incredibly tiny risk…still, one third dis-agreed; they would worry.

The irony is that probability blindness is itself dangerous. It can easily lead people to overreact to risks and do something stupid…it’s not just the odds that can be erased from our minds…it’s the costs.

The risks of paying too much for a goalie in the NHL are therefore likely swept away by probability blindness because the perceived risks of having poor goaltending are far more obvious and anxiety provoking to an NHL general manager. No single player exerts more influence on a clubs goal differential than it’s goalie and that’s true in both directions – a good puck stopper can save a club 20-30 goals above a similar replacement level player. On the other hand, a bad one can cost a team 20 or 30 goals. That’s a huge swing in results in a league with as much parity as the current NHL, meaning the rewards for purchasing a good goalie are potentially big, while the adverse effects of playing with a sub-par goalie are conversely huge (and unpleasant).

As such, NHL GMs tend to be very willing to pay to lock-up a goalie who has seemingly proved to be effective, even if that proof is a fleeting spike in performance. Paying a little too much for a guy is certainly not an optimal use of cap space, but as Gardner notes, humans are more than willing to forgo efficiency of costs in the face of anxiety-inducing risk. As such, it’s probable in many ways that GMs who make Mason or Leighton type overpayments are purchasing goalies who they perceive to be “not bad” rather than “good”. The distinction is a meaningful one: it means they are not necessarily seeking the most efficient solution, but merely looking to avoid the risk of the worst possible outcome. In this calculation, it’s much worse to be stuck with a terrible, bargain basement goalie than it is to be stuck with a average but over-priced one.

In Risk, Gardner discusses a Rottenstriech and Hsee experiment where participants were given the choice of paying to avoid a bad outcome:

Students were divided into two groups, with one group told the experiment would involve some chance of a $20 loss and the other group informed that there was a risk of ‘a short, painful but not dangerous shock’…Students were then told the chance of of this bad thing happening was either 99 per cent or 1 per cent.

When there was a 99 per cent chance of losing $20, they said they would pay $18 to avoid this almost-certain loss. When the chance dropped to 1 per cent, they said they would pay just one dollar to avoid risk…The students asked to think of an electric shock did something a little different. Faced with a 99 per cent chance of a shock, they said they would pay $10 to stop it. But when the risk was 1 per cent, they were willing to pay $7 to protect themselves. Clearly, the probability of being zapped had almost no influence. What mattered is that the risk of being shocked is nasty – and they felt it.

Clearly when the outcome is emotionally loaded, people stop rationally assessing the costs associated with avoiding it. As mentioned, a bad goalie is the only lone skater on an NHL team that can seemingly sink an entire club: all the other efforts by a NHL GM can be rendered fruitless if the guy between the pipes can’t stop the puck. This is likely why it’s very difficult for general managers to resist signing a guy who has (seemingly) gotten the job done in the recent past – even if the resultant contract is inefficient from a purely rational point-of-view. As such, signing, say, Michael Leighton may be less about maximizing performance or bang-for-the-buck and more about operating under the precautionary principle.