How Should We Feel About Steve Mason?

(Photo by Jong Kim/

(Photo by Jong Kim/

When the Flyers picked up Steve Mason at the 2012-13 trade deadline, he was arguably the worst statistical starter in the NHL. A strong rookie campaign had been followed by four abysmal years of .899 Sv% goaltending in front of a struggling Columbus team. After watching him finish the season off well in Philadelphia (7 games, .944 Sv%), some were ready to declare that the Flyers had found their answer in net. Too soon. When his 2013-14 season started off fairly well (36 games, .916 Sv%), he was rewarded with a 3 year, 12.3 million dollar contract extension. While the mainstream media celebrated, bloggers and statisticians mostly saw the extension as a piano hanging above their heads from a quickly fraying rope. After a strong finish in 2013-14 a good start in 2014-15, the same question remains: At what point is it no longer too soon to say that Steve Mason is the Flyers answer in goal?

(Note: Statistics in this article are from 5v5 play only. All statistics were obtained from


A lot of fantastic work on goaltending statistics has been done by some of the top stats gurus in hockey, including Eric Tulsky, Brian MacDonald, and Hawerchuk (@BehindTheNet). Many of the findings from their research will have significant value in helping to evaluate Steve Mason and predict his chances of future success. Here are some of the important ideas that I will look at.

Even Strength Save Percentage (ESSv%) – Tulsky has found that ESSv% tends to be a better predictor of future goaltending success. Because shorthanded Sv% tends to fluctuate largely from year to year, ESSv% turns out to be more useful when predicting future performance. The predictive value also increases as the sample size grows. The graph below illustrates this concept.

Screen Shot 2014-12-19 at 12.27.50 PM

Chart by Eric Tulsky

Luck In Goaltending – Hawerchuk’s work (below) illustrates the idea that small samples of great goaltending from league average netminders can be significantly fueled by luck. This is one of several reasons that career statistics tend to be a much better predictor of future success than single season results. The chart below shows the chances that a run of good goaltending is being exclusively driven by luck.

Screen Shot 2014-12-19 at 12.32.41 PM

Chart by Hawerchuk (@BehindTheNet)

Predicting Future Success – MacDonald and Tulsky combined to develop a very useful chart for predicting future ESSv%.

This chart uses career ESSv% along with shots faced to produce a range (confidence interval) for each goaltender. One can expect with 95% certainty that the true talent level of a goaltender will fall somewhere within this interval.

Screen Shot 2014-12-19 at 12.40.56 PM

Chart by Eric Tulsky and Brian MacDonald


When Mason was acquired by the Flyers, he had posted a Sv% of .9135 after almost 5000 shots at even strength during his career. This is well below the league average, which tends to hover around .919. At that time, Mason’s 95% confidence interval for future performance was about .909 – .922, meaning that he was very likely to perform below the league average in the future

Mason played very well during his first 10 months as a Flyer, ultimately earning himself a 3 year, 12.3 million dollar extension. Statistics, as well as most of the blogosphere, suggested that this extension wasn’t well calculated by Philadelphia’s management.

From the time of the trade until the date of the extension, Mason saw 1038 shots at even strength, posting a ESSv% of .9241. Referring to Hawerchuk’s chart above, stretches like that from average or below average goaltenders are fueled completely by luck close to 30% of the time. When you also take into consideration the 4 seasons of poor goaltending that preceded this 43 game stretch, it seems like the Flyers took a somewhat poorly calculated risk with that contract.

After finishing 2013-14 well and getting off to a solid start this season, Mason’s career ESSv% currently sits at .918 after nearly 7000 shots faced. During 89 games with Philadelphia, this number has risen by close to .5% – a very significant jump. Tulsky/MacDonald’s chart from earlier would suggest a 95% likelihood that Mason’s future ESSv% will fall somewhere within the (estimated) range of .9145 and .9245. Certainly this is a solid step up from Mason’s projections at the time of his acquisition in April 2013.


For the Flyers and for fans of Mason, there are both good and bad takeaways here.

The bad takeaway is that Steve Mason’s level of play (.9286 ESSv%) since arriving in Philadelphia still seems highly unlikely to be sustained. According to MacDonald’s model, the upper end of Mason’s confidence interval is about .9245. Based on his career numbers, it is not realistic to expect him to sustain a level of play that is a full .0041 higher than that. While there may be some legitimacy to the idea that a change of scenery has helped Mason in Philadelphia, it is highly probable that he will ultimately come back down to earth.

The positive takeaway is that Steve Mason is almost assuredly better than the disaster of a goalie that we saw during his final 4 seasons in Columbus. He came to Philadelphia with a career ESSv% of .9135, which is now lower than the bottom of his 95% confidence interval. There is a reasonable chance that Mason can continue to provide close to league average goaltending for the Flyers, which certainly was not a realistic expectation when he first arrived.

Given the statistics at both the time of his acquisition and extension by the Flyers, this most definitely seems like a win. While Mason will almost certainly regress towards his career averages at some point, those averages are much better than they were just a short time ago. We still can’t say that his 3 year contract is great, but it is likely to not be a total disaster, and the term is short enough that it won’t drastically hurt the team if it is.

Add a Comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.