This has been cooking for quite some time, but I'm finally getting around to posting a little bit of data I've collected. In an effort to better understand goaltenders, I've compiled even strength save percentage numbers from NHL.com and have begun filing through them.

Jonathan Willis wrote a post about Dave Tippett and Ilya Bryzgalov over at Cult of Hockey where he concluded that a lot of goalies were prone to fluctuations in save percentage and that coaches can't really control a goaltender's save percentage. I was interested, because I'd been looking a bit at Ken Hitchcock's work.

We know the story: Ken Hitchcock comes to St. Louis and instantly, Jaroslav Halak and Brian Elliott play like all-stars. Elliott, who had put up save percentages of .909 and .893 in two prior years benefits from Hitchcock's defensive system and posts a .940. Jaroslav Halak goes from a .910 a year ago to .924 thanks to Hitch's defensive system, one that secured his Blues the third William Jennings' Trophy of his career.

The previous winners under Hitchcock? Ed Belfour and Roman Turek in 1998, and Roman Cechmanek with Robert Esche in 2003. He also led Columbus to their only playoff spot and, since leaving his post behind the Columbus Blue Jackets' bench, Steve Mason has put up awful numbers after winning the Calder Trophy in 2009, the only time the Jackets made the playoffs.

Now, one thing that people fail to spot at first is that Brian Elliott actually had a .949 save percentage at even strength in seven games before Davis Payne was even hired in St. Louis. Not that it makes a tremendous difference sample size-wise, but I think it does show that some goaltender performance can be random.

The Hitchcock Effect

Hitchcock has coached every year in the NHL since even strength shot data was recorded by the NHL starting in 1998. Here is how his goaltenders fared in each of these years:

EV SV+ is an adjusted save percentage indicator measured against NHL EVSV%, which is the combined save percentage of all NHL goaltenders. For the purposes of EV SV+, a .900 is exactly league average.

All of Hitchcock starting goaltenders combined end up at above NHL average. Half years are all recorded here because the NHL doesn't split up its data, but it should be fine for us for these purposes.

But split up, what does that make them? For instance, Roman Cechmanek had an EV SV+ of .920 with Hitchcock, but .912 without, despite an awesome 2001 campaign where he had an all-star appearance and a Vezina nomination. That didn't come because of Hitchcock.

I totalled up every goaltender that played time with and without the Blues bench boss. The numbers vary slightly from the chart because I used starters numbers above. I didn't include goaltenders like Fredrik Norrena who didn't play a season without Hitchcock. For multiple years to determine NHL EVSV%, I totaled up every goalie's season and calculated the number of saves an average goaltender would have made that season and divided that total by the total shots faced.

Here are the goalies with Hitchcock:

And without:

This is just a jumble of numbers when it isn't totalled up. Niittymaki, Elliot, Belfour, Halak, Fernandez, Leclaire, Cechmanek, Esche and Mason all fare better under Hitchcock. Notably, Roman Turek fares better without. He won a Jennings his first season after departing Hitchcock, and remains the only goaltender to win two consecutive Jennings' Trophies (or pre-1982 Vezina) for the lowest goals against total in the league on different teams.

Manny Fernandez has two Jennings' to his name , winning as a backup in 2007 and 2009 with Minnesota and Boston.

But what's the total? What's the difference between with and without Hitchock? It's unfortunate because a goalie like Belfour played some of his prime years in the days before EV SV% and for a slight while after his Dallas career when he dipped below average. There's probably no clear way of doing this and there may be a few counting errors, but these numbers are close to as accurate as we can get:

Hitchcock goaltenders had a slightly better save percentage in a slightly weaker goalie pool. Overall, the difference in EV SV+ is .904 to .900. Over the 1500 shots a starter may face in a season, that's a difference of about 4.8 goals.

Is that 0.4% difference statistically significant? On 18,256 shots, random variance gives us a standard error of about 0.2%, so we can be more than 95% confident that the goalies really did play better (and/or face easier shots) when they were with Hitchcock and reasonably confident that that impact would be a couple of tenths of a percent.

It is a small enough effect that some of it could be random chance and some could be Hitchcock playing goalies who were close to their primes. However, random chance probably doesn't explain all of it, and not all of the goalies were in their primes -- Garon played with Hitchcock towards the end of his career, as did Hackett.

It would seem that goalies really do play a little better under Hitchcock, but how much better? We'll see how Elliott plays out his contract with the Blues, but I have a feeling he won't post a .945 over his next 806 shots...