January 23, 2011
(Ed. Note: Welcome to Stat Nerd Sunday, where we occasionally obsess over hockey numbers like a Dungeon Master obsessing over the level of his warrior elf. Here's Matt Barr, formerly of LCS: Guide To Hockey and Trolleytracks and now blogging hockey at Kertwang.me.)
Everyone knows the Boston Bruins' Tim Thomas(notes) is having an All-World season in goal. How many realize that Nashville Predators' Pekka Rinne(notes) is right behind him? Would you believe Corey Crawford(notes) is a top-five goalie so far this year? Or that Marty Brodeur is out of gas? (Ok, you probably heard that one already.)
Hockey has long been satisfied evaluating and comparing goaltenders using Goals Against Average and Save Percentage, which each have their own problems but share a big one: They fail to account for even strength vs. shorthanded play.
Scoring league-wide goes up some 165 percent when one team has a man advantage. No surprise, right? But that's not because goalies get worse when their team is a man short (well, maybe Marty Turco(notes) does -- see below), or for that matter, better when they have the man advantage. The difference in scoring has little to do with goalies at all.
So why do we assign them one GAA and one Sv% and pretend they tell the whole story?
Let me show you rather than tell you. Here's a sample of some starting goalies' GAA at even strength, with his team shorthanded and with a power play this season (all 2010-11 stats through Thursday, Jan. 20's games):
Taking account of the rate of scoring at even strength and with a man advantage forms the basis of a more thorough and accurate rating of a goalie's overall performance. But then, those situational stats aren't all accrued in the same amount of game play in each situation. This season, the league as a whole is playing just about exactly 80 percent of the time at even strength, and 20 percent with a manpower imbalance. But of course each team has its own split, how much time it, and its goaltender, plays five a side, a man short and with a man advantage.
Let's look a little more closely at two goalies whose overall minutes played are nearly the same, so that superficially they might seem to be handling an equal workload. The estimated situational minutes played for Roberto Luongo(notes) and Marc-Andre Fleury(notes):
Both men have logged close to the same number of overall minutes. But Luongo has played almost a full (overtime!) game's worth of minutes at even strength more than Fleury. Fleury has had to contend with the other guys' power play for about 40 more minutes of ice time than Luongo - a situation when league scoring is way up, no matter who's in goal. But consider, too that Fleury has played another 20 or so minutes with his team with the man advantage - when the shorthanded team can be expected to score far less.
Differences like these in expected scoring, performance and situational minutes played vary with each and every goalie, or at least team, in the league. Happily, armed with each goalie's situational GAA and their estimated situational minutes played, we can compare both to the league as a whole, weight the data appropriately for expected scoring, and quantify how much value each man adds (or doesn't) with his play. Like so:
Since it's in the rules that you should name your new rating system with an acronym that spells an obscure former player's name, I call the seasonal proration of this rating the Situational Offense Expectation & Time Adjusted Evaluative Rating Technique, or SOETAERT. The group above consists of any goalie who has played at least 50 percent of his team's minutes this year. I added together (and weighted) Dwayne Roloson's(notes) Islander and Lightning stats.
What do we see? Rinne gets his due under SOETAERT thanks to his immaculate 3.805 GAA in more than 170 minutes played with the Preds shorthanded. Thomas comes out on top overall, though, largely thanks to a league-best even strength performance (1.684 GAA in about 1,600 minutes).
Bunch of usual suspects and 2010-11 All-Stars fill out the top 10, along with the Caps' Michael Neuvirth, who's doing well compared to league-average in all three situations, and Corey Crawford, currently a monster at even strength (1.799 GAA in about 1,200 minutes) and a slight liability shorthanded (7.186 GAA over about 130 minutes). As a job-security matter, Crawford needn't worry too much about being a "slight liability" a man short: His netmate, Marty Turco, sports a 9.124 GAA shorthanded. We think it's Marty, not his team: In 2009-10 in Dallas, Turco's shorthanded GAA was 9.181; in 2008-09, a relatively pristine 8.065.
Pick a couple guys in the middle of the pack: Kari Lehtonen(notes) and former teammate Ondrej Pavelec(notes). Both can trace their middling SOETAERT ratings to being squeaky clean at even strength but struggling a man short. But there's struggling - like Lehtonen's 7.313 GAA - and STRUGGLING, as in Pavelec's 9.032.
Have a look at 2009-10, minimum 60 percent of team minutes:
"Chris Mason?" you seem to be saying. Well, yeah. He played 11.6 percent of the time for the Blues last year a man short, more than any other starter (by quite a bit), and acquitted himself with a 5.007 GAA, compared to the league's 6.828. He excelled at keeping the power play goals against down, and this rating system recognizes that.
As you see, SOETAERT largely agreed with the Vezina conversation last year, with the winner and a runner-up coming out one-two - Marty Brodeur was eighth, but accumulated the best actual (non-projected) numbers for the season, thanks largely to his 77 very good games played. Fluke? You decide! Here's 2008-09:
SOETART picked the winner and a runner up again 1-2, and the other runner-up, Niklas Backstrom(notes), was second in actual accumulated rating points for the season behind only Henrik Lundqvist(notes).
Whatever you might find interesting in the individual stat lines here, it's important to account for differences in expected offense according to manpower situation when evaluating goalie performance, and SOETAERT is a step in that direction.
Tell your friends.