The Legend of Luu Bledsoe (or, who should be the Canucks starter this postseason)?

Thomas Drance
March 27 2012 01:33PM

The relationship between Roberto "Luuu Bledsoe" Luongo and Cory "Schneider-Brady" Schneider has always seemed to be exceedingly supportive and professional. One might even go so far as to call it "chummy." But don't be fooled their eerily similar butterfly styles or their mutual difficulty handling the puck, because when it comes to their respective personalities, they are impossibly dissimilar. Where Cory Schneider is relaxed and as cautious with his choice of words as a politician, none of those adjectives really apply to Luongo. It's what makes him so good at Twitter.

Recently, for the first time, Cory Schneider revealed what's been obvious to anyone with empathy who understands the nature of competitive sports: that he'd like nothing more than to be the Canucks starting goaltender this postseason. Of course, he managed to convey that sentiment diplomatically, but Schneider is a competitor and a stellar performance on a long playoff run could earn him an extra million dollars this offseason (or more). Schneider believes he's good enough now, and so do most Canucks fans and seemingly all of the Vancouver sports media, for that matter.

When Schneider got the surprise start in game six of the first round last season against Chicago, after Luongo's mighty stumble in game five, it was probably the biggest story in Vancouver sports history up to that point (Aaron Rome's suspension, the tire pumping back and forth, and the game 7 riots surpassed it soon after). The tenor if this discussion "who should start" is already beginning to ramp up, and it'll continue unabated throughout the postseason.

Even once the Canucks are eliminated (or somehow, miraculously manage to hoist the cup) how the team resolves the "good problem" caused by Cory Schneider's impending restricted free-agency, and Luongo's immovable contract will be obsessed over all summer, and then for several years afterwards. But we've got plenty of time to digest that once the season is finished, for now let's look at the two goaltender's performance this season and ask the question: which goaltender should the Canucks be counting on when the playoffs start?

(Below is a basic table breaking down each goaltender's performance this season according to some metrics you're probably familiar with. From left to right I've included total save percentage, even-strength save percentage, short-handed save percentage and two more stats you may not be familiar with. QS% is " quality-start percentage " and it's calculated by dividing the total number of starts in which a goalie has a save percentage above .913% or allows only 2 goals against, by his total number of starts. BU% is "blow-up percentage" which I invented in October in an effort to measure goaltender inconsistency.  A blow-up is classified as any game in which a goaltender has a save percentage below .850% or allows five goals on under forty-shots against, and it's similarly calculated by dividing the number of blow-ups by the total number of starts).

 By the numbers, Cory Schneider has had the more impressive season:

2011-12 Season SV% EV SV% SH SV% QS% BU%
Roberto Luongo 0.920% 0.929% 0.871% 60.7% 12%
Cory Schneider 0.934% 0.929% 0.954% 72% 8%

There's a few things worth noting, but let's start with the most simple fact: Cory Schneider has given the team a good chance at winning the game in a higher percentage of his starts, and he's also been significantly more consistent than Luongo has been. Schneider has only been "blown up" on two occasions in 25 starts, while Luongo has had ugly games six times in 51 starts this season. Cory Schneider also has the higher overall save-percentage by .014% (a not insignificant number), which, is what most of the beat guys look at when they write about Schneider giving the team its best chance to win (also they look at GAA and wins and losses, but those are totally meaningless).

The qualifier here is a big one, of course, and thats that both goaltenders have a functionally identical even-strength save percentage (Cory Schneider's save-percentage is actually three ten thousandths higher). This means that Cory Schneider's edge in save-percentage is built on his exceptional performance while the Canucks are short-handed this season, and short-handed save percentage is a not a particularly predictive stat. As Gabriel Desjardins has noted:

Basically, Even-Strength save percentage, which is measured over a large number of shots, is a significant predictor of future performance, though it must be heavily-regressed to the mean save percentage. SH save percentage, on the other hand, is essentially random.

The reason short-handed save percentage is of limited value is that the sample sizes are just too damn small. Luongo hasn't had a short-handed save percentage above .900 for a full season since 06-07, but that's partly because he's usually been such a workhorse. Cory Schneider, on the other hand, is well over .930 for his career, but it's based on a 300 shot sample, so it's dubious to say that he has some sort of "skill" for stopping more shots than Luongo does while his team is shorthanded. In all likelihood, Schneider's short-handed save percentage is an aberration and he'll see that figure regress dramatically over the next couple of seasons (especially if he's given a full time starters job). 

Statistically, both goaltenders have both played extremely well this season, however, Schneider has been considerably more consistent. I'd give him the slight edge, by the numbers he's been the better goaltender (albeit in more limited duty) this season. Will that, however, have any impact on Vigneault's decision for "who should start" an elimination game in mid-May? I'll bet you it does. 

After all, the Canucks have gone out of their way to give him opportunities in hostile environments, it's like they've been giving him a crash-course in "big game experience." Knowing how deliberate this Vigneault/Gillis/Gillman hockey operations team is, that has almost certainly been done that for a reason.

So, while Roberto Luongo remains the teams incumbent starter, and will absolutely start game one of the first round - expect him to be on a much shorter leash this time around. When it comes down to it, the Canucks will play whoever is hot and even if Luongo is flawless, expect Schneider to get into some road games in an effort to reduce Luongo's fatigue and travel burden. It will be a team effort in net this Spring, just as it will be for the Canucks' positional skaters.

3136ae487fac57943f99a50e66e4d6cf
Thomas Drance lives in Toronto, eats spicy food and writes about hockey. He is an NHL News Editor at theScore, the ex-managing editor of CanucksArmy.com and an opinionated blowhard to boot. You can follow him on twitter @thomasdrance.
Avatar
#1 Scott McKenzie
March 27 2012, 11:56AM
Trash it!
0
trashes
Props
0
props

Good stuff. I've been on Team Schneider for a while and because I've been a crappy goaltender my whole life there's no story that's interest me more for the past year or so.

For me, it's the fact that he's cheaper, younger, probably more talented, and comes with zero baggage. Luongo's got the experience, but he's also got the potential to completely fall apart (so does the rest of the team, but it just seems like they feed off of him).

But I've gotta come down to the egos of Gillis, ownership, and to a certain extent AV because they all stuck their neck out on the Luongo contract and to bail on it this early would look terrible, even if it's probably the right move to make should they be able to do it.

Avatar
#3 aidanbc
March 27 2012, 01:28PM
Trash it!
0
trashes
Props
0
props

Hey guys,

I was interested when you noted that Luongo's and Schneider's save percentages are different by by .014% (a not insignificant number). With the relatively small sample sizes by statistical samples, can we say that their save percentages are statistically significantly different? It's a tough thing to measure, and I took some iffy liberties in trying to do so, but let me know what you think:

Basically, I narrowed it down to each goalie's starts (I think score effects impact relief numbers). I then took an average of the save percentage per start (rather than just dividing total saves in starts by total shots faced in starts). These calculations give us the following:

Luongo: 51 starts, 0.911 average save percentage Schneider: 25 starts, 0.914 average save percentage

Now, Schneider's numbers are heavily skewed by a huge blowup vs Nashville, where he allowed 3 goals on 5 shots for a putrid 0.4 SV%. If we drop that obvious outlier, his numbers go to 24 starts and a 0.935 average SV%. Significantly better than Lu's, right? Well, maybe not.

WARNING!! THINGS ARE ABOUT TO GET EXTREMELY STATISTICAL!!

I then ran a 2 sample T-test using this helpful site: http://www.usablestats.com/calcs/2samplet

Basically, the t test in this application treats the game results as samples of a "population" save percentage, essentially what a goalie's SV% would be over an extremely large sample size. It creates an interval of what that SV% could be, given the small size of our actual sample.

The 95% confidence interval of the difference between the two means (.911 and .935) included 0. In other words, we can not say with 95% confidence (the statistical standard) that Schneider's average save percentage is better than Lu's.

I ran the same tests for 19/25 vs 29/51 quality starts, 2/25 vs 9/51 blowups, and 780/835 vs 1381/1513 SV%. In each case, sample sizes make the differences statistically insignificant at the 95% confidence level. (Please let me know if you think I've made any methodological errors, or have misused the T-test).

END OF EXTREME STATISTICALITY

So what does this all mean for analysing hockey stats? In many cases, differences between most players will be smaller than the margin of statistical error, leaving us with nothing to say about them. I'm actually not sure of the answer to my question, so maybe I'll stop blathering and see what you think.

Avatar
#4 aidanbc
March 27 2012, 01:28PM
Trash it!
0
trashes
Props
0
props

Hey guys,

I was interested when you noted that Luongo's and Schneider's save percentages are different by by .014% (a not insignificant number). With the relatively small sample sizes by statistical samples, can we say that their save percentages are statistically significantly different? It's a tough thing to measure, and I took some iffy liberties in trying to do so, but let me know what you think:

Basically, I narrowed it down to each goalie's starts (I think score effects impact relief numbers). I then took an average of the save percentage per start (rather than just dividing total saves in starts by total shots faced in starts). These calculations give us the following:

Luongo: 51 starts, 0.911 average save percentage Schneider: 25 starts, 0.914 average save percentage

Now, Schneider's numbers are heavily skewed by a huge blowup vs Nashville, where he allowed 3 goals on 5 shots for a putrid 0.4 SV%. If we drop that obvious outlier, his numbers go to 24 starts and a 0.935 average SV%. Significantly better than Lu's, right? Well, maybe not.

WARNING!! THINGS ARE ABOUT TO GET EXTREMELY STATISTICAL!!

I then ran a 2 sample T-test using this helpful site: http://www.usablestats.com/calcs/2samplet

Basically, the t test in this application treats the game results as samples of a "population" save percentage, essentially what a goalie's SV% would be over an extremely large sample size. It creates an interval of what that SV% could be, given the small size of our actual sample.

The 95% confidence interval of the difference between the two means (.911 and .935) included 0. In other words, we can not say with 95% confidence (the statistical standard) that Schneider's average save percentage is better than Lu's.

I ran the same tests for 19/25 vs 29/51 quality starts, 2/25 vs 9/51 blowups, and 780/835 vs 1381/1513 SV%. In each case, sample sizes make the differences statistically insignificant at the 95% confidence level. (Please let me know if you think I've made any methodological errors, or have misused the T-test).

END OF EXTREME STATISTICALITY

So what does this all mean for analysing hockey stats? In many cases, differences between most players will be smaller than the margin of statistical error, leaving us with nothing to say about them. I'm actually not sure of the answer to my question, so maybe I'll stop blathering and see what you think.

Avatar
#5 John Andress
March 27 2012, 02:27PM
Trash it!
0
trashes
Props
0
props

@Scott McKenzie

I am one of those in both camps. Perhaps the reason we see Schneider as less likely to implode is because we have seen him in so many fewer games. We have seen him in less than a full season's worth of regular season games and only a very few post season tilts. Do we really have enough info to say that he won't break down but Luongo will when, say, his team's scoring has gone AWOL, the powerplay is as limp as a long dead mackeral and his defense is playing give away in his goal mouth? I don't think so and I am sure that even Mr, Drance would say that reading statistics becomes an increasingly precise art the larger the sample size they are based on. I am going into the playoffs confident that we have two good goalies on our side and by god we are going to need then because I can't remember a season when the depth of the competition in the playoff pool, particularly the Western Conference, has been so high and, as we know to our chagrin, there is only one winner every year.

In regards to Roberto's contract, I, too, think that we should get out of it if we can. Not because I don't like Roberto or because I think Schneider is going to be a better goalie but simply because we have now seen enough of these long term contracts around the league to have understood that, for the most part, they were a misconceived idea in the first place. They very quickly turn into millstones and millstones are one thing a hockey organization can do without.

Avatar
#6 sgolesor
March 28 2012, 08:48AM
Trash it!
0
trashes
Props
0
props

Thom, are there any stats similar to QoT for goaltenders? While I won't argue that Luongo has been less consistent than Schneider, there are definitely some nights where the team just lets him down. Is there any way to account for how good the defense in front of the goalie is when we look at their stats?

Comments are closed for this article.