Navigation: Jump to content areas:


Pro Quality. Fan Perspective.
Login-facebook
Around SBN: Heating Up: Miami Evens Series; LeBron, Wade Take Charge

Price vs. Halak. The Return.

Time to break out an old gimmick to look at last night's return of the prodigal son.

Unfortunately the return involved the Canadiens treating Halak more politely than any time during his Canadiens tenure. I still believe that the Price vs. Halak debate is going to bubble up and I wouldn't be surprised if it is sooner rather than later based on the perfect ingredients brewing in Ken Hitchcock's pot.

Star-divide

The 2011 season was humbling for Canadiens fans who were loyal to Halak. He struggled during his original tenure as a starter with the Blues. After spending his spring posing as Moses holding the tablets, the expectations placed upon him were highly unrealistic.

Halak also found himself in an environment in which his team was hurting his performance. In a shot quality study conducted by Michael Schuckers from Statistical sports consulting he concluded that

Among those with over 1000 shots faced, Halak had to face the most difficult set of shots, on average.

I wouldn't conclude that this is 100% accurate because it is a basic shot location/probability exercise and there are greater factors that determine the quality of a shot, but it is much more reliable than your basic every shot is 100% equal model that 99% of fans base their opinions on.

Without advanced stats, comparing Halak's career stats with 2011 made the Price deal look even more like a steal and fans put their knives back in their drawers and moved forward.

Jaroslav Halak GP WIN LOSS OTL GAA SV% SO
Career 101 56 34 7 2.65 .919 9
2011 57 27 21 7 2.48 .910 7

With wins being a team based stat, it is easy to identify how simplistic goaltending analysis has become. A lower GAA and a better shutout ratio was virtually ignored because of a drop in SV%. With expectation screaming for a repeat of 2010 Jaro looked like a disappointment. He wasn't. Same goalie, poor environment, different expectations.

After a disastrous start to 2012 the best thing for Jaro's resurrection occurred. Ken Hitchcock.

Jaroslav Halak GP WIN LOSS OTL GAA SV% SO
Pre-Hitchcock 64 28 27 7 2.56 .905 7
Post-Hitchcock 15 9 1 5 1.69 .937 2

When Hitchcock was hired, I immediately e-mailed the EOTP crew and offered up that we might be about to see the resurrection of Jaroslav Halak. With the Habs struggling and Price's job getting more difficult by the day and with it a declining SV%, I think we are about to have the tape removed from the muted Halak fans.

The whole point of the original series of breakdowns was pointing out that we needed to dig deeper for a better perspective of what was occurring during the 2009-10 seasons. Forcing fans to look past surface level analysis is something that we still battle to do to this day at EOTP. It isn't a stretch to point out that the majority of fans use one stat to determine a goaltender's worth. That stat in my mind is one of the most flawed stats in sports and was on full display last night.

If we look at the shot clock alone we are left with 25-19. Fairly close. How about adding in misses for Fenwick purposes? 35-31. Corsi? 53-43. All numbers that don't really identify the dominant Blues performance last night. Remove the late powerplay when the game was already decided and the numbers were even closer.

If we peel back the onion the story becomes clearer.

Layer 1 - Shots

Pricevshalak_2012_shots_medium

GREEN BLUE ORANGE RED WITH WHITE
Shot Miss Block Goal


Looking at the shot chart, it doesn't really look all that lopsided. Plenty of dots, fairly well balanced. If writing a column based on the highlights and this alone, we get the narrative that Halak walked into the Bell Center and stymied the Habs by setting the tone with an early breakaway save.

Layer 2 - Scoring chances

Add in the second layer provided by Olivier over at En attendant les Nordiques and we get a little more clarity.

Pricevshalak_2012_chances_medium

GREEN BLUE ORANGE RED WITH WHITE
Shot Miss Block Goal

Removing all the excess data from shot locations with extremely low success rates, we are provided with a clearer picture of what transpired last night. Although both goaltenders faced 14 scoring chances, we see that Price was forced to record a save on over 90% of them. Of Halak's 14 registered chances he was only required to put leather on about one quarter of them. Only four of the Canadiens scoring chances were the result of a registered shot and as much as I would love to credit Halak with great depth and angle coverage, the Canadiens, Mike Cammalleri in particular, were spectacularly inaccurate.

Layer 3 - Rebounds and passing

This layer is totally absent in any metric based stat you can find. Nobody tracks rebounds or shots immediately preceded by a pass. This is why save percentage is extremely flawed. Watch any great passing center and the second they hit the blue line they begin to move laterally. Watch any shootout and it is the primary reason that the shooter will break wide and cut to the net late. They are attempting to alter the goaltenders angle and possibly gain an angle advantage. The ability that separates the average from the elite is usually recoveries. The lockout removed the Robert Esche's and the John Grahame's of the game because of the increased lateral focus.

Pricevshalak_2012_passes_medium

GREEN BLUE ORANGE RED WITH WHITE
Shot Miss Block Goal

In 1985 it was not out of the question to score off the wing with a goaltender set and on his angle. In 2012 allowing a goaltender to telescope out and not worry about a backside recovery or a reckless rebound results in Brian Elliott going from a career .901 SV% to a .940 and likely All-Star appearance.

Last night Carey Price faced 21 shots in which he was forced to set and re-set in transition or recover from a rebound. All three goals were a result of either a rebound that was not scooped up by his defense or a lateral pass that left him in recovery mode. Jaroslav Halak did not have to deal with defensemen not picking up rebounds or one shot preceded by a pass that took place within the home plate area. Every major lateral transition resulted in a blocked shot or a miss of the net.

The Blues were spectacular last night at puck pressure and removing the Canadiens time and space. The Blues bottled up the passing lanes and made Halak's life simple. Outside of the early breakaway and the Tim Thomas like sprawling save on Mathieu Darche, Halak could have completed the win dressed in a speedo and sandals.

After witnessing the Blues last night I am convinced that Halak is on his way to a .920+ SV% this season. After watching the barrage of high quality shots that Price is exposed to on a nightly basis I wouldn't be surprised to see him settle in at the .910 mark that he finds himself around today. Considering that the Canadiens fanbase jumps from scapegoat to scapegoat with the same frequency George Clooney does models, I wouldn't be surprised to hear Halak's name more frequently as they continue to struggle.

Considering that this fan base booed Ken Dryden after six Stanley Cups in eight seasons and mocked Patrick Roy after multiple Conn Smythe trophies, I don't expect Price's Molson Cup greatness to mute the masses.

He's not gone, he's never gone.

Comment 22 comments  |  1 recs  | 

Do you like this story?

Comments

Display:

Awesome analysis. One of these days, I’ll have to find a way to con you out of how you get to make those charts.

Chris Topham asked me once about the next step after scoring chances. I don’t beleive in touches, too fine and labor intensive. Zone-Entries are a bore too, but may be more doable (the Broadstreet Hockey guys are doing it this season and I’m eagerly waiting for the results). Christopher Boucher is in a league of his own with his scouting reports.

The whole notion of lateral movment is really an interesting data point. Fairly easy to include in scoring chances too. Hmmm….

by Olivier on Jan 11, 2012 3:59 PM EST reply actions  

It is totally ignored. The problem with the analysis lies in no existing database, just like your scoring chances it needs to be done manually. Nobody wants to do that. Or do they?

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 11, 2012 4:08 PM EST up reply actions  

BTW, we do need to do a project together where we analyze and chart your data.

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 11, 2012 4:09 PM EST up reply actions  

It would be fantastic to have graphs like this for every game.

Puck Worlds: Chasing Pucks from here to Turku.

For Twitter Updates on Puck Worlds, follow @puckworlds. For updates plus additional witty banter from yours truly, follow @saskhab.

by Bruce Peter on Jan 11, 2012 4:32 PM EST reply actions  

If only I had the time. The graphs are nothing, it is tracking the game which becomes time intensive.

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 11, 2012 4:34 PM EST up reply actions  

This is an amusing pre-emptive strike.

Remember when the debate was at its hottest, it was decided that it would probably take a lot of years to finally rule on which way things would have turned out better?

That became a silly notion to some when proof was so fullsome in Year 1. Now Year 2 might be turning up something different… Well the debate is settled, why are you digging up the old hatchet…

Amusing.

It does obviously bring out the very creative, as many of the most interesting graphs and statistical twists for goaltending evaluation seem to come out of this debate. The graphs above are just fantastic in showing what happened, though the unforunate conclusion is that we’ll just never know how well either guy would have played for the opposite defence on the same night.

I hope we continue to follow this as we all committed we would. I still track Price and Halak thoroughly with different analyses and though halak overtook Price in save percentage last night, it’s still clear that he’s had the worse season to this point by most criteria.

People are going to want to compare goalies around the league, jsut as they are other players. I’d love to see the day when we could have all three levels presented here for every game, but it’s far off. I can’t even find scoring chances for the Blues to compare true save percentages of the two guys.

One of my favourite things I have found is expected save percentage based on shot distance/location. it’s not that sophisticated, but obviously a step above save percentage. It gives an indication of whether a high save percentage comes from good defending (ie.e., it was expected) or from good save-making (lots of saves that an average goalie would never have made).

I think what we’d see with the Blues is that the turnaround for Halak has been a bit of a rebound of his own combined with a significant bump in defensive coverage.

That said, the Blues allow the least shots in the league as well, so their goalies might have complaints about few low quality save percentage padding shots that others benefit from.

by Topham on Jan 11, 2012 5:18 PM EST reply actions  

here is never full definitive proof of anything unless you are comparing Patrick Roy to Steve Penney. If you try to compare Roy to Brodeur your success will ultimately lie with the bias of the individual reading the comparison.

Ken Dryden won 6 Cups in 8 seasons. His numbers are absolutely ridiculous, yet he is never placed in the top 2-3 goaltenders of all-time by ANYBODY. Yet people who will argue that Dryden was great because of his team will place Brodeur at number one.

Is Chris Osgood better than Curtis Joseph? Baseball which is an individual game with zero outside influence still has debate with all the data they have.

I am not digging up the old debate. I am telling you that it hasn’t been resolved. The Halak/Price debate is always there waiting to be ignited. As far as it being a pre-emptive strike, I made that strike two months ago. I have believed in system goaltenders since my old fantasy website and watching Elliot’s numbers and that clinic the Blues put on last night only confirmed my beliefs.

As for expected SV%. It is flawed data, just not as flawed as regular SV%. The distance based metrics are extremely flawed because they judge an 18 foot shot from the slot as the same as an 18 foot shot from the goal line. The study I linked to above is based on the data stripped from cbs sports shot locations (co-ordinates, not distance). It is more accurate, but ultimately flawed as well.

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 11, 2012 5:42 PM EST up reply actions  

I think this goes back to discussions we`ve had previously on how save percentage doesn`t really measure goaltending as defined solely by the goaltender`s contribution but ``goaltending`` as the team`s ability to stop shots from becoming goals. Much like shots against really doesn`t measure the contribution of defenseman alone but the 5 man unit as defenders.

It is however, far fairer to netminders than GAA, or puke wins. At least it says something about what the man in the mask is doing out there.

Writer for http://www.habseyesontheprize.com/

by Stephan Cooper on Jan 11, 2012 6:14 PM EST up reply actions  

Flawed, flawed, flawed.

What’s the point of us talking any stats. They’re all flawed. If a person can’t trusted to estimate shot distance, why should he/she be trusted to guess whether a chance was even on net? Why should we trust someone else to do it?

The only comparison we know that we can trust reliably is a controlled environment, where we change one variable at a time. But if we insist on comparing players on different teams because we find this enjoyable, then we might as well just accept that it’s all flawed and hope that it evens itself out in the same way for everyone involved. Might as well go with expected SV% as far as I’m concerned, because it’s unlikely that scorers are systematically shortening the distance on shots that become saves vs. goals.

by Topham on Jan 13, 2012 1:24 PM EST up reply actions  

You are free to accept what you choose to believe. It is unimportant to me whether you choose to look for a more accurate assessment of shot quality.

Shot distance is flawed. Not because the person tracking the distance has made a mistake, but because listing a shot from 18 feet implies that 1000 shots from 18 feet are equal. You would then come to the conclusion that 100 goals on 1000 shots is a .10 probability of a goal from 18 feet. So that would lead us to the conclusion that 10 shots taken from 18 feet would result in one goal.

Now what if I told you that a shot from 18 feet could result in 50 different locations in the offensive zone? Do you still trust the .10 probability when it could be broken up to 50 different variables within that 18 feet?

Regular SV% breaks every shot into an equal opportunity to score. Distance based metrics break it down into a smaller probability and Schuckers study breaks it down even further based on zone co-ordinates.

As the number gets higher, the variables increase. A 50 foot shot has over 100 different locations in the offensive zone. Using a distance based metric breaks down the offensive zone into 76 different offensive zone distances. When you use a co-ordinate based metric it increases the variable to 4725 different locations.

So call me crazy to distrust the 1 to 1 ratio. Call me crazy to distrust the 76 to 1 ratio and call me even crazier that I don’t trust the 4725 to 1 ratio. Why? Because none of these metrics include rebounds or shots altered by passes.

So while you decide to fluff off my two+ years of research as "amusing" or "creative", I will continue to push to find a way to discover an analytic that provides a better understand of defensive independent save percentage instead of pissing on other peoples research while doing very little of it myself.

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 13, 2012 10:21 PM EST up reply actions   1 recs

Shot distance is flawed. Not because the person tracking the distance has made a mistake, but because listing a shot from 18 feet implies that 1000 shots from 18 feet are equal.

This thinking, I think, is the reason so many people reject microstats out of hand.

The question is not whether 1000 shots from 18 feet are different from one another; the question is whether those 1000 shots are significantly different from another sample of 1000 shots from 18 feet collected in the same manner.

by MathMan on Jan 14, 2012 12:05 PM EST up reply actions  

This is amazing Chris, and like Olivier said it would be amazing to have it every game.

Unfortunately for that I’m guessing you’d have to stop having a real job and work full time on it!

I can’t even imagine how excellent it would be to see a full season’s worth of this data.

Co-editor of Eyes on the Prize
Follow me on twitter: http://twitter.com/andrewberkshire

by Andrew Berkshire on Jan 11, 2012 6:06 PM EST reply actions  

Thanks.

Well, if they continue to shit the bed, it may be more interesting for me to chart the games live than come in here and rehash all my repetitive complaints.

We’ll see.

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 11, 2012 6:10 PM EST up reply actions  

Can’t blame you there, last night was tough to watch and I’m assuming it’ll be even worse in Boston.

Co-editor of Eyes on the Prize
Follow me on twitter: http://twitter.com/andrewberkshire

by Andrew Berkshire on Jan 11, 2012 6:29 PM EST up reply actions  

My dental work was more enjoyable than watching last night. I should have taped the first period and sent it to the dentist. He’d save tons on Novocaine and laughing gas.

Kevin van Steendelaar

http://www.twitter.com/kvansteendelaar

but don't forget...

http://www.twitter.com/HabsEOTP

by Kevin van Steendelaar on Jan 11, 2012 7:45 PM EST up reply actions  

Great stuff

Nice work, Chris.

Btw, I find that the “shot quality” difference between Price and Halak is 0.2% at 5-on-5:

http://behindthenet.ca/goalie_shot_quality.php?sort=5&mingp=1000

Price has an expected save percentage of 941 (using shots + misses), while he posted 946 over the last four seasons. Halak has an expected save percentage of 939 and posted a 941.

So I’m not convinced of the statement “it is much more reliable than your basic every shot is 100% equal model that 99% of fans base their opinions on.” The error bars on “shot quality” are so large, and the results so different depending on the method used to calculate it, that it does not offer any additional information relative to assuming all 5v5 shots are identical.

by Hawerchuk on Jan 11, 2012 6:12 PM EST reply actions  

Crap, misread my own chart.

Halak shot quality: 939; save percentage: 948
Price shot quality: 941; save percentage: 946

Toughest shot quality: Hiller 936

by Hawerchuk on Jan 11, 2012 6:21 PM EST up reply actions  

Oh, and if we include all shots:

Halak shot quality: 930; save percentage: 942
Price shot quality: 932; save percentage: 939

Toughest shot quality: Hiller 929

by Hawerchuk on Jan 11, 2012 6:23 PM EST up reply actions  

But does your expected goals take into account how much of an impact a pass pre-shot has on the expected result? That is not tracked in your data.

Unless you have access to a data base that I am unfamiliar with.

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 11, 2012 6:24 PM EST up reply actions  

No, I was just pointing out that you have at least five shot quality models that I know about – Schuckers, Awad, Ryder, Krywicki and the one I posted. And none of them agree…They also show a very small spread in “shot quality.”

I am not disputing your results (or any shot quality results) at the single shot level or game level. I just haven’t seen them amount to a large difference over the course of a season (eg – we’d see a larger spread in scoring chance/shots than we do in “shot quality” but we don’t.) Maybe if we had touch info and other details we’d find some other trends.

Also, Hitchcock didn’t demonstrate much of an impact on “shot quality” in his previous jobs:

http://www.arcticicehockey.com/2010/10/28/1777880/better-late-than-never-minnesota-wild

by Hawerchuk on Jan 11, 2012 7:22 PM EST up reply actions  

I won’t dispute the current shot quality models just because I know the data is incomplete. So whether Hitchcock can or not is irrelevant if the model is flawed IMO.

The struggle as a goalie following these studies is watching these things unfold in front of you on the ice and not having a way to quantify it. I am sure eventually a way will be discovered and I look forward to trying to poke holes in it and not being able to.

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 11, 2012 10:07 PM EST up reply actions  

And thanks for the kind words.

If you want 30 tweets a day, don't follow me. @ChrisBoyle33

by Chris Boyle on Jan 11, 2012 10:08 PM EST up reply actions  

Comments For This Post Are Closed


User Tools

Welcome to the SB Nation blog about the Montreal Canadiens.

FanPosts

Community blog posts and discussion.

Recent FanPosts

Triple_low_five_small
The Bergevin Bunch – Who should be on the Canadiens’ Shortlist?
Triple_low_five_small
Anti-GM Race: The Search for the Worst Candidate
Triple_low_five_small
Habs GM Search: My Personal Shortlist
Ts_small
Vintage Hockey Jersey Scam
Predsgame_small
So yeah...
Lokomotiv_yaroslavl_logo_small
Welcome Mount Royal Soccer to SB Nation
Triple_low_five_small
The Montreal Canadiens and the Future
Picture_1_small
Other Crazy Eklund Trade Rumours
Picture_1_small
Randy Cunneyworth's Link To Homer Simpson

+ New FanPost All FanPosts >

A Glossary of Analytical Stat Terms

For those unfamiliar with common terms on EOTP such as Fenwick, PDO or relative Corsi, here is a handy explanation.

Click here to access the glossary


Managing Editor

2987845178_b30976f7f9_small Kevin van Steendelaar

Feature Editors

A_new_eotp_logo_small Robert L

P1020029_small Andrew Berkshire

Butch-montreal__2__-_copie__4__small Francis B.

Small Chris Boyle

Lokomotiv_yaroslavl_logo_small Bruce Peter

Small Stephan Cooper

Contributing Writers

Small Olivier

Jp_small Joe Pelletier

Profile_small Melissa_Boufounos