More blogs about goalies and food

Two things are dominating the Oiler fan (and hopefully Oiler management) thought process:

  • Connor McDavid!
  • Without improving our goalies and defense, we won’t be winning much, even with Connor McDavid!

Now while it is clear that both G and D shortcomings need to be addressed, I have gone on record in stating that improving the goalie situation will have a bigger impact than fixing the defense.  Guest columnist Frjohnk found much the same thing, that while you can’t say very many good things about the Oiler defense, the goaltending was undeniably awful.

So, fixing the goaltending is imperative.  But not easy.

Life’s hard, so is fixing the goaltending

None of the various options the Oilers have for fixing the problem come without some warts.

Do you sign a UFA veteran like Niemi, costing you hefty cap dollars and possibly buying into the down side of that goaltenders career?

Do you sign a younger UFA like Neuvirth and live with the risk of not getting a great goalie, or worse still, repeating last years ‘less experienced goalie’ debacle? (I think this would be a good route and acceptable risk, personally)

Do you trade for a veteran like Anderson, costing you assets and cap, and again risking aging veteran problems?

Do you trade for a youngster like Talbot, which should be an asset and cap friendly move, but risk getting a player where early indicators of talent are misleading?

Woodguy has posted a number of relevant articles recently, and his posts (here, here, and here) are required reading on the topic.  (He makes some pretty powerful arguments for Talbot, and so it’s good to hear that Chia Pet is apparently in the Talbot hunt)

Now with all that (and more) work on the topic, why YALAG (yet another look at goalies)?

Well …

Because I’m going to try and take a slightly different approach to the matter and see if that helps gain any insight that might help with the decision making process.  (We are of course, not involved in that at all, but I suppose conversely it gives us something to cheer for, and a thought process on assessment once/if something gets done).

What I’m going to look at is the career arc of the sv% of a few select goaltenders, with the purpose of trying to get a sense of:

  • How quickly or slowly does a goaltender’s sv% stabilize relative to his career sv%?  (if ever)
  • What can a goalie’s sv% early on (say at 50 games) tell us about his long term career? (if anything)

What I’ve done is selected a handful of goalies, and pulled and graphed their EV cumulative sv% over their careers.  On that graph is a line showing career sv% and sv% at 50 games.  Then I pulled a handful of younger goalies of interest and compared their career sv% pattern to see if that puts any of them in a better or worse light.

A few notes on the process before jumping into the charts:

  • This is more like a longitudinal study rather than a large scale statistical study.  I’ve focused on long service, mostly elite goalies as my archetypes.
  • As a result, as is so often the case, small sample warning!  Also selection bias warning!  I’ve only done this for a handful of selected goalies, so any conclusions have to be taken with a grain of salt.  But!  As measurement science tells us, even looking at a small sample gives us infinitely more information than a non-sample, so I do think there is value to be had here.
  • All sv% shown here are supposed to be EV Regular Season sv%, hopefully I didn’t screw any of the settings up when I downloaded the data.  If you notice a bust, please tell me.
  • All data from the awesome folks at war-on-ice.

The King

OK, without further adieu, the charts … first up is The King.


  • Sv% was volatile and started to stabilize between 120 and 160 games.  Prior to that it was very volatile on both the upside and the downside
  • Sv% at 50 games underestimates career sv%
  • He gets better his whole career after that.  No wonder he’s the King.


Next up, Roboito:

  • Again, sv% at 50 games underestimates career sv%
  • Not a great start to his career, though less volatility early compared to Lundqvist
  • Depending on how you choose to interpret the graph, sv% stabilized between 90 games and 180 games
  • Notice the substantial weakening from 160 to 240 or so?  We’ll bring that up later!

The Volatile One

Any look at goalies has to include at least one French Canadian goalie, no?

  •  Again: volatile early, early games overestimate sv%, sv% at 50 games underestimates career
  • I’d hazard stabilization occurred around 120 games, but improved a fair bit thereafter
  • Like every goalie so far, there is a period of significant weakness either prior to stabilization, or in the first season or two after stabilization could be argued to have occurred

Trade Target X

A look at one veteran goalie who has been in fan discussions regarding a trade:

  • Stabilization at 80 games, which is surprisingly early.
  • Sv% at 50 underestimates career, but by a very small margin (reflective of the early stabilization)
  • You can see the drop off late in his career relative to games 250 to 480 or so.  It’s not consistent with his career arc, which was quite stable, so I would indeed be worried that we’re seeing a permanent dropoff.

Applying the patterns

And now we’ll kick off applying our tentative ‘lessons learned’ from these patterns with someone who’s been an extensive topic of discussion in Oilerland:

What conclusions might we draw?

  • It’s reasonable to think that Sv% at 50 games, about .905, is a low bound, and that Dubnyk is going to be quite a bit better than backup level for most of his career.
  • The dropoff we saw from 130 to 180, the one that got him shipped out of Edmonton?  Not unusual.  Goalies do that kind of thing …
  • If I were to bet where Dubie ends up, he’s going to stabilize his career around .915.  The train wreck of  two years ago is not indicative, but I wouldn’t expect this years Vezina performance is either … it looks like more of a rebound effect (and not an unprecedented one, if you look at games 80 to 130).

Anyone who gets Dubie signed is going to get a good goalie, I’m pretty sure.  But if he gets paid based on this last years performance, I suspect it will be an overpay.

Medium-sized Ben

Now the man who replaced Dubie, Our Man Ben:

Well, now things get interesting, don’t they?

Ben’s decline this last year is not unprecedented among goalies, even elite ones.  Early career volatility is the norm.  It did get a little long in the tooth as declines go, though.

Conversely, if sv% at 50 games is typically an underestimator for career, Ben will rebound this year and have one hell of a career.

What’s the catch?  Sv% at 50 games has been an underestimator for good goalies.  What if Ben isn’t a good goalie?

This is far from definitive, but what I would personally conclude is this:

  • I would not get rid of Ben Scrivens.
  • He may very well just have had a bad season, and be in for a rebound of Dubnyk-like proportions next year.
  • His career arc is just entering the point at which most of the goalies in our study set start to stabilize.
  • At worst, if he settles in on a career somewhere between his 50 game sv% and his sv% after last years disaster, he will be a .910ish goalie, which is a capable backup.

Getting rid of Ben Scrivens carries the very high risk of repeating the mistake the Oilers made with Dubnyk.

Newbies and Youngsters

Let’s look at a couple of other much-discussed options.  First up, The King’s understudy.

Well, if sv% at 50 is a career underestimator, Cam Talbot may very well be the next Dominik Hasek.  I’m getting more on board with Woodguy’s conclusions.  This guy’s game has been so amazing early, that even if he isn’t that outstanding the rest of the way, there’s an excellent shot that he’s still going to be darn good.


If we do sign him, expect volatility in his game, above and below that mark.   If the typical (elite) goalie stabilizes between 120 and 180 games, we should see plenty of bad to go with plenty of good in the next season or two from Cam Talbot.

These patterns suggest that Talbot may very well be a brilliant long-term solution, but a risky short-term solution for a team that needs to shore up goaltending next year, not two or three years from now.


And last but not least (or is he?), Michal Neuvirth:

This is a remarkably stable and consistent goalie already.  (One of my arguments for why I think the Oilers should get him)

He’s not a lights out goalie at about .915, but very solid.  That’s about the same career level that Dubie is on track for as well.  He’s well into the ‘stable’ part of a typical goalie’s career arc, and he’s doing pretty well.

It might turn out that signing Neuvirth, which will almost certainly be cheaper than Dubnyk, might turn into a bargain bin signing in comparison.  Buy low, etc.

Delusions and Conclusions

OK, let’s be careful before we draw too many conclusions: small samples and selection bias, right?

So I won’t put forward that these conclusions are airtight.  They are indicative rather than strongly supported.

That said, here is what I would conclude would be a reasonable bet for next year, if it can be made to happen:

1. Keep Ben Scrivens. Doesn’t seem likely we’d get good value in a trade (sell low), and a rebound year is a reasonable expectation.

2. If you sign Cam Talbot, you may have solved your goaltending for the next decade.  But next year might be pretty risky.

3. Neuvirth is the steady eddy bet.


– A Neuvirth/Talbot tandem may solve the Oiler goalie problem next year and the next ten years.  At worst, we have a capable tandem.  But what do you do with Scrivens?  Three-headed monster?

– A Scrivens/Talbot tandem is risky for next year if Talbot has a bad year (good chance) and Scrivens doesn’t rebound (I’d say he will, but continued suckage is certainly possible).

– A Scrivens/Neuvirth tandem is likely to solve most of the problems in the short term, but might not provide top rank goaltending in the long term. That would depend on where Scrivens gets to.

Me?  I’d go get both Neuvirth and Talbot, as long as the cost for either (cap for the former, trade assets for the latter) is not too dear.  If Ben rebounds and it turns into a three-headed monster, I say that’s a good problem to have.

What say you?

P.S. I lied about the food


27 thoughts on “More blogs about goalies and food

  1. It seems to me that we should more or less expect stabilization as a product of the cumulative measure. What does it look like with rolling n-game averages (n=10, 25, 60, for example)?

    Would like to see a different selection/cobort: promising goalies that turned out crummy, or had bad runs along the way. Thats the Talbot risk. Thinking more Steve Mason.

    Liked by 1 person

  2. Yeah, absolutely, cumulative measures are inherently stabilizing. What I was looking more for was the relationship between early returns and long term stabilization, and also for any pattern as to when an initial stabilization occurred that was relatively close to the career stabilization level.

    There’s lots of follow on work that can be done, for sure. I may get to that, but I’m also hoping that others may find it interesting enough to take on the challenge of follow on work (hint 🙂 hint).

    Steve Mason would be interesting, as would some representatives of mediocre goalies who still had reasonably long careers (Raycroft maybe?). I will make all of my spreadsheets freely available if it helps!


  3. I’m really behind getting Talbot if the price is right. Even if he has an up and down season Scrivens should be able to provide reasonable goaltending during rough patches if the D is shored up a bit. Long term, Talbot could be a steal. Great analysis.


    1. Thanks. Agreed, Talbot would be a good ‘un. I was a bit skeptical when Woodguy first pushed him to the top of his list, but he’s been so good that I’m more comfortable that even if he weakens markedly, he’s still going to be outstanding.


  4. Great article!

    I think part of the problem with the analysis of these patterns is the comparison to successful goaltenders, which of course you acknowledge. As you say, if Ben Scrivens is a ‘good goalie’, he should rebound and end up closer or above his 50-game save percentage. However, perhaps the more important question is how many goalies had Ben Scrivens’ 50-game save percentage and DIDN’T end up staying in the league 2 or 3 years out. That may end up being the more relevant question.

    I’m not sure how to get that data set. Is there a way to look at all goalies who played at least 50 NHL games but no more than, say, 150 NHL games? Or maybe 200 as the outer limit? What does the subgroup of goalies that get a foothold in the NHL but can’t hang on look like? How are they different than the subgroup of goalies that hang on and carve out careers? Do they have lower save percentages at 50 games or do they just have larger drop-offs that are unpredictable?

    I am not at all convinced of my ability to crunch numbers in an effective way, but if I knew where to find that data set to manipulate, I could try my hand at it. Actually, even having a list of those goalies that meet that criteria in the salary cap era would probably indicate if the sample size is large enough to warrant exploration.


    1. Hi Wheat, I think you’re asking *exactly* the right question! The counterpoint is how comparable goalies who washed out did, and whether that contradicts any of my initial conclusions. The data to repeat my analysis is all available on war-on-ice.

      To find a sample set of good goalies for comparison, I think a good source might be Rob Vollman’s super spreadsheet of goalies:

      You should be able to find what you need there, though it might take a bit of Excel wrangling. The link text is actually out of date, his spreadsheet covers all the way to 2014.


      1. Almost done my look at all the washed out goalies. Hard to find time with a toddler… who knew? 17 goalies played at least 50 games since 1998 but washed out before 200. Two suffered significant injuries (Dan Blackburn who lost use of his glove hand and Jani Hurme who lost two whole seasons to injury and then went to Europe). Of the remaining 15, only two had 50-game even-strength cumulative save percentages that were in the range of Scrivens, Lack and Talbot (>0.920).

        One: Erik Ersberg. Started off like a house on fire and then started to cool down. His numbers were still pretty good but L.A. already had a goalie and so they tried to send him down to the AHL. He refused to go and so his contract was nulled and he took off for Europe. He’s had a pretty good career in the KHL. There might have been a solid NHL goalie there!

        Two: This one’s going to hurt… Jussi Markkanen. Great looking 50-game Sv% (0.923)… and then crashes and burns like none of the other goalies.

        I can’t figure out how to insert a graph into the comments section like TheGreatMutato did below, but I’ll send the whole data set to you once I clean up a few things (war-on-ice only goes back to 02/03 and I added a couple goalies with seasons before then… but accidentally added regular save percentage pre-2002 instead of just even-strength). That’ll probably get fixed up tomorrow… unless baby doesn’t agree.

        Speaking of which, I should sleep… there’s a non-zero chance the boy’s up in 5 hours… I assume to spite me. However, the data does seem to support your 50 game culumative save percentage as having some value, even in those that end up washing out.


  5. if you look at this from an objective point of view it seems that a scrivens/talbot duo might be the best bet for long term here. Yes we might have a year or two of suck but that will allow us to stockpile a few more assets to develop during cap crunch years. It would also give us Talbots best years during the prime of McDavids Oiler career.


    1. Agreed. A good long term bet. I guess the question is how patient Oiler fans and management are willing to be – will we tolerate another year in the toilet because of goaltending next year?


      1. Connor buys alot of patience especially if we see moves are being made to make us a contender for years instead of short sighted quick fixes


  6. Nice post. I like the idea of keeping Ben. He is no doubt a smart guy and under a new coach with a better defensive system he has a good chance to rebound.


    1. I’m looking forward to TMc’s systems, which pressure puck carriers on the way in and use really effective set play breakouts on the way out – both huge weaknesses (which ought to be strengths) for this Oiler team.


    1. I suspect that any real trade involving McDavid would get Chia Pete tarred and feathered. But in one sense I agree with you – basically the only trade I would consider for McDavid would have to include one of Doughty, OEL, Hedman, Subban, or Karlsson; plus a goalie; plus a top 1C or at least 2C to pull the trigger. Say Doughty, M. Jones, and Kopitar. That’s not as exciting or saleable a team as with McDavid, but it probably wins more Stanley Cups.


  7. Fantastic work. As another commentoer mentioned, I’d love to see this technique applied to Steve Mason – a goalie with a long period of questionable performance who may have (or may not have) finally begun to stabilize.

    Pretty please!


    1. Thanks for doing that, that’s a very interesting chart. First one where the Sv@50 was higher than career (but I suppose in the end, not that surprising, there’s bound to be those, and also still indicative of some pretty promising career trends starting around game 200.


  8. After noticing that I typo’d the formulas for the cumulative sv% in the images I posted above (and didn’t bother to check it) , I found something worth asking about:

    Curious about the method you used to calculate cumulative sv% – I only ask because the even strength numbers seem kind of low. For example, looking at Hank’s cumulative 5on5 sv% on puckalytics gives a career number of 92.93 whereas his chart above implies a number in the low 92s. Did you use any adjusted numbers or get data for 4on4 as well?

    FWIW, here’s what I had for Hank’s chart using the basic 5on5 data from War on Ice. Trend looks the same but ends up shifted on the y a wee bit:



    1. Hmm, I’m not sure, I definitely *intended* to use 5×5. I end up with 92.3% as his career number. I calculated it by cumulatively summing the shots and goals for each game, then calculating sv% as ( shots – goals ) / shots. I pulled up the war-on-ice number and the last game matches (20 shots, 1 goal) my spreadsheet. One thing I did is stick to regular season numbers, is it possible your numbers include playoffs?


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s