Crashing the Crease: A Tale of Two Bryzies

We believe, Bryz.  Well, when you win we do.  Not so much when you lose.

Image Courtesy of Bruce Bennett/Getty Images North America) / Zimbio

Goalies, by and large, are fragile.  As you look at more elite goalies, you see that they’re able to better overcome this inherent fragility, but even the best out there have developed coping mechanisms for the rigors of the position.

“Routines” or “rituals,” as you’ll hear them called, are euphemisms for superstitious routines that let goalies exert a sense of ownership and control over a situation that is essentially dictating their activities.  Patrick Roy talked to his posts and skipped over the red and blue lines on the ice.  Braden Holtby, the Caps’ rookie netminder, was known for a “one-for-you, one-for-me” ritual with his water bottle and the ice.  Even yours truly has to have a Sugar-Free Red Bull and handful of Nerds before a game, although I’m more in line with the worst of the worst.

The point is that there’s no escaping our own fragility.  We simply institute and develop these routines to manage the stress and nerves by introducing a little more regularity.  We carefully tread on that broken glass, but we still get cut now and then.

When you look at how well these methods work, your ultimate goal is to look at performance in the clutch.  Does he stand tall with the extra attacker in the zone and a minute left on the clock?  Can he steal a game seven when the team is nearly spent?  Is he good for one game-changing save on a shot he shouldn’t have had?

After watching a goalie for a season or two, I tend to get that qualitative feel of just how clutch he is.  As I remember the negatives and track how many games he “blew” or recount the bad goals that beat him late in a game, my observations transform into opinions, and before I realize it, I’ve made my judgment.  Just as an example, from years of being a Flyers fan, if I could pick any goalie in his prime to make one save on one shot to win it all, I’m going to pick Martin Brodeur.  I have no stats to support this, but I’ve [painfully] seen him win Cups and raise himself up to a new level in the toughest situations to get it done.  Some among us may pick Dominik Hasek, some Roy.  If we have any Western Canadian readers, I might even see a vote for Kirk McLean (look up “the save” if you’re wondering why).  Since I’ve seen more of Brodeur than those guys, though, I’m likely going to pick him.

The takeaway here is that we don’t typically look at clutch performance in a quantitative fashion.  It’s very feel-based, somewhat anecdotal and, ultimately, innately subjective.  So, if you’ve read my past couple of articles, you may guess where I’m going with this.  In this article, I want to try and start to quantify that hard-to-measure clutchness, and who better to use as a guinea pig than our very own Ilya Bryzgalov.  Because I want to develop this statistic with input from readers, I’ve only looked at his turbulent first half of 2011-2012.  My next article on the topic will look at his second half while incorporating enhancements and refinements to the metrics explained below that were proposed in that time.

And with that, let’s get to brass tacks…

I started the process pretty simply: To measure how clutch Bryz really is, I thought about what I want from a goalie in order to consider him a go-to guy.

The first thing I figured out is that I want his performance to get better as the game goes on.  In my opinion, starting strong is nice, but finishing stronger wins games.  A goalie that will win you key games will start to own those games, even with increased pressure late in the game.

Thus, the first key statistic was pretty straightforward: save percentage as a function of time into the game.  If I see a positive linear trend in save percentage from start to finish, I’m going to say that this guy is more likely to steal me a game than a guy who gets worse by the 55-minute mark.  By breaking down shots, saves and save percentage into two-minute buckets, save percentage can be viewed from the first puck’s drop to the final horn. Why two minutes?  It was arbitrarily selected.  Two minutes is a nice number, and it’s also an interval that one of my biggest goaltending influences, Steve McKichan, has suggested for game management.  The time can really be broken up however works best, and it doesn’t have to be in evenly distributed buckets.  Given that, here’s Bryz’s first-half performance for all games he appeared in.  I didn’t differentiate based on started, finished, pulled or anything like that.  If the Flyers goalie at the time of the shot or goal was Ilya Bryzgalov, it hit his numbers.  You can see the outcome of this approach below:

After a brief analysis, this graph essentially begins to visualize issues I know we’ve all seen.  The first period starts poorly, with a save percentage slightly above 80 percent.  It seems to level out, though, before another decrease at the end of the first.  The second period shows a good deal of variability.  It’s hard to say whether he’s gotten better or worse, but he has gotten more inconsistent, as the standard deviation on his save percentage jumps from an overall value of 8.6% to a second period-specific value of 10.4%.  Going into the ever-important third, it immediately jumps out that the save percentage trend looks worse here.  This is confirmed mathematically, as the third period is his worst one for save percentage through the first 41 games (see below).

Save Percentage, by Period

1st period: 89.4%

2nd period: 90.1%

3rd period: 88.5%

Overtime: 81.8%

Clearly, this is a bad start for Bryz’s proving his “clutchness.”  Due in large part to the relatively poor third period and very poor OT performance, you can see the dotted blue line passing through the data showing that he trends downwards, on the whole, over the course of the game in the first half of the season.

Without doing a detailed statistical  determination of significance, I wanted to next see how much data I had in each bucket.  Was I looking at four or five shots in many instances, or were the shot counts higher?  Fortunately, the sample size seems decently large (not saying statistically significant, just relatively large), as shown below.  The graph below is the same as above, with the addition of shot count in red:

A lot of what I see here is focused on the third.  Again, I’ve not attempted to prove any sort of correlation, but through inspection, there’s a solid trend in the first period: The more shots allowed, the lower the save percentage.  Interestingly, this doesn’t seem to carry as evidently into the second and third.

What does this say to me about clutch capabilities?  Nothing really, but it does make me think that Bryz’s warmups and pre-game prep aren’t potentially where they could be, correctly or incorrectly disregarding the effect of defensive performance on the shot count and quality.  There’s some anecdotal evidence to support this, as we heard from Claude Giroux during this year’s first round series when he said the team basically knows how Bryz is going to play based on his pre-game and warmups.  To evaluate just how stark a difference there is between Bryz’s performance in games the Flyers won and ones they lost, I split out the time-dependent save percentage for each.  I think, to put it bluntly, the graphs speak for themselves.  Let’s look at the games they won, first:

Two things jump out almost immediately to me on inspection: (1) There’s far less variability in the save percentage, and (2) the variability decreases as Bryz progresses into the third.  On top of those two things, you can easily see that, in games they won with Bryz in net, his performance actually trended upwards as the game progressed, although his second period was far and away his best.

Save Percentage in Wins, by Period

1st period: 91.5%

2nd period: 93.4%

3rd period: 91.6%

Overtime: 100.0% (obvious, or else it’d be a loss)

In other words, a good Bryz will only become a better Bryz (an oversimplifcation, but work with me).  Now, with that in mind, let’s contrast it with the games the Flyers lost and Bryz saw ice time:

This graph, also put bluntly, is painful.  The first and third periods are flat-out ugly.  There’s drastic variability in these periods, and poor save percentage performance that progressively declines:

Save Percentage in Losses, by Period

1st period: 85.9%

2nd period: 84.2%

3rd period: 80.6%

Overtime: 71.4%

Adding a rancid cherry on top is that performance in the final two minutes, where his save percentage broached a gut-wrenching 30%, albeit on three shots.

There is, undoubtedly a team performance aspect to this data that needs to be ferreted out.  In the most simplistic fashion I can think of, I would imagine plotting this data against shot differential should highlight how much of an impact that team performance aspect really has, as I’ve noted in previous analysis that the second period had proven to be a very strong period for the Flyers.  Their shot count seemed to get much higher and they generated more offense here.  From that point, though, they trended back down to a more first period-like performance.

Tying it all together, it looks like the mid-to-late first period range is where the game turns one way or another for Bryz.  The performance curves look similar in shape to this point, but when he was on last year, he could more effectively rebound from his notoriously poor starts and right the ship.

Going back to the overall view, the salient points for Bryz that I walked away with were the negatively trending performance overall, the very poor OT performance and the poor start and finish in the first.  Looking at the wins versus losses, it’s as though he was made of two different goalies.  It was a far, far better thing he did in the wins than whatever it was he was doing in the losses (shocking, I know), but this simplistic statement ties all the way back to the main point of mental strength and how well his routine prepares him for the game.

Working on honing this routine, building that comfort level with his surroundings and starting to be able to own the game a little more will go a long way in cutting back on the Jekyll-and-Hyde Bryz we’ve come to love and fear.

  • James Minger

    Brilliant work! I love the idea of being able to metric “clutch” goaltending however, Bryzgalov is a far too complicated first analysis. Bryz may turn out to be a good test representation but I’d be curios to see what Henrik Lundqvist, or Jonathan Quick’s data looks like as a control.

    Also, something that you may consider into your metric is save percentage vs. lead differentials. Meaning that I’m curious to see how winning and/or losing a game affects save percentage through the course of a game.

    Regardless, great read. Really thought provoking stuff.

    • Justin Brennan

      James, thanks! This is a good idea, and I think it ties in well with Kate’s feedback. Game situation does definitely impact attack and shot strategy, and I think this method is too simplistic to currently account for that.

      If I can develop a faster method to pull the data and sort it, I plan on expanding to other goalies. Because of how ridiculous both of those guys were this year, I’d love to see what their numbers look like.

    • Justin Brennan

      James, thanks! This is a good idea, and I think it ties in well with Kate’s feedback. Game situation does definitely impact attack and shot strategy, and I think this method is too simplistic to currently account for that.

      If I can develop a faster method to pull the data and sort it, I plan on expanding to other goalies. Because of how ridiculous both of those guys were this year, I’d love to see what their numbers look like.

  • Kate Rhea

    You’ve left out one important factor: every goalie’s per-period save percentage increases over the course of a win, because the losing team take more and worse SOG in an attempt to make up the difference. You can see this in your own data – in the “Flyers Wins” breakdown graph, there are a ton of shots-against in the third period, as losing teams try to put in dirty goals, but in the “Flyers Losses” graph, there are relatively few shots-against, as the winning teams play more conservatively.

    You might get some useful data out of a league-wide comparison, but even then, the results are skewed by each team’s win-loss record. Interesting idea, if you believe in the whole “clutch” thing, but it needs some work to be statistically useful.

    • Justin Brennan

      Kate, thank you – that’s great feedback, and I think you’re exactly right. There are definitely flaws if I were to try and use this as an end-all, be-all approach. Luckily, this type of venue allows me to air ideas and let the light of day clean them up a little.

      I think a per-game bucketed average save percentage approach will allow for more effective representation of what you’re talking about, as I can work with the shot differential a little more effectively to adjust the save percentage. Since I’m aiming to measure “clutchness,” I can move away from the pure save percentage and start to do a little more with it.

      Any other insight or feedback you have is greatly appreciated.

  • Goalie Guru

    If everyone had listened to me, the Goalie Guru when I stated that goalies are only as good as the teams in front of them, then the Flyers coaches, scouts, and GM would know that there was never a goalie problem in Philly, there was a defensive system problem. Bryzgalov is not a better goalie than any of the goalies they have had there last ten years. I, the Goalie Guru could have saved them 51,000,000.
    Bryzgalov was good on Phoenix because the defensive systems are better there than Philly. That’s all! That’s it! What a waste of money! Goalies are not what they appear.

    Interesting how the new Phoenix goalie got better. What a coincidence!

  • Pingback: Flyers Faithful: Crashing the Crease: "Clutchness," Revisited

  • Pingback: Flyers Faithful: Crashing the Crease: Shortened Season, Increased Expectations