For most of the time I’ve been writing here and tracking passes and shots generated from passes, it’s been about how teams generate offense. Recently I gave you a preview of the future of these metrics as they relate to on-ice offensive and defensive performance. However, I realized I was leaving out the goalies. That changes today.
Before I get into it, I want to take some time to highlight some existing research on goalie analysis that I found to be both relevant and intriguing. @garik16 writes at Hockey-Graphs and posted on goalie projection using "Marcels," a predictive tool that weights a goalie’s more recent seasons heavier than those from several years ago. It also adjusts for aging when predicting several seasons.
Chris Boyle wrote on a goalie’s workload and its impact on save percentage. He breaks down Henrik Lundquist’s performance when facing fewer than twenty shots and more than forty.
Greg Balloch suggest we need to scrap Goals Against Average as a stat (rightly so) and offers a detailed summary of its faults. He goes on to discuss some of the better metrics when anaylzing goalies: Save Percentage, Even Strength Save Percentage, Adjusted Save Percentage, and Goals Saved Above Average. It's a solid resource if you're curious on goalie analysis.
Robb Vollman ushered in Quality Starts, another stat attempting to provide some context involving a goaltender's save percentage as it relates to the rest of the league.
Finally, Steve Valiquette, who coined the term "Royal Road", has been doing the best analysis of the types of shots specific to how they were created. He labels them as green and red shots in terms of not only where they were taken from, but also how they arrived there (e.g. pass over the Royal Road or a player moving across it). Boyle has written a detailed article for those that are unsure of what constitutes a Royal Road event.
Compared to the depth of analysis we can perform on skaters, we lack that level of depth when analyzing goalies. Apart from Boyle and Valiquette’s cited works above, your best bet is to seek out a goalie’s adjusted save percentage on War on Ice to account for how many shots a goalie faces from specific zones on the ice—you can also view these zonal breakdowns via War on Ice’s hextally charts and I recommend you do.
However, adjusting for shot location, while important, doesn’t explain anything about how the shot was created. We don’t know which goalies are forced into making tougher saves by virtue of the pre-shot movement. Inspired by Valiquette's work, I wanted to take a look at our passing data and see what, if any, benefit there would be in using it to analyze goalie performance. After all, If teams score more often from passes, then the opposite side of that is they are more difficult to stop for goalies. Why? The more you make a goalie work to track the puck, the greater your advantage over him. It's what I call "shot sequencing."
I talk a lot about sequenced events. What do I mean by this? What I mean is that how we track the games tells us the story of the shot attempt before the goalie is required to make a save. We know if there’s one or two passes that occurred prior to the shot attempt. We know whether the passes were made in transition or if they originated in the offensive zone. We also know if the pass crossed the Royal Road, was sent into the home plate area for a scoring chance, or was a less threatening pass on the outskirts of the home plate area. Finally, we know whether the shot was a one-timer or not. Once you go through the permutations, you come away with twenty-six different sequences that can occur before the shot is attempted. Which ones are more dangerous? Well, that’s what we’re going to discuss in this article.
Terms: Most of the acronyms I use are self-explanatory and this community is likely familiar with things like A1 and A2, SC and OZ, and knows what transition means. In order to ensure you’re following along, listed below are some examples of they appear as sequences with a few new terms:
A2 Trans. A1 Trans.: This sequence would be both secondary and primary passers completing passes outside of the offensive zone and the recipient of the primary pass skating into the zone and shooting the puck.
A2 Trans. A2 Not Trans. RR/SC/1T: This sequence begins with a secondary pass in transition followed by a primary pass within the offensive zone that goes over the Royal Road line, remains in the home plate area, and the recipient one-times the puck on goal.
A1 Not Trans. OZ: This sequence involves only one passer and the pass is outside of the home plate area, does not cross the Royal Road, and the recipient shoots, but it is not a one-timer.
You’ll see the various mix-and-match sequences, but these are a few examples.
Note: Similar to the article on on-ice passing metrics, this is something rather new and so I don’t have a lot of data on goalies just yet. I will post comparisons as I gather more, but I wanted to at least introduce this concept to the community and, hopefully, get some feedback on where to take it from here.
Sequences: As you see below, I’ve organized the various events so you’re aware of each sequence. The columns to the right denote Cory Schneider’s saves made, goals allowed, and save percentage for each sequence from the Devils’ visit to the Los Angeles Kings on January 14th through their March 3rd game against the Nashville Predators. We’ll begin with shots generated in transition against him. All data is from 5v5 situations.
So, in this sample size, we see Schneider has faced eighty-six shots in transition and conceded a single goal. The Devils typically give up more shots in transition than they generate themselves, so this is not surprising. Eventually, I’ll compare Schneider to the goalies of the other teams we’re tracking, but my guess is that this is a high number.
Now, we’ll look at those sequences that begin with a secondary pass in transition, but the primary passer advances into the offensive zone prior to passing it again. The most common situation is a pass completed prior to a shot from within the general "OZ" offensive zone area. This would be any pass completed anywhere in the offensive zone except for the scoring chance area. The fact that it is not a SC or RR (Royal Road) event helps to identify where on the ice it could have occurred: shot from the above the faceoff circles or a shot from outside the faceoff dots are the two most likely locations. Unsurprisingly, Schneider has saved all fourteen of these.
Schneider saved four out of five that resulted in a SC and one out of three that were the result of a Royal Road, Scoring Chance, and one-timer (1T). You’ll see that by requiring the goalie to move laterally (RR), coupled with the proximity to goal (SC), as well as the speed from pass to shot (1T), it becomes exponentially more difficult for Schneider to save these types of shots. More on this later.
With these sequences, both passers are inside the offensive zone. This will capture cycles in the corner, back to the point, cross-ice movement from multiple passes, etc. We see that Schneider saved all but one of the twenty-seven shots from outside the scoring chance area. He was victimized once on four shots that were RR/SC/1Ts, similar to his one-in-three rate above when the A2 is in transition. He was also allowed two goals on three shots that were RR/SC, but saved all four of the SC/1Ts, highlighting the importance of the pass crossing the Royal Road and forcing Schneider to move is more important than the SC location or 1T shot type.
Now, these are sequences involving just a single passer within the offensive zone. This is the bulk of the workload Schneider has faced. You’ll see that he failed to stop all five of the A1 RR/SC/1T events, but performed admirably on stand-alone SCs (12/13) and SC/1Ts (11/13). If the opposition completed a Royal Road pass into the Scoring Chance area, but did not one-time their shot, Schneider stopped all five of those. Schneider allowed a single goal from forty-eight "basic" shots.
Overall Goals Allowed by Sequence Type
The charts above take us through all twenty-six sequences and how Schneider has performed. I’ve sorted them by category in the next chart so that you can see how he fared against all Royal Road events, all Scoring Chances, all One-Timers, and so on.
So, we see that Schneider saved more shots from a single pass than from two and was nearly unbeatable in transition and from basic offensive zone shots. As you move from right to left on the chart, you see that of the shots that were one-timed, Schneider saved 82.1%. Any shot sequence involving a SC, Schneider saved 76.5%, and any sequence involving a Royal Road event, he saved 61.5%.
Clearly the faster a shot is taken, the closer it is taken to the goal, and the more the goalie has to move are going to impact their ability to make a save. What also impacts said ability to make a save is the build-up play leading to the shot (number of passers and location of passers). This all makes sense and now we can quantify how a goalie performs in these passing situations.
Comparison with Schneider's Overall Save Percentage
During this same time period (1/14 - 3/3), War on Ice has Schneider's 5v5 Adjusted Save Percentage at 95.5% (Overall 94.7%). Clearly, the man has been on fire since the start of the year. His overall Passing Save Percentage, or PSV%, during this same time frame? 93.6%. The issue with Save Percentage is that it does not take into account anything other than the shot - all shots are equal. Adjusted Save Percentage takes the next step in adjusting for location. Passing Save Percentage is another improvement on taking into account the speed from pass to shot, the build up play, location, and the movement of the puck prior to the shot.
Player’s Impact on Save Percentage
Using this data in conjunction with the on-ice passing metrics I previewed here, we can identify which players are on the ice for these most dangerous sequences, both for and against. If a defenseman is particularly skilled at breaking up or not allowing passes across the Royal Road, then that is a quantifiable key event in limiting quality chances. This would take into account positioning, stick work, and movement to get into passing/shooting lanes. Also, if players are better at limiting one-timers, perhaps they close down their man quicker than others and make the shooter think twice about releasing it. The same goes for players continually out of position, failing to get in proper lanes, or losing coverage as the opposition moves around, both with and without the puck.
Isolating which sequences are most dangerous for the goalie is step one. Step two would be to determine which players are on the ice for more or less of those events than their teammates. We'd then be able to point to that and quantify a player's impact on save percentage simply by on-ice events. That would be a basic approach, but an incredibly useful and intuitive one at that.
Future Comparisons
As I mentioned above, as I get more data and sort through the goalies for the teams we track, I’ll be able to compare how Schneider performs against passes to Roberto Luongo, Henrik Lundqvist, Jaroslav Halak, Braden Holtby, and Corey Crawford. Those are the goalie we’d have the most data on. So look for that in the near future as we gather more data.
Your Thoughts
What do you think of this method of evaluating goalies? Would we be able to properly identify more "athletic" goalies? Or, would this quantify advantages/disadvantages of certain goalie styles? Perhaps it also tells us about the systems these goalies play behind. What do you think? Sound off below!