Redefining Defensemen based on Transitional Play

Last time, I showed how passing data is a better predictor of future player scoring than existing public metrics. In this piece, I’m going to spend some time talking about how we can more reliably evaluate offensive and defensive contributions from defensemen, which has been difficult due to a lack of data. Not only due to a lack of data, but from a lack of flexibility regarding the identity of the position. Traditionally thought of as existing to defend and “make a good first pass,” I feel this limits the scope of both how we evaluate the position and its responsibilities.

In order to better evaluate defensemen, we need to identify specific metrics that we can tie to future goals. In looking at entry assists (a pass occurring in the neutral or defensive zones that precedes a shot), both for and against, we can quantify how effective that defensemen is at generating offense in transition, as well as suppressing those chances. The importance of those things at the team level is something I’ve previously discussed (transition here and defensive work here with Matt Cane). Once we identify these metrics as having a strong impact on future scoring and goal-suppression, we naturally then reevaluate what the proper roles are for a defensemen, which in turn forces us to reevaluate how we evaluate them.

Personally, I’d like to see us think of them more as fullbacks or midfielders in soccer (this is part of a larger concept of redefining positions and responsibilities, which will be posted in the next month or so, I hope). There are still going to be various types of players based on their individual skill set and team tactics, but supporting play, overlapping on the attack, and distribution are all pillars of what teams should look for. Let’s get to it.

All data is from 5v5 situations and special thanks to Dr. McCurdy for pulling the on-ice player data for me. All non-passing project data is from Corsica.

Accounting for Shot Quality

I used the same sample as in my previous piece. This consisted of 135 defensemen with >400 minutes of recorded data – 200 minutes in each half of the sample. The expected goals formula is derived from the expected primary points formula – a simple weighting of passing sequences based on the likelihood of a goal. Factors included were: Passes back to the point, royal road, behind the net, odd man, center lane, outer lanes, as well shot types like one-timers and rebounds. Passes were not double-counted. All shots not recorded by us, but recorded by the NHL, were weighted together as well.

Below are the R correlations for repeatability and predicting Goals per sixty minutes. There are four passing metrics, Passing Expected Goals For, Entry Assists For, Shot Assists For, and Danger Zone (Royal Road and Behind the Net Shot Assists) For, and three non-passing metrics, Corsica’s Expected Goals For, Goals For, and Shots For. There are a multitude of metrics that can be derived from the passing data, so if you’re curious about repeatability and predictive power of odd man rushes, stretch passes, multiple passing sequences, etc, feel free to reach out and ask.

def_xg

For defensemen, similar to individual production, there isn’t as wide a range in terms of predictive power. The reason for this is that defensemen do not impact offense nearly as much as forwards. So, different metrics need to be developed to measure their impacts, though it may be further away from the actual goal. Similar to their individual production, the rate at which defensemen are on the ice for entries that directly lead to shots is both repeatable and predictive of future scoring. While the expected goals is a few points higher in terms of predictive power, entry assists is a specific skill to measure how effective a defensemen transitions from the defensive zone to the offensive zone. From these correlations, it is an important skill at that.

Now, let’s take a quick look at how predictive entry assists against is of future goals against*.

def_eaa

We don’t see a great correlation (predicting goals against is hard), but we see enough repeatability to assess that preventing entry assists that lead to shots is as much as skill, given this sample, as overall shot suppression. Furthermore, we see a tiny improvement in predicting goals against. More on this below.

So What? 

Why does this matter? What can we gain from this? Lots.

First, it identifies skills and areas of play that bring about positive game outcomes. This builds on existing analytics work that predicts outcomes by explaining what makes them occur. It’s from there that tactical or personnel changes can be made to optimize a team or player’s performance. This is how every video analyst in the NHL should be working: using data to identify optimal plays/players/team strategies and tendencies, reviewing video to come up with ways to assert these tactics upon your opposition, and then preparing a report with how to do that.

Second, not only does accounting for passing in the player environment provide us with improved accuracy in predicting that player’s on-ice performance, but it provides yet more evidence that passing is a skill general managers and coaches should be keying in on when evaluating talent and tactics. Defensemen have been traditionally difficult to evaluate and predict scoring for, but measuring their impact on how teams transition and generate offense is a reliable way to do that. This should lead to smarter decisions in player acquisition.

Third, quantifying the number of entry assists against with a specific defenseman on the ice provides us with the defensive side of this metric: transition defense. Whether through good stick work, good gap control, good puck recovery skills, or a combination of all three, this sample of players highlights the importance of this phase of the game.

Making Smarter Decisions

I have a few things I’m working on concerning on-ice tactics and how best to create advantageous situations, but let’s look at an example of how a front office should use data like this. Basically, this is where we get into the prescriptive analytics, or “what should we do with this information?” You can answer this from a tactical perspective, or from a front office perspective.

I had several issues with the New Jersey Devils letting David Schlemko walk last summer in Free Agency. He signed a modest deal with the San Jose Sharks and the Devils, having already traded away Adam Larsson, then brought in Ben Lovejoy. However, when we look closer at the numbers, we can quickly see it was a mistake to sign Lovejoy when they could have just paid Schlemko.

Of all Devils defensemen who played more than 200 minutes during the 2015 – 2016 season, no one contributed to entry assists more often than Schlemko. Furthermore, no one on the Pittsburgh Penguins contributed less often than Lovejoy. Also, when we control for each Defenseman’s Quality of Teammates and the team’s off-ice entry assists (accounting for how good their defense partner was and the impact the team’s system has on a player’s transition game) Schlemko ends up outperforming Lovejoy overall. The data, after controlling for QoT of defense partner and team effects, is here. You can also find the expected goal and other offensive on-ice passing metrics via that link as well. Forward on-ice data is forthcoming.

This is the type of value that you miss when you analyze players based on raw rates or percentages – be it QoC, shot rates, or shooting and save percentages. You see this sometimes and it’s okay if you’re just having fun, but it’s tantamount to the eye test that is often derided in the analytics community; if you are simply eyeing up several metrics without testing them, you’re just performing another eye test of data.

Video Support

Speaking of eye tests…Naturally, once you have your data (which includes testing to ensure the metric you’re paying for today will produce goals tomorrow), you’ll want to go to the video to further evaluate play.

So, let’s look at some video of Schlemko from last season.

schlemko_regroup

Schlemko attempts to enter the zone, but curls back to regroup when he sees Minnesota is in good position. Notice how he doesn’t just dump it in. He’d rather go back to his partner, settle things down, and give the forwards a few seconds to regroup for another entry. This second entry has given time for forwards to build up speed and win the race to the puck. While a chip-and-chase isn’t an ideal play, Schlemko’s patience manufactures increased odds of recovering the puck.

schlemko_pinch_stretch

This is a continuation of the previous play (patience paid off). Schlemko pinches to keep the puck in the zone, then, when Minnesota finally forces the puck out, calmly threads a stretch pass to Adam Henrique for another entry.

schlemko_breakout

Here is an example of Schlemko leading the breakout. Ideally, you’d want to make the first forechecker miss or find a way to neutralize him. Here, some quick hands by Schlemko allows him to make a pass to Henrique for an entry.

schlemko_breakout_2

And another breakout led by Schlemko. Here, he makes the first man miss and then occupies the second, while still getting a tape-to-tape pass away for another entry. Being able to get past the first forechecker, or at least not have them interrupt the breakout, is hugely important when teams transition.

Besides leading the breakout or making smart decisions as far as when to regroup in the neutral zone,  defensemen bring immense value by simply being an option in transition. I can’t find a link to the Mike Sullivan quote from earlier this season, but he talked about this idea as well: having a fourth player up at the blue when you’re attacking forces the defense to account for them, which creates advantageous situations for your attacking players. Simply being an option in transition makes entering the zone much easier. (Update: s/t to Bill West for providing video on Sullivan’s philosophy. Starts around 4:50).

schlemko_joining_rush_2

If Schlemko doesn’t jump into this play, it’s likely that Jiri Tlusty is challenged much sooner. Trevor van Riemsdyk has to account for Schlemko as an option and can’t over-commit. The result isn’t a great shot, but this likely ends up as a dump-in or no entry at all if Schlemko doesn’t provide that option on the left.

schlemko_joining_rush_3

Schlemko recovers a puck in the neutral zone and then takes a few strides to the center lane to give the forwards time to regroup. One of the really nice things Schlemko does is he follows his passes and fills that space really well. He’s available just as Kyle Palmieri curls back and is looking for a good play to make. With Travis Zajac driving to the net, Shlemko finds some open ice and has plenty of time to put this one in the net.

By jumping into the play, defensemen often can confuse defensive zone coverage as well. See the below clip.

schlemko_joining_rush

Adam Henrique occupies Jake McCabe before making his pass to Mike Cammalleri. Schlemko overlaps on the outside and this forces Tyler Ennis to try and account for him on the fly. When a defender has their back to you, the attacking player always has the advantage and here is no different. Schlemko is able to present an easy target for Cammalleri and put this one home. Knowing when to jump into the play to create advantageous situations is key for any offensive-minded defenseman.

schlemko_rush_rr

This is a similar play to entry with Tlusty, but this time Schlemko quickly pulls up upon entry to give himself a better lane to access his options. He finds Sergey Kalinin for a good chance. Jumping into the play creates the opportunity, putting the brakes on and looking up creates a better one.

Now, these are all offensive plays, but a few moments in there (pinch and regroup against Minnesota, recovering the puck against Detroit) explain why Schlemko would suppress them as well. Here is one more showing how good decision-making and a quick up can shut down an entry opportunity for the opposition.

schlemko_nz_recovery

Here, Schlemko makes a simple pass after sucking in a forechecker. Towards the end of the sequence, he cuts off the Red Wings player’s path to the puck and quickly transitions the other way with an immediate pass. The ability to suppress entry assists – relative to Schlemko’s defense partner and how the Devils’ defensive style of play impacted results -exceeded all other Devils defensemen. In fact, when looking at total residual performance (how much a player outperformed his offensive and defensive expectations, Schlemko has some stellar company).

Conclusions

Accounting for passing and pre-shot movement, given this sample of players, improves upon existing methods for predicting future goals. How teams move the puck and generate offense continues to be something of immense importance when trying to predict scoring. It also provides much more detailed data to launch specific video analysis (i.e. we know Team X is great at passing, here are some metrics that show that, how can we stop them?).

The real takeaway here is not just that passing data can help us make better predictions, but that it reveals which areas of the game are important and attaches a predictive value to that. Specifically, how can we measure a defenseman’s offensive contributions beyond just individual shots and on-ice shot differentials? We hear it all the time that “puck-moving” defensemen are valuable and sought after at the trade deadline and in free agency. Who would have marked David Schlemko as a puck-mover in NHL circles? My guess is very few, (thought some teams are certainly smarter than others).

Be smarter about evaluating defensemen. With proper data and testing, we can identify new metrics that are of vital importance in hockey. This may lead to a preference for a new type of defensemen and unearthing previously miscast players. We need to move beyond the traditional, positional requirements of the defensemen and think of them more as attacking fullbacks and box-to-box midfielders. This will in turn redefine the rest of the team’s positions and responsibilities, ultimately arriving at a much more fluid and supportive style. But, more on that in future posts.

Thanks again to Micah for pulling data, as well as Alex and Matt for feedback while writing this piece.

*This is a quick preview of further defensive work and metrics developed by Matt Cane and myself. The extent of our research will be released in a future article with Matt, which builds off of our previous work here.* 

 

 

Leave a comment