What Do We Need to Measure Goalie Puck Handling Outcomes

How to Prove or Refute the “Tend the Goal” Refrain

If you’re a goaltender… TEND THE GOAL!
—Steve Dangle

Wednesday, November 27th, 2024 Charlie Lindgren (left) tries to clear the puck in the 3rd period, but puts it into his own net giving Tampa a 4-3 lead. Puck-handling mishaps make fans feel exasperated and can lead to goals against. But how would measure the impact of all goalie puck-handling — good or bad — and determine if the benefits of all successful puck-handling is worth the occasional hiccup. Washington would win this game 5-4

As I understand, a lot of people like Steve Dangle. He does a lot of hockey media outside of traditional media spaces. He does playoff game livestreams, he has a good YouTube presence, and —most importantly— he screams at the Leafs. This is all good. It’s not how I like to consume my sports¹, but it’s good for hockey for this content to exist and be successful.

From the my outsider perspective, I see Mr. Dangle referenced in relation to a few things: (1) the Leafs losing in the playoffs against the Bruins, (2) the Leafs’ 4-1 loss to Carolina, in which eBUG David Ayres famously was the goalie of record, and (3) his oft-repeated statement “if you’re a goaltender, tend the goal” uttered in reaction to goalie turnovers that lead to goals or other puck-handling mishaps.

I will be honest, I think this statement is lazy. Obviously, you can say that Steve Dangle is an entertainer primarily and thus has different priorities here than I do. But I’ve seen the statement echoed in many places, and at this point I’m not too sure how much of it is facetiousness. I find this statement too totalizing for my tastes.

We can’t just measure against every time a goalie leaves the net: we need to measure against every time a contested (or loose) puck enters the D-zone because that paints a fuller picture of the risk selection than all the times a goalie feels it’s safe enough to play it.

How Do We Measure This?

So what’s the antidote? Well, there’s no actual way to measure the impact of goalies playing the puck right now, so, we can’t really do a cost-benefit… right away, at least. We could do a basic rate of frequency: how many turnovers/goals are there per each opportunity for that to happen? The answer would be ‘not many’, of course (I don’t have actual data at the moment), but that’s unlikely to convince people of anything as they’d point out that just because a really bad outcome doesn’t happen a lot, doesn’t mean it makes sense to risk that outcome happening. After all, people don’t get struck by lightning frequently, but that doesn’t mean they don’t prefer to be indoors during a thunderstorm.

But we can imagine that there are a very general set of outcomes that result from a goalie playing the puck². Since goalies generally don’t skate with the puck enough to start a rush, this limits our typical outcomes to 2 primary outcomes: the play continues after the pass or the play is dead.

You might not realize it, but play being dead is always a possibility — it can be shot into the bench, into the netting, over the glass, sent for icing or even an offside. The other play-dead option is covering the puck — which does happen frequently, but is also technically spatially confined to the crease, and sometimes, when improperly done, results in a penalty.

The play continuing is a far more frequent outcome and gives us 6 more possible general outcomes. Any time anyone makes a pass outside the offensive zone, 3 things can typically happen: (1) the pass is received by a teammate and possession is maintained; (2) the pass is not received by a teammate and leads to a puck battle somewhere else on the ice; (3) the pass is not received by a teammate and the puck is possessed by the other team. For sake of simplicity, we’ll consider a pass successful if possession is maintained for 1-1.5 seconds after reception or until the next “event” (e.g. pass attempt) occurs. But we don’t have to get bogged down in the definitions because we haven’t started measuring yet.

Next thing to consider is whether or not the puck leaves the defensive zone. For our purposes, the neutral and the offensive zones are equally good (referred to here as “OutZone”) because offsides prevent direct passes to someone already in the offensive zone. The pass reception must be carried into the O-zone by the receiver.

Diagram of Goalie puck-handling outcomes. Actual outcomes are at the bottommost branches of the tree.

This is great! But it’s only really useful if we know how “good” each outcome is. Since some outcomes are better than others we can rank these by preference. Remember: the alligator (> & <) wants to eat the better result.

The general hierarchy is as follows:

OutZ-maintained > InZ-maintained > OutZ-contested > Out of play OR covered, no penalty > offside³ > icing > OutZ-turnover > InZ-contested > Out of play, penalty > InZ-turnover

Within these outcomes, looking at the categories used to derive this list, there are some generalities:

• Maintained possession > Contested possession > Turnover possession

• Out zone > In zone

• |Continued| > play dead

Obviously, it’s better to have the puck than not have the puck just as it’s better to have the puck out of your zone than in. But what’s interesting here is the comparison between the play continuing compared to the play being dead. The play continuing doesn’t result in universally preferable outcomes: it results in outcomes that are either better or worse than dead-play outcomes.

Now these are simply general categories of plays that can look really different even within the same category, so how do we know how good an individual play is without comparing it. Typically, xGoals would be a good way of determining this, but there’s one major problem: xGoals doesn’t measure passes, it measures unblocked shot attempts.

So this is a dead end? Well, not quite. xGoals has a variable in its formula for where a shot takes place on the rink —weighing some locations as inherently better (all other factors equal) than others. Theoretically, this could be extended to the entire rink. Now, the value of these positions are measured relative to their value as shooting locations. This is a bit of an issue, because being directly behind the net in the offensive zone is not a bad spot to be offensively, but it is if you’re trying to score by yourself. What this tells us is that the value of a location on the ice has different value for passing vis-à-vis shooting. But this is fine because we don’t actually care about shooting right now: remember, goalies can’t pass directly to the offensive zone!

We still have a spark of an idea with location, so what about possession? This is the most intuitive part of this exercise. Using xGoals as a proxy measurement, I’m going to assume that puck possession is always > 0 xG, whereas the opposition possessing the puck is always < 0 xG. Again, actual expected goals are measured only on unblocked shot attempts and typically range from 0.01 to just approaching 1. But if we imagine mere puck possession carrying some — admittedly very small (say, 0.0001) — equivalent value, it’s possible to better understand these outcomes of a goalie playing the puck.

Now, contested possession poses a problem, but this is where we reintroduce the value of location. A loose puck might carry no intrinsic value to either team, since it is not actually possessed, but its inherent value is derived from the value differential its position on the ice has to each team.

Contested possession value = TeamAPossValue - TeamBPossValue for the same position of the puck.

In other words, a loose puck in the D-zone is bad because its position is more conducive to the other team scoring than it is to your team scoring immediately. This, of course, assumes that all loose pucks are 50/50, but I think that’s just an assumption we’ll have to take in the teeth.

Some of the dead-play outcomes we mentioned above have byproducts we can’t necessarily measure the same way. Penalties have a prolonged consequence, but aren’t immediately worse than turning the puck over in your own end. Icings, meanwhile, can be dangerous as well, since you’re unable to change, but also have a higher potential of innocuousness. There’s also a miscellaneous effect in any outcome that results in out-of-zone position, because any attempt by the other team to re-enter the zone will have to be on-side. Of course, sometimes the other team will already be on-side, but I’ll consider it anyways. The last positive miscellaneous effect is that any non-icing face-off allows the both teams to change. The option of changing, especially for a D-zone face-off, is a positive in my mind but not a big one.

Now, taking everything we’ve learned so far, before we dive into exact numbers, I’m going to rate each outcome mentioned above on a +, =, or - scale based on (in order of appearance) (1) possession, (2) location, (3) miscellaneous by-products (ie what we just talked about last paragraph), and (4), if the play continues (C) or is dead (D). You might noticed that my nomenclature for live plays is ordered position-possession, but my scale is (possession, position, miscellaneous) and that’s because I believe possession is more determinant of a good outcome than position, whereas the nomenclature just feels more natural to me.

• OutZ-maintained (+, +, +, C)

• InZ-maintained (+, -, =, C)

• OutZ-contested (=, +, +, C)

• OutOfPlay/Covered, no penalty; Offside (=, -, +, D)

• Icing: (=, -, -, D)

• OutZ-turnover (-, +, +, C)

• InZ-contested (=, -, -, C)

• OutOfPlay/Covered: penalty (=, -, - -, D)

• InZ-turnover: (-, -, -, C)

You might notice a couple of things looking at this and have a few questions, such as: why are InZ-contested and OutOfPlay/Covered, penalty so low if their possession is technically neutral? Well, remember that when possession is neutral, the value becomes determined by the difference between value of possessing the puck vs not possessing the puck for a certain location. Hopefully, that makes sense. But what might not make sense is why InZ-contested is considered a worse outcome than icing. For that, I’d remind you that in terms of play status |continue| > dead, meaning that continued plays will generally be judged as both more beneficial and more dangerous than dead plays. Also, these are general theoretical estimates. Empirical data might cause rerankings. I tried to address the most glaring oddities within my model here, but if you have any further questions PLEASE ASK! I’d be happy to discuss.

Now, I’ve being talking a lot in vague terms of positives, negatives, and neutrals, so let me put down an actual number. I’d estimate that true neutral (i.e. 0 equivalent xGoal value) is somewhere in the middle of OutZ-contested. Wait, but that’s the third-best outcome! So if so many of these outcomes are negative, albeit only slightly for most, it raises the question stated in the header below:

Why do goalies play the puck at all?

If a loose puck in the D-zone (InZ-Contested) has outright — albeit slightly — negative value, then we can understand goalies playing the puck as trying to flip that value to positive by regaining possession, which is why most times goalies will just set up the puck for their defensemen. That is why goalies go out to play the puck. Goalies can’t live in fear of “what if I mishandle this puck” because at the NHL level, everyone has to — to some degree — believe in their ability to execute. This concept is not always optimally understood, and I’m sure my friend

Robbie Marriage

could regale you with many an instance of football coaches cowardly calling plays.

But for a hockey example, I think pulling the goalie makes an interesting analogy. I was talking with someone recently about why a team might pull the goalie with around 5 minutes left and down 3 goals. Their argument was that 5 minutes left is a lot of time to mess up and turnover the puck, wouldn’t it be better to try your luck making up that first goal 5v5 rather than failing and giving up a goal to make a low-percentage comeback even less likely? My argument was something like well, there are a set of conditions where you could be comfortable taking that risk early. If your best players are rested and you have control in the O-zone, then yes, you need to believe that your best players can maintain and handle the puck and generate scoring chances without surrendering a shot on goal. If those conditions are only met with 2 minutes left, you might not bother pulling the goalie unless you can cut the lead to two. But none of this is predicated on the fear of allowing a goal once you have possession because losing is a worse — and more likely — outcome.

Every time a goalie leaves the net, it’s a risk — I’m not going to deny that: shots on goal against an empty net are, by definition, goals — and the further a goalie goes, the riskier it is. But for playing the puck, we can’t just measure against every time a goalie leaves the net: we need to measure against every time a contested puck enters the D-zone because that paints a fuller picture of the risk selection than all the times a goalie feels it’s safe enough to play it.

Consider how frequently pucks are dumped-in in the NHL. Consider now how frequently goalies go out to play the puck. Goalie puck-handling mishaps are not as rare as, say, goalie goals, but are still relatively rare —especially compared to dump-ins or other puck-playing opportunities that lead to these mishaps.

So, essentially, goalies aren’t trying to make extraordinarily high-value play even though their mistakes are high-cost. But we aren’t talking about comparing them to zero, but instead to whatever value a contested puck has. Let’s talk about dump-ins.

Dump-Ins: Punting For Offence

So why would someone dump the puck in under this model if puck possession has been judged more valuable. We’ll look at this as a tactic rather than a strategy (ie, a contextual decision by the player, not part of a coach’s game plan). A main reason is that you are past your red line and you’ve judged in more beneficial to chase down the puck in better ice rather than carry it there and risk turning it over in worse ice. The decision to dump wouldn’t be measured in a puck-playing model, because it’s irrelevant: what’s important would be the value of the resulting loose puck.

But dump-ins rarely stick the landing — and even less so if they’re rims. So, how do we measure the value of a continuously moving puck. The way we account for this is to generalize pucks by sub-zones. We also say that all pucks shot in along the boards around to the other side have a value of x based on all the ice from the edge of the trapezoid to the goal line and above the goal line outside the face-off dots. That’s the value a goalie is trying to beat when they go out to stop the puck behind the net. A lightly flipped puck, meanwhile, can be determined to have a more simple measurement: where it would end up (usually roughly along the boards). Typically a goalie would intercept this puck before it’d reach the boards.

To reiterate, it’s not the neutral xGoal value the goalie tries to beat, it’s that loose puck in the D-zone value — which is, again, always negative — that they’re trying to improve on, so we need to measure based on that. If we control for strength, power play puck-handling should prioritize damage while PK puck-handling should deprioritize possession maintenance.

But now that we’ve established enough principles and gotten some rough estimates, I want to talk about why I think measuring goalie puck-handling is actually a super important goal.

Why Should We Measure This?

I’ll be totally honest, in post-lockout world, puck-handling is the best it has ever been in terms of talent, but it’s not really a needle-mover in terms of assessing a goalie’s value. In my article on the Theoretical Soundness of Rick DiPietro’s Contract, you might recall me citing Rick DiPietro’s puck-handling as one of the tools that made him a can’t-miss draftee. DiPietro was drafted in 2000 during the dead puck era. In his brilliant video essay on the trap, Hockey YouTuber Pinholes Graham — whom I’m frankly shocked I haven’t mentioned sooner on here — breaks down what made the trap a perfect storm of play-driving stagnation in the dead puck era. The section linked is now required viewing, and the rest of the video is strongly recommended viewing. Graham underscores the effect of the two-line pass rule choking up a lot of the rush, but also on the ability of goalies to negate dump-ins being major contributing factors to the low-event scoring environment.

It’s not so anymore. Rush chances are as plentiful and dangerous as ever, but dump-ins are also viable, with the 2024 cup-winning Panthers finding a lot of success using their hard forecheck to create offence. Graham cites the removal of the two-line pass rule (absolutely) and the introduction of the trapezoid (not so sure about, personally) as reasons for why this is now the case.

But even still, perhaps due to the change in environment, puck-handling in the 2020s isn’t as much of a value-add as it previously was. And perhaps that’s due to its ubiquity. Even though one of my takeaways from the early 4-Nations games was “wow! Jordan Binnington can really handle the puck” that wasn’t nearly as valuable as him doing his primary job — stopping the puck — very well.

So why measure this at all? Well, I did say going into this that it was to prove that puck-playing impacts can make up for puck-playing mishaps which are inevitably all over the highlights. And those do leave impressions. But how do we know if those impressions are correct? We don’t because our impressions are only what we notice and what we notice is definitionally what we deem to be note-worthy.

It’s not noteworthy when a goalie goes out to play the puck on the powerplay, because typically those pucks are uncontested. The hand-off is uneventful. In fact, on the powerplay, the biggest impact a goalie has is typically deterring a whole-sale change. But when a goalie makes a high-risk play on a puck — usually high-risk due to an incoming forechecker — it’s much more noteworthy, especially when things go awry. But as much as fans may groan about this, this is the NHL and just as goalies — or anyone in this league — can’t take risks while doubting their ability to execute those risks, so this is just an attempt to measure the benefit against the catastrophe.

1. Full disclosure, (1) I am a weirdo nerd and (2) as a student, I don’t actually have time to watch that much hockey — nor money for the requisite cable subscription — and used radio to listen to playoff games last year.

2. Spoiler Alert: none of them are goalie goals! I don’t care about goalie goals. It’s something that only happens when extra-attacker teams are careless with puck possession.

3. Since any offside off a goalie pass would certainly be called intentional, thus resulting in an own-zone face-off, think of an offside in this scenario as functionally equivalent to an icing, but with the added benefit of being allowed to change.