The hockey power play explained: why four cannot cover five

Take a skater off a hockey team and the arithmetic looks mild. Five becomes four. One fifth of the outfield labour has gone, so the team should be about a fifth worse off, and a power play should be a modest edge rather than the most dangerous ninety seconds in the sport.

That is not what happens, and any power play hockey explained in terms of headcount has stopped one level too early. Removing a skater does not make a defence twenty per cent weaker. It breaks the geometry of the defence outright, because defending five attackers is not a job you can do a fifth less well. It is a job that requires a specific number of bodies in specific places, and below that number the shape stops existing.

Here is the mechanism in one line. Defending in your own zone is a trade: somebody pressures the puck, everybody else covers a man or an area. At even strength one defender pressures and four cover four attackers, and the ledger balances. Take one away and one defender still has to pressure the puck, which leaves three to cover four. Somebody is free, always, by construction, and the attacking team gets to choose which of its players that is and where he stands.

Everything else here is a consequence of that sentence. The formations are arguments about where to put the free man. The penalty kill's structures are attempts to make him the least dangerous of the five. The battle at the blue line exists because a kill that never lets the puck into the zone never has to solve the problem at all.

• 2Minutes of advantage from a minor penalty
• 4Minutes from a double minor, served as two consecutive minors
• 5Minutes from a major penalty, served in full whatever the score
• 3Fewest skaters a team may ever have on the ice

Rule-defined values only. No success rate, conversion percentage or scoring figure is asserted here or anywhere in this article.

Why a power play is not a twenty per cent advantage

The offensive zone in a standard rink runs sixty-four feet from the blue line to the goal line, across a sheet eighty-five feet wide. Four skaters cannot cover that. Neither can five, which is the point: five cannot cover it either, and the whole apparatus of defensive hockey exists to make the uncoverable parts unattractive rather than to cover them.

At even strength the trade works because the numbers match. One player goes to the puck. The remaining four account for the remaining four attackers, and although nobody is marking man to man in the football sense, every dangerous area has a body responsible for it. The defence concedes the perimeter and defends the middle, and it can do that because it has enough people to hold a shape while one of them leaves it.

Now take a skater away and run the same trade. One player still has to go to the puck, because a puck carrier given time in the offensive zone will eventually find something. Three are left to account for four. The shape has a hole in it, and the hole is not a small one, because the areas defensive hockey cares about are few and large: the slot, the two seams that feed it, the net front.

A four-man kill has a choice, and both branches lose something.

It can refuse to pressure. Four players stay home, the shape stays intact, and the puck carrier is left alone. That buys structure at the cost of time, and time in the offensive zone is exactly what a power play converts into chances. An unpressured passer at the top of the zone can hold the puck until the shape moves for reasons of its own, and shapes always move eventually.

Or it can pressure, and accept that somebody is now free. This is what most kills do, because the alternative is death by patience. The moment the first killer commits, the attacking team has a spare man, and the entire tactical exercise is deciding in advance where that spare man should be standing so the puck can reach him before the kill recovers.

60%20%20%

• Attackers a four-man kill can account for once one killer goes to the puck3
• Attackers left unaccounted for by construction1
• Attackers the fifth killer would have covered if he were not in the box1

A structural count, not a statistic. One defender must pressure the puck, which leaves three to account for the four attackers away from it.

Show the numbers

Where the free man comes from

Item	Value
Attackers a four-man kill can account for once one killer goes to the puck	3
Attackers left unaccounted for by construction	1
Attackers the fifth killer would have covered if he were not in the box	1

The number of free men is one. What varies is how long he stays free and whether the puck can reach him in that window, and those two things are what formations, passing routes and the one-timer are all about.

Every way a power play in hockey can end

Before the tactics, the clock, because the clock has more rules attached to it than most people watching realise.

A minor penalty is two minutes. The team with the advantage keeps it until the two minutes expire or until it scores, whichever comes first. The scoring exit exists for an obvious reason. Without it a single tripping call could cost three goals, and the punishment would be wildly out of proportion to the offence.

A bench minor works the same way but is served by a player who was on the ice at the time, chosen by the coach. A penalty on the goaltender is served identically, by a teammate, because the goaltender does not go to the box. Too many men on the ice is the most common bench minor and the most avoidable penalty in the sport, which is why coaches take it personally.

A double minor is four minutes, and it is not one four-minute penalty. It is two two-minute penalties served back to back, and the difference matters enormously. If the team on the power play scores during the first minor, only the first minor ends. A fresh two minutes begins from the moment of the goal, and the penalised player stays where he is. So a goal thirty seconds into a double minor leaves two full minutes on the clock rather than three and a half, and a goal with ten seconds left in the first minor barely saves the offending team any time at all. The high stick that draws blood is the classic double minor, and it is a heavier punishment than the arithmetic first suggests.

A major penalty is five minutes and it is served in full. Goals do not shorten it. A team can concede twice inside those five minutes and go straight back to killing, which is why a major is the one call that genuinely changes the shape of a game rather than merely interrupting it. A match penalty also carries five minutes of advantage, served by a teammate, with the offender removed for the rest of the night.

A misconduct is ten minutes and creates no power play at all. The player sits, his team substitutes for him immediately, and the game continues at even strength. People see a player go to the box and assume an advantage. Not for a misconduct.

A penalty shot replaces a power play rather than accompanying it. The referee decides the offence denied a clear scoring chance, and the remedy is one player against one goaltender instead of two minutes of five against four.

Coincidental minors cancel. One player from each side goes off, nobody gets an advantage, and the teams play four against four with a great deal more open ice than usual. This is why a scrum after the whistle that produces matching roughing calls is not a gift to anybody, and why the smart response to a cheap shot is often to skate away.

There is a floor. A team may never have fewer than three skaters on the ice. If a third penalty is called against a team already two men down, the third player goes to the box but his time does not begin until one of the earlier penalties expires. A player whose penalty runs out while his team would otherwise be over the limit stays in the box until the next stoppage. The queue can therefore stretch a long way past the arithmetic of the individual penalties, and the rules governing which penalty starts when are worth reading properly in the full account of how penalties are classified and served.

Two more clock details change how a power play feels. The penalty clock runs only while play is running, so a two-minute minor with three stoppages in it can occupy five minutes of broadcast. And penalty time carries across intermissions and into overtime, which is how three-on-three overtime becomes four-on-three, the most lopsided situation the regular season produces.

The delayed penalty, and the only free attacker in hockey

An infraction does not always stop play. If the offending team does not have the puck when the arm goes up, the referee holds the whistle and lets the play continue. It ends when the offending team gains possession and control, and not a moment before.

The team about to go on the power play therefore has a window in which it cannot concede. Any touch by the other side ends the play, so the goaltender sprints for his bench and a sixth attacker comes over the boards. Six against five, in the offensive zone, with the penalty waiting regardless.

Two things are worth knowing about that window, because both get misunderstood.

The first is that it is not entirely risk-free. The whistle blows when the offending team touches the puck with control, which does not cover a defenceman putting the puck into his own empty net. That goal counts, and it happens often enough to be a recognised category of disaster.

The second is stranger. Score during a delayed minor and the minor is never assessed at all. The goal is the punishment, the offending player does not go to the box, and there is no power play. So the team with a free extra attacker is playing for a goal that will cost it the two minutes it was about to receive, which is a trade almost every coach takes and almost nobody explains. A delayed major is different: the goal counts and the five minutes are still served.

What this produces on the ice is a distinctive kind of possession. Teams on a delayed penalty rarely force anything. They circle, they reset, they wait for a genuinely good look, because a bad shot that misses the net ends the six-on-five for nothing while a patient one might not. The extra man is used to keep the puck rather than to attack with it, which is the opposite of how six attackers behave at the end of a game.

The face-off you get for free

After a penalty is called, the face-off takes place in the offending team's end zone. That is a rule, and it is the most undervalued part of the whole sequence.

It means a power play does not start in the neutral zone with a problem to solve. It starts with the puck on the dot inside the offensive zone, five attackers already in position, four killers already boxed in. Win the draw cleanly and the unit is set up with no entry to make, no blue line to cross, and the full two minutes available for the thing it actually practises.

Lose it and the reverse applies. The kill wins the puck, and here the rulebook helps the short-handed side: a team that is short-handed may shoot the puck the length of the ice without being called for icing. The exemption exists precisely so a kill has a way out that does not depend on skill, and it turns every won defensive-zone draw into a chance to erase ten or fifteen seconds. That asymmetry is one of the few genuine advantages the rules hand the penalised team, and it sits alongside the other blue-line and red-line rules set out in the guide to icing and offside.

Which is why coaches send out a face-off specialist for the first draw of a power play even if he is not part of the unit they want on the ice, and swap him off at the first stoppage. One clean win is worth more than his shift.

1. 1The infractionAn official's arm goes up. If the offending team does not have the puck, play continues under a delayed penalty until they gain possession and control.
2. 2The whistlePlay stops, the penalised player goes to the box, and the penalty clock is set at two minutes.
3. 3The face-offThe draw is taken in the offending team's end zone. The power play begins already inside the offensive zone, five skaters against four.
4. 4Win the drawThe unit sets up immediately in its formation and the two minutes are spent on offence rather than on transport.
5. 5Lose the drawThe kill takes possession and is permitted to shoot the puck the length of the ice, because icing is not called on a short-handed team.
6. 6Retrieval and regroupThe power play collects the cleared puck in its own end, comes back up the ice as a five-man group and looks for a controlled entry.
7. 7The blue lineThe kill stands up at the line. A denied entry means a turnover or an offside stoppage, and the clock has kept running.
8. 8The entryA carry-in or a drop pass gets the puck over the line with possession. The unit sets up.
9. 9The rotationThe puck moves half-wall to point to the far half-wall, forcing four killers to rotate. Each rotation is slower than the pass that caused it.
10. 10The seamThe free man receives in a lane the kill has just vacated, most often the bumper in the middle or the far flank.
11. 11The shotA one-timer through a screen, or a feed to the net front for a tip. If the puck is blocked or frozen, the sequence restarts from a face-off.
12. 12The clearIf the kill wins the puck and gets it out, the power play goes back and does all of it again with less time left.

The rule-governed sequence, step by step. No timings are asserted, only the order of events.

The umbrella, and what it is actually trying to move

The umbrella puts three players high and two low. One sits in the middle near the blue line, two more sit wide of him and slightly lower, and the arc they form gives the shape its name. Below them, one player works the net front and one operates around the goal line.

What the umbrella wants is not a shot from the point, whatever it looks like. It wants the box to expand. Three players spread across the top of the zone are too far apart for two killers to reach, so the box has to stretch upward to contest them. A stretched box is a thinner box, and the moment it stretches, the seam down to the net front opens and the low players become available.

The puck movement across the top of the arc is quick and lateral, and it is designed to arrive faster than four skaters can shuffle. Each pass costs the kill a rotation. Rotations are slower than passes. Do it three times and somebody is a stride out of position, which is all the shape needs.

The umbrella suits a unit with a heavy shot from the middle of the arc and a large body in front of the goaltender, because the shot is not really trying to beat the goaltender cleanly. It is trying to arrive through traffic for a tip or a rebound, with two attackers standing exactly where the rebound goes.

Its weakness is distance. A lot of the offence originates fifty feet from the net, and a kill willing to pressure the top of the arc hard can take the whole arrangement away. Blocked shots from up there are also the most dangerous turnovers in hockey, because the man who took the shot is the furthest player from his own net and the puck is now travelling the other way with nobody in front of it.

The 1-3-1 and the bumper: standing where the box has nobody

The 1-3-1 is the shape that takes the geometry argument seriously.

One player at the top. Three across the middle of the zone, one on each half-wall and one in the middle. One at the net front or behind the goal line. Written out, that is a shape with a player standing in the exact centre of the area a four-man box surrounds.

That middle player is the bumper, and he is the whole idea. A box has four corners and an empty middle. It is a rectangle, and rectangles have interiors. Put a skater in the interior and the kill faces a question with no good answer: collapse inward to take him away and the half-walls are unguarded, stay wide to contest the half-walls and the bumper is standing alone in the slot with his stick on the ice.

The half-wall player is the unit's distributor and usually its best passer. There is a reason he is often a left-shot standing on the right wing wall, or the mirror of that on the other side: it puts the puck on his forehand while he faces the middle of the ice, so every option is a forehand pass and none of them requires him to turn his back on the play. From there he has four routes. Down to the goal line. Inside to the bumper. Back to the point. And across to the far half-wall, which is the pass the kill spends the whole two minutes trying to prevent, because it is the one that produces the shot nobody can get to.

The 1-3-1 is not a shooting formation, and units that treat it as one get frustrated. It is a rotation-forcing formation. Every touch is intended to make four tired players move sideways, and the shot arrives when they have moved sideways once too often. The counter-structures a kill uses against it, and the trade-offs each one accepts, are the subject of the full breakdown of penalty kill systems.

The one-timer is the shot the kill cannot reach in time

A one-timer is a shot taken on the arriving pass, without receiving it first. It is harder than a normal shot, it is less accurate, and units build entire sequences around it anyway. The reason is time, and time is measurable.

Consider a constructed example with round invented numbers. Say the pass has nine metres to travel from one flank to the other, and it moves at twenty-five metres per second. It arrives in a little over a third of a second. Now consider what the goaltender has to do in that third of a second: read the pass, push off the post, cross the crease, stop, and be square and set before the shot leaves the stick. Consider what the nearest killer has to do: recognise the pass, leave the man he was covering, and get a stick or a body into a shooting lane nine metres away.

Neither is possible at that speed, which is the point. The one-timer is not dangerous because the puck is faster. It is dangerous because it removes the interval in which anybody could have prepared for it.

Receive the pass first, settle it, then shoot, and you have handed back somewhere between half a second and a second. That is enough for the goaltender to arrive and set, and enough for the killer to close. The shot becomes more accurate and vastly less dangerous, and that trade is why coaches accept a rate of mishits that would be intolerable anywhere else on the ice.

The cost is real. A one-timer struck a fraction late is a rolling puck into a shin pad. A one-timer missed entirely at the top of the zone is a loose puck behind four attackers, with two defencemen stranded at the blue line. The shot that best exploits the free man is also the shot that most exposes the men behind it.

Which explains the kill's priority order. Killers do not primarily try to block the one-timer. They try to prevent the pass that makes it possible, with sticks in lanes rather than bodies in front of pucks, because a stick laid flat in a seam takes away a shot that has not been struck yet.

The net front is a job, not a place

Standing in front of the goaltender looks like the least skilled assignment on the power play. It is the one most units get wrong.

The screen is the primary task, and it works on a specific mechanism. A goaltender sets his angle from where the puck is, and he cannot set an angle he cannot see. Take away his sight of the release and you have removed the first half of the save, the half where he decides where to be. What is left is a reaction to a puck that appears late, from a position that may already be wrong.

The second task is the tip. A puck deflected a few degrees on its way through changes direction after the goaltender has committed, which is the same trick as the one-timer performed at close range and with even less time available.

The third is the rebound. Every shot through traffic produces one, rebounds tend to land in the slot, and on a power play there is a spare attacker available to be standing exactly there. Rebound goals on the power play are not luck. They are the predictable output of shooting into traffic with more bodies than the other team has.

All of this is contested physically, and the boundary is not always obvious. A screen is legal. Contact that stops the goaltender moving in his crease is not, and the line between them is argued over every week under the goaltender interference rule. The net-front player's night consists of cross-checks in the back, sticks under his arms, and a defenceman trying to move him a metre to the side so his goaltender can see. He loses most of those individual battles and wins the two minutes anyway, because moving a man a metre takes a defender's full attention, and a defender giving one attacker his full attention has stopped covering anybody else.

The entry problem: a kill's best moment is at the blue line

Two minutes is a hundred and twenty seconds. Every one of those seconds spent anywhere other than the offensive zone is gone.

This is the part casual viewers miss entirely, because television shows the set-up and the shot. The best penalty kills in the world do not spend much time defending their own zone. They spend it standing at their own blue line making sure the puck never gets there with control.

A denied entry is enormously expensive for the attacking team. The puck turns over at the line or the play goes offside. Either way the power play has to retreat, collect the puck in its own end, organise five men, and come the length of the ice again. Call that fifteen seconds, as a constructed example. Do it twice and thirty of the hundred and twenty seconds have been spent on transport rather than offence, and the unit has had two-thirds of a power play rather than a whole one. Do it three times and the penalty is essentially over before it started.

Offside is the kill's friend here, and it is the one stoppage in hockey that helps the team being punished. The play resets to a neutral zone face-off with the clock having run in the meantime.

The power play's answers are three, and each is a different bet.

Carry it in. The best outcome and the hardest. One player crosses the line with the puck on his stick and possession is never surrendered. It requires beating a defender who knows exactly where the line is and has three teammates behind him.

The drop pass. The standard power play entry, and the reason those entries look choreographed. The carrier crosses the red line, draws the first killer towards him, and leaves the puck behind for a teammate arriving at speed with the whole width of the ice in front of him. Done well it defeats a stand-up kill, because the man who committed is now behind the puck.

Dump and chase. Shoot it in and race for it. This concedes possession deliberately, and it is worth doing only when the retrieval is genuinely likely, which usually means the kill has nobody back or a particular forward is quicker than the defenceman going to get it. Units that dump the puck in because they cannot think of anything else are handing the kill a free clear.

The trade-offs between the three, and why controlled entries produce so much more than dumps, are worked through in the piece on zone entries.

• Unit A
• Unit B

Retrieval and regroup after a clear15s45s

Neutral zone and the entry attempt10s20s

Set up in formation with possession95s55s

Constructed example with invented round numbers, not measured data. Unit A gains the zone at the first attempt after a lost face-off. Unit B is turned back twice at the blue line, at a cost of 15 seconds each time.

Show the numbers

Worked example: where two invented units spend the same 120 seconds

Item	Unit A	Unit B
Retrieval and regroup after a clear	15s	45s
Neutral zone and the entry attempt	10s	20s
Set up in formation with possession	95s	55s

Ninety-five seconds of organised offence against fifty-five. Same penalty, same personnel, same formation on the whiteboard. The difference was made entirely at a line sixty-four feet from the goal the unit was trying to score in.

The box, the diamond, and the kill that refuses to sit

The box is the default and it is honest about what it is. Four players at the corners of a rectangle in front of the net, collapsing towards the middle when the puck goes low, expanding slightly when it goes high, blocking what they can and conceding the perimeter entirely. The box is built to concede shots from bad places and to lose slowly. Against a patient unit it will eventually lose, and everybody involved knows it. The bet is that two minutes is not long enough.

The diamond is the box's answer to shapes with a man in the middle. One killer high on the point, two at the sides, one low. The low man's job is the bumper, the high man's job is the top of the umbrella, and the diamond exists specifically because the box has an empty interior and the 1-3-1 puts somebody in it. The cost is the flanks. A diamond gives up the half-walls more readily, so shots come from wider angles more often, and the kill accepts that as the better of two bad outcomes.

The wedge plus one splits the difference. Three killers hold a tight triangle around the slot and never leave it, while the fourth chases the puck wherever it goes. The triangle guarantees the middle is never empty, the rover guarantees the puck carrier is never comfortable, and the arrangement gives up the point shot almost entirely.

Then there is the kill that refuses to sit. Aggressive pressure attacks the half-wall the moment the puck arrives, sometimes attacks the entry before the zone is even gained, and tries to force a hurried pass rather than wait for a good one. It concedes the free man by definition, and more of him than a passive box does. What it buys is rhythm, and rhythm is what a set unit runs on. A power play is a sequence of pre-agreed reads, and a player who receives the puck half a second earlier than he expected, with someone closing on him, makes the safe pass rather than the intended one. Enough safe passes and the two minutes evaporate with nothing in them.

The best kills switch between these inside a single penalty, and they do it on cues: after a face-off win, after the power play changes personnel, after a shot is blocked. What none of them are trying to do is defend for a hundred and twenty seconds. They are trying to touch the puck four times and get it out four times, because four clears with fifteen seconds of retrieval each is the whole penalty, and nobody had to make a save.

Short-handed chances are worth more than their frequency suggests

Short-handed goals are rare, so they get treated as noise. They are not noise, for three reasons, and the third is the largest.

The first is the size of the swing. A goal against the run of the situation moves a game further than one goal, because the expectation attached to those two minutes pointed the other way. The team that drew the penalty has gone from favourite to score in this passage of play to a goal behind in it.

The second is that a short-handed goal does not end the penalty. This surprises people every single time. The minor was assessed against the penalised team, and scoring does not discharge it. The team that just conceded remains on the power play with whatever time is left, now trying to get a goal back while a man up, which is a strange and uncomfortable state to be in.

The third is deterrence, and it is worth more than the goals themselves. A power play that has been burned once stops pinching. Its defencemen hold the line rather than jumping into the zone, its half-wall player thinks twice about the cross-ice pass, and the unit becomes measurably more conservative for the rest of the night. A kill that has scored once has bought itself an easier evening, and none of that shows up as a goal against.

The exposure is structural rather than accidental. A power play sets up with both defencemen at the blue line and three forwards below the tops of the circles. That is the most stretched formation any team voluntarily adopts. A blocked shot, a missed one-timer or a poached drop pass at the line all produce the same thing: a loose puck with two attackers going the other way against one retreating defenceman, or against nobody at all. The chances a kill generates are few and they are of a quality nothing else in the game matches, which is exactly the sort of distinction raw goal counts hide and chance-quality measurement exists to expose.

Five on three is a different game, not a bigger power play

Two overlapping penalties produce a two-man advantage, and it is worth understanding that this is a change of kind rather than of degree.

Three skaters cannot form a box. They form a triangle, and a triangle in front of the net leaves two attackers unaccounted for rather than one. Run the same trade as before: one killer goes to the puck, two are left, five attackers minus the carrier is four. Two free men, and the goaltender now faces the cross-ice pass problem with the additional knowledge that there is a second option available if he cheats towards the first.

The kill's method changes accordingly. Clearing the puck stops being realistic, because three skaters cannot both hold the middle and win a puck on the wall. Possession stops being the objective. Getting the puck two hundred feet away by any means available, and then getting it there again, is the objective, and beyond that the kill blocks shots, keeps sticks in seams, and counts.

The attacking shape usually spreads low and wide, because the triangle can rotate once and cannot rotate twice. The puck goes side to side rather than up and down. Every pass across the middle of the ice forces all three killers to move together, and three players moving together leave two men standing still with time.

The duration is usually short, and the arithmetic is worth doing once. A constructed example: a minor is called, and forty seconds later a second minor is called against the same team. The two-man advantage lasts eighty seconds, until the first penalty expires. Then it reverts to five on four for the remaining forty seconds of the second penalty. The five-on-three portion is the overlap, nothing more.

One rule catches people out. If a goal is scored during a two-man advantage created by two minor penalties, the penalty with the least time remaining is the one that ends. So the team on the power play scores, one player returns, and the other stays in the box with his own clock still running. The advantage drops from two men to one rather than disappearing.

The goaltender's problem changes shape, not just size

At even strength a goaltender's job is mostly geometry against a shooter he can see. He takes an angle, controls his depth, and the shot either beats him or it does not.

The power play removes both halves of that. He sees less, because there is a body in front of him whose entire assignment is to remove his sight of the release. And he moves more, because the shape is built on lateral passes and every lateral pass is a demand that he cross the crease and re-set.

That produces the one adjustment you can see from the stands if you know to look for it. Goaltenders play deeper against a power play. Coming out to challenge a shooter on the half-wall cuts down his angle beautifully and leaves nothing whatsoever if the puck goes cross-ice, because a goaltender high in his crease has further to travel to the other post. So he sits back, concedes more of the net to the direct shot, and buys himself a chance of arriving on the pass. It is a deliberate trade of the shot he can see for the shot he cannot.

Rebound control matters more than usual for the same reason it always does, only with an extra attacker in the slot to collect anything he spills. And he has a decision to make on every puck he can reach: freeze it or play it. Freezing stops the clock, which is time the kill no longer has to defend, and it hands the power play a fresh offensive-zone face-off with its best unit already set. Playing it keeps the play alive and might produce a clear. Freezing also lets four exhausted killers change. That trade, made a dozen times a night in a fraction of a second, is one of the least discussed and most consequential judgements in the sport.

Why special teams decide a disproportionate share of tight games

Hockey is a low-scoring game and a large share of its results turn on a single goal. That is the first half of the explanation.

The second half is that the power play is the most productive state in the game per minute played, by a wide margin, and that the number of those minutes is small. A handful of minutes a night carry several times the scoring value of the same number of even-strength minutes. That is concentration of leverage, and concentration of leverage is how small differences in preparation turn into results.

There is a compounding effect underneath it. Penalties are not distributed at random. A team chasing the puck takes more of them, because chasing produces sticks in the wrong places, and a team that is behind chases. So the side already in trouble hands its opponent the highest-value minutes available, which is one of the quieter reasons leads in hockey grow rather than shrink.

Overtime concentrates the same effect further. A penalty in three-on-three overtime produces four on three, on open ice, in sudden death, which is about as close to a decided situation as the sport gets. The way the extra period is played, and why it magnifies every special teams mistake, is set out in the rules governing overtime and the shootout.

In a playoff series the loop tightens again. The same power play unit meets the same kill ten or twelve times over a fortnight, both coaching staffs watch the same video, and the adjustments run back and forth until one of them stops working. That repeated head-to-head is where special teams stop being a percentage in a table and start being the reason a series ends, and it is one of the clearest windows into how the sport is actually coached across the rest of the modern game.

What to watch on the next power play entry

The next time an arm goes up, stop watching the puck and watch the blue line. Five things are visible from the sofa, and together they tell you how the two minutes will go before the unit has taken a shot.

Where the kill stands as the power play comes up the ice. Four killers strung across their own blue line are playing to deny the entry, and the first thirty seconds will be a fight in the neutral zone. Four killers already retreating into a box have conceded the zone and decided to defend it instead. Those are different games, and you can tell which one you are getting before the puck crosses centre.

Whether the entry is carried, dropped or dumped. A carried entry means possession was never surrendered. A drop pass means the unit has a plan and is executing it. A dump-in from outside the line usually means the plan failed twenty feet earlier, and you should start counting the seconds it takes to get the puck back.

Who is standing in the middle of the box. If a player is sitting in the interior of the kill's shape with his stick flat on the ice, the unit is running a 1-3-1 and the whole two minutes will be about whether the half-wall can find him. If nobody is there, watch the top of the zone instead: it is an umbrella, and the shot is coming through traffic.

Which pass the kill is refusing to allow. Watch the sticks rather than the bodies. Four sticks laid flat in the same lane means the kill has decided the cross-ice pass is the thing that beats it, and the power play will be pushed to shoot from where it does not want to shoot.

Whether the goaltender ever has to cross his crease. If he does not, the power play has spent two minutes moving the puck without moving the defence, and the shape has failed however busy it looked. If he does, and he is late, that is the free man arriving, and the free man was always the point.

The unit that has understood all this is not trying to shoot more than the other team. It is trying to make four people stand in five places at once. They cannot, and every good power play is simply a well-organised demonstration of the fact.