Penalty kill percentage is really a goaltending stat
Penalty kill percentage is the number every broadcast reaches for when a team goes shorthanded: the share of opposing power plays a club prevents from scoring. A kill rate in the mid-80s is treated as proof of a disciplined, well-drilled unit — aggressive forwards pressuring the points, defensemen boxing out the slot, a coaching staff that has the structure humming. It reads like a clean verdict on how well four skaters defend a man down. But a large part of what PK% actually measures is standing in the crease, not skating in the box. The goaltender’s save percentage while shorthanded drives the kill rate as much as the skaters’ structure does, and because that save rate bounces around from year to year, so does the penalty kill percentage teams are praised and blamed for.
What PK% actually records
Penalty kill percentage is simply the percentage of times a team kills off an opponent’s power play without allowing a goal. It is an outcome number, and every shorthanded goal against is a joint production: the skaters determine how many shots and how many dangerous chances the power play generates, and the goaltender determines how many of those chances actually beat him. PK% bundles both halves into a single figure and attaches it to the unit, but the two halves are not equally in the skaters’ hands. A penalty kill can defend beautifully — conceding few shots and fewer genuine looks — and still post a mediocre kill rate if the goalie is not stopping the pucks that do arrive. The number reports the result and stays silent on which layer of the defense produced it.
The goalie carries the number
Shorthanded save percentage is one of the most volatile inputs in team hockey, and it swings the kill rate directly. A team whose goaltender runs hot while down a man for a season will sport an elite PK% that says as much about the crease as the coaching; when that save rate regresses the following year, the kill rate sinks even if the skaters defend identically. This is the fingerprint of a stat riding on goaltending: the unit’s ranking rises and falls with save percentage more than with any change in structure. Penalty kills that finish at the top of the league one season routinely tumble the next, not because the box collapsed but because the goaltending behind it came back to earth. The skaters get the credit and the blame for a number their goalie is quietly steering. The effect is large enough that penalty-kill rankings can be reordered almost entirely by which teams got the best shorthanded goaltending in a given year, which is why year-to-year correlation in team PK% is so much weaker than fans assume. A unit’s reputation as a great penalty kill often outlives the hot goaltending that built it.
Rate stats hide how much work the kill did
Because PK% is a percentage, it erases the volume and the difficulty of what a unit faced. A kill that allows very few shots and suppresses the dangerous chances is doing excellent defensive work even if a couple happen to go in; a kill that surrenders a barrage of high-quality looks can post a shiny rate for a stretch purely because the goalie or good fortune bailed it out. Two penalty kills with the same percentage can be doing completely different jobs in front of the net. The rate tells you how the season happened to end up, not how well the skaters actually defended, and it will mislead anyone who reads a good number as proof of a good process or a bad number as proof of a broken one.
Small samples make it noisier still
Teams kill only a few hundred penalties a season, and the goals against pile up in the low dozens, so PK% sits on a small, noisy sample. A handful of bounces — a deflection, a scramble, a seeing-eye point shot — can move a team several places in the league rankings, and short-run kill rates over a few weeks are close to meaningless. That noise is why penalty kill percentage is one of the least stable team stats from month to month and one of the more regression-prone from year to year. Analysts who track these things have long noted that a team’s shot-suppression on the kill predicts its future goal prevention better than its current kill rate does, precisely because the rate is contaminated by save-percentage swings that do not carry over. The stability lives in the process, not in the percentage the standings display. Reading firm conclusions about a unit’s quality off a rate built on so few events is exactly how analysts end up crediting or firing a penalty-kill structure for what was mostly variance and goaltending.
What the skaters actually control
The repeatable, skater-driven part of penalty killing shows up when you stop looking at goals and start looking at chances. Shots and shot attempts allowed per two minutes shorthanded, and especially expected goals against and high-danger chances against on the kill, describe what the four skaters genuinely control: how much the opponent’s power play is allowed to generate. A unit that consistently suppresses shot volume and dangerous looks is a good penalty kill whether or not the kill rate reflects it in a given season, because it has done the part of the job that persists. Zone clears, faceoff wins shorthanded, and pressure on the entry all feed that suppression. Those are the inputs a coaching staff can actually build and sustain.
How to read a penalty kill
Start by separating the goalie from the skaters. Read PK% as a joint result, then look at shorthanded save percentage to see how much of it the goaltender is carrying, and at chance-suppression metrics to see how the four skaters actually defended. A team with a strong kill rate but a hot shorthanded save percentage and heavy chances allowed is a regression candidate wearing a good number; a team with a middling rate but excellent chance suppression is probably a better penalty kill than its ranking says. Treat short-run kill rates as noise, weight the underlying chance data over full seasons, and you will stop mistaking a good goalie or a lucky sample for a well-built unit.
The honest read
Penalty kill percentage measures something real — the bottom-line result of defending a man down — and killing penalties matters enormously to winning. But it is not the clean verdict on structure it is treated as. A large share of it is the goaltender’s shorthanded save rate, it rides on a small and noisy sample, and it regresses hard when the goaltending behind it does. To judge the skaters, read the chances they allow, not the goals that happen to go in, and keep the goalie’s contribution in a separate column. The box did its job or it didn’t regardless of the final rate — and a lot of the time, the number on the scoreboard belongs to the man in the crease.