PDO is regression wearing a jersey
PDO is the hockey stat with the least descriptive name and one of the most useful jobs. It is just a team's shooting percentage plus its save percentage, expressed on a scale around 100 (or 1.000, depending on the convention). The whole appeal is that it tends to drift back toward 100 over time, so a team running hot at 102 is widely flagged as due for a fall and a team stuck at 98 is flagged as due for a bounce. That intuition is mostly correct, which is why PDO became shorthand for luck. But "mostly correct" is doing real work in that sentence, and treating PDO as a pure luck dial is its own kind of error.
Why the two percentages sum to a constant-ish number
The reason PDO hovers near 100 is close to arithmetic. Across a whole league, every shot is both a shot for one team and a shot against another, so league-wide shooting percentage and the complement of league-wide save percentage are two views of the same events. No single team can durably keep both its own shooting percentage and its goalie's save percentage far above the league norm, because the league norm is what everyone is averaged against. Over a full season the spread in team PDO is narrow, and most of the early-season extremes shrink. That compression is not a mysterious force. It is the law of large numbers operating on percentages that are bounded and centered.
What regression to the mean really claims
Here is the part that gets garbled. Regression to the mean does not say a hot team will get cold to balance the books. It says future results are a blend of past results and the underlying true rate, weighted by how much signal the sample carries. Shooting and save percentage over twenty games carry little signal and a lot of noise, so the best forecast of the next sixty games leans heavily on the league baseline rather than the hot start. The hot team is not owed a cold streak. Its future is simply expected to look more average than its past, because the past was mostly noise. PDO is a way of measuring how much of a team's record is sitting on that noisy foundation.
The shot-quality crack in the luck story
The clean "PDO is luck" story assumes every team shoots and saves at the same true rate, so all deviation is variance. That assumption is wrong at the edges. A team that systematically generates higher-danger chances — more shots from the slot, more off the rush, fewer from the perimeter — has a genuinely higher true shooting percentage, and it can sustain a PDO modestly above 100. The shot-quality work that grew out of the Corsi and Fenwick era is precisely about separating the volume of attempts from the quality of them, and quality is the channel through which real talent leaks into PDO. Expected-goals models exist largely to estimate that true rate so you can tell a deserved 101 from a lucky one.
Elite goaltending and elite finishing are real
The other crack is talent at the position level. A genuinely elite goaltender raises a team's save percentage above the league mean in a way that is not noise and does not fully regress, which is part of why goalies are so hard to evaluate and so valuable when you find a real one. A roster stacked with true finishing talent can hold a shooting percentage a tick above average across a season. Neither effect is large — the sustainable edge in PDO is a point or two, not five — but it is real, and a model that assumes every team regresses to exactly 100 will keep predicting a collapse that never comes for the few teams that earned their number.
How to use PDO without being fooled by it
The productive way to read PDO is as a triage tool, not a verdict. A team winning while posting strong shot-share metrics and a PDO near 100 is winning on process and is likely real. A team winning with a mediocre shot share and a PDO of 103 is winning on percentages that probably will not hold, and the record is built on sand. The combination is what carries the information: PDO alone cannot tell you whether a high number is deserved, but PDO paired with expected goals and shot location can. Used that way, PDO is an early-warning light, flagging the standings positions most likely to move once the percentages come back to earth.
The trap of the round trip
The most common misuse is the round-trip prediction: a team sits at 97, the analyst says it is unlucky and due to climb, the team climbs, and the analyst claims a forecast. But some teams are at 97 because they are actually bad — thin finishing talent, weak goaltending, a true rate genuinely below average. For those teams the low PDO is not stored-up good luck waiting to be released; it is a fair reflection of the roster. Distinguishing the unlucky-good team from the accurately-bad team is the entire skill, and PDO by itself does not do it. The number tells you variance is present. It does not tell you which direction the truth lies.
Individual PDO is even noisier
The team-level number is shaky enough over twenty games; the on-ice version for a single skater is shakier still. A forward's on-ice PDO — the shooting and save percentages the team posts while he is on the ice — gets cited to argue a player is driving results or being dragged down by bad luck, but the samples are tiny. A skater might be on the ice for a few hundred shot events across a half-season, and percentages built on a few hundred events swing wildly. A bottom-six grinder can post a gaudy on-ice PDO for two months purely because the goalie stood on his head during his shifts, and an elite playmaker can post a dismal one because the finishing around him went cold. Neither tells you much about the player.
This is the level at which PDO gets most abused, because it is the level at which the narrative is most tempting. A breakout scorer riding a high personal shooting percentage looks like a new star until the percentage normalizes and the goals dry up, and a slumping veteran with a brutal on-ice save percentage looks finished until his netminders stop letting in soft ones. Reading individual PDO without first asking how many events it rests on is the fastest way to mistake a goaltender's hot stretch for a skater's breakout, or a cold one for a decline.
The honest read
PDO is one of the most honest stats in hockey precisely because it admits what it is: a measure of how far a team's percentages have wandered from the league mean. Read it as a flag for unsustainability, not as a law guaranteeing a reversal, and never read it alone. Pair it with shot share and expected goals to decide whether a high number was earned through chance quality and elite talent or simply borrowed from variance. Most extreme PDOs do regress, which is why the stat works. The ones that do not regress belong to the teams that found a real edge in finishing or in goal, and telling those apart is the whole game. PDO is regression wearing a jersey — useful, honest, and dangerous only when you mistake the jersey for a guarantee.