NHL Betting Regression & PDO Metrics

Hockey has a luck problem, and PDO is the metric that measures it. In a sport where a fraction of an inch separates a goal from a post-and-out, where a deflection off a shin pad can redirect a harmless shot into the top corner, and where a goaltender can have the game of his life on any given night for no discernible reason, luck plays a larger role in short-term results than most fans and bettors want to admit. PDO quantifies that role, and understanding it is one of the most powerful tools available for predicting which teams are headed for a rise, a fall, or a continuation of their current trajectory.

PDO is calculated by adding a team’s five-on-five shooting percentage to its five-on-five save percentage. Both components are expressed as percentages and summed. A team shooting at 8.5% and saving at 92.0% has a PDO of 100.5. The league average PDO is always 100.0 by definition, because every goal scored by one team is a goal allowed by another, and every save by one goaltender is a missed shot by the opposing team. The system is zero-sum, and that zero-sum property is what makes PDO so useful for regression analysis.

Analyzing NHL Team PDO Statistics

A PDO above 100 means a team is getting above-average results from the combination of its shooting and its goaltending. A PDO below 100 means the opposite. The critical insight is that extreme PDO values are not sustainable. Shooting percentage and save percentage are both heavily influenced by randomness over short samples, and teams at the extremes of the PDO distribution are almost always riding luck that will eventually run out.

A team with a PDO of 103 is experiencing an extraordinary confluence of hot shooting and hot goaltending. Their forwards are converting shots at an unsustainable rate, their goaltender is stopping everything, and the win-loss record looks spectacular. But the historical data is unambiguous: teams with PDOs above 102 regress toward 100 over the following months with near certainty. The shooting percentage cools off, the goaltending normalizes, and the wins become harder to come by. A team that went 15-5 during a PDO spike might go 8-12 over the next twenty games as the luck drains away.

The same logic applies in reverse. A team with a PDO of 97 is suffering from cold shooting and below-average goaltending simultaneously. Their skaters are hitting posts, missing open nets, and generating chances that simply are not going in. Their goaltender is leaking soft goals that he would normally save. The record looks terrible, but the underlying process may be perfectly fine. When the PDO rebounds toward 100, the team’s results will improve even if nothing about their system or personnel changes.

The betting market tends to follow results rather than process. A team on a five-game winning streak fueled by a 104 PDO sees its moneyline shorten. A team on a five-game losing streak with a 96 PDO sees its price drift longer. In both cases, the market is reacting to outcomes that PDO tells us are temporary. The bettor who recognizes these PDO extremes and bets against them is betting on regression, which is one of the most reliable forces in sports statistics.

Shooting Percentage: The Noisier Component

Of the two components of PDO, shooting percentage is the noisier and more prone to regression. The NHL league average for five-on-five shooting percentage sits around 8.0-9.0%, and individual teams rarely sustain rates far above or below that range over a full season. A team shooting at 11% through twenty games is almost certainly benefiting from puck luck — fortunate deflections, bounces off goaltenders into the net, and high-percentage finishes on chances that normally do not convert.

The reason shooting percentage is so volatile is that it depends heavily on factors beyond the shooter’s control. The goaltender’s positioning, the traffic in front of the net, the angle of deflection off a defender’s stick, and the precise location where the puck strikes the post or the mesh are all partially random. Elite shooters do maintain slightly higher shooting percentages than average players, but the gap between the best and worst shooters at five-on-five is much smaller than most fans believe. Team-level shooting percentage, which averages across all skaters, regresses even more strongly toward the mean.

For betting purposes, a team with an abnormally high shooting percentage is a fade candidate. Their goals are built on a foundation that is likely to erode, and the moneyline is pricing in results that will not persist. Conversely, a team with an abnormally low shooting percentage is a buy candidate, because their offense is underproducing relative to the quality of chances they are generating.

Save Percentage: Skill With a Heavy Dose of Noise

Save percentage is less noisy than shooting percentage because goaltending skill is more stable than team-level shooting skill. An elite goaltender can sustain a .925 save percentage over a full season, and the gap between the best and worst goaltenders in the league is meaningful and persistent. However, over shorter samples of ten to twenty games, save percentage still fluctuates significantly due to shot quality variance, lucky or unlucky bounces, and the inherent randomness of individual shot outcomes.

The distinction between skill and luck in save percentage is captured by goals saved above expected, or GSAx. A goaltender with a high save percentage but a mediocre GSAx is benefiting from facing a low volume of high-danger shots rather than from personal excellence. His save percentage will regress if the shot quality against increases. A goaltender with a mediocre save percentage but a strong GSAx is being let down by his defense, which is allowing too many high-danger chances. His save percentage is suppressed by the difficulty of the shots he faces, and a defensive improvement would produce better raw numbers without any change in the goaltender’s own performance.

For PDO analysis, the practical takeaway is that save percentage regresses less aggressively than shooting percentage but still regresses. A team whose PDO is inflated primarily by goaltending — say, a PDO of 102 with a league-average shooting percentage and an elite save percentage — may be more sustainable than a team whose PDO is inflated primarily by shooting luck. The former scenario can persist if the goaltender is genuinely elite. The latter scenario is almost always temporary.

Applying PDO to Betting

The most direct application of PDO for betting is identifying regression candidates on both sides of the spectrum. At any point during the season, pull up the league’s PDO rankings sorted by five-on-five data. The teams at the top with PDOs above 102 are overperforming their process and are candidates to fade on the moneyline and the puck line. The teams at the bottom with PDOs below 98 are underperforming their process and are candidates to back at longer odds than their true quality warrants.

The timing of regression is the difficult part. PDO does not tell you when a team will regress, only that it will. A team can ride a 103 PDO for a month before the correction begins, and during that month, they will keep winning and your regression bets will keep losing. This is why PDO is best used as a medium-term indicator rather than a game-by-game tool. Over a 20-game window, the regression signal is strong. Over a single game, the noise dominates.

Combining PDO with possession metrics like Corsi produces the sharpest regression signals. A team with a high Corsi percentage and a low PDO is the strongest buy candidate in hockey: they are controlling play at five-on-five but getting killed by shooting and goaltending luck. When the luck normalizes, the strong possession translates directly to wins. A team with a low Corsi percentage and a high PDO is the strongest fade candidate: they are being outplayed but winning anyway through unsustainable shooting and saving rates. The eventual collision of poor process and normalizing luck produces a losing streak that the market does not see coming because it was focused on the record rather than the underlying numbers.

Totals betting also benefits from PDO analysis. A team with an inflated PDO is scoring more goals than its process supports, which means the game totals in its matchups may be set higher than the team’s true offensive output warrants. Betting the under in games involving high-PDO teams captures the eventual normalization of shooting percentage, which suppresses goal output over time.

Regression Is Not a Dirty Word

Bettors who use PDO are often accused of ignoring what is happening on the ice in favor of abstract numbers. That criticism misses the point. PDO does not ignore what is happening. It tells you what is happening at the most fundamental level: whether a team’s results are built on repeatable skill or on temporary luck. The ice-level events that produce a 103 PDO — the lucky bounces, the highlight-reel saves, the deflections that find twine — are real. They happened. But they are not going to keep happening at the same rate, and the bettor who understands that is the bettor who positions himself on the right side of the correction.

Regression is not a prediction of failure. It is a prediction of normalization. The teams that regress from a high PDO do not become bad teams. They become teams whose results align with their process, and for most of them, that process is still good enough to win plenty of games. The profit comes not from the team’s collapse but from the market’s adjustment: the price was set for a 103 PDO team, and you collected the edge when reality delivered a 100 PDO team instead.