Explaining PDO and why it matters in hockey

Hockey is random. Actually, the most random among North American major sports.

The sheer amount of events that happen over the course of a single game makes it hard to truly predict how sustainable a team’s success is. But, there have been attempts to get close to quantifying this.

One example is that of PDO (which doesn’t actually stand for anything). As a statistic, PDO is simple – shooting percentage plus save percentage. While that might not seem like much, this statistic is attempting to capture “puck luck” with a number, a proxy measure of how fortunate a team or player is. While there are absolutely flaws to PDO, not many stats attempt to do what it is trying to communicate.

A high PDO means that a team is scoring a high proportion of their shots and/or conceding a smaller proportion of goals against them. This is what teams ideally strive for, but what PDO tracks is how teams regress to the mean over the course of a season. Teams should usually always end up with PDOs around 100, as with every shot there are two outcomes: a save, or a goal. By the end of the year, each team’s shooting percentage plus save percentage should equal 100 as they play the same number of games against the same number of teams (a closed system).

Think back to secondary school pre-calculus and calculus. PDO exists as a standard distribution curve as a normal distribution, with the most probable value being 100. For a player to sustain high PDO, they and their linemates would have to keep scoring at the same high rate, while also getting consistently elite goaltending, something that just isn’t feasible over the course of 82 games.

What PDO can tell us about teams and players is that despite swings in their play, either positive or negative, they will tend to trend back to the mean as the sample size grows. This can be helpful when looking at how likely a player is to repeat the same scoring performance from years prior, or how effectively a team can maintain an excellent save percentage. The same works in the reverse direction, a team that was awful at scoring usually swings back up to the mean.

Generally speaking, PDO has low repeatability, meaning that from year to year, the values will vary. That itself isn’t helpful, but when the data is charted, it’s possible to create a regression line through regression analysis, creating a trend line that can predict the “luck” a team can experience over the course of a season or timeline. When taken in conjunction with other statistics such as shots for percentage and goals for percentage, PDO can help show where a team was lucky in a previous season or less fortunate in another.

Talent does impact save percentage and shooting percentage. This is something that PDO doesn’t entirely account for, which is shown by discrepancies between the expected values and the actual values. However, full-season values for both categories are not defined by talent. Rather, the randomness outweighs talent a large majority of the time. For a statistic to be absolutely determined by skill, it is to remove everything else around to it make a vacuum. That isn’t how hockey exists, and will mean that most players will be impacted by randomness more than they will be in control of the situations.

PDO is something that won’t be discussed frequently in The Statsies this upcoming season but is absolutely something interesting to keep an eye on. The regression analysis from season to season, from team to player, lends itself well as a proxy gauge for how lucky or unlucky they have been.

The PDO of someone like Andrei Kuzmenko is something noteworthy, for example, and this season could be fun to track to see how much ‘luck’ he will retain. Or on the contrary, J.T. Miller and if he is more primed to have some better ‘luck’ next season after posting a 0.975 PDO.

PDO is also handy in fantasy hockey when looking at how sustainable player production is, whether a cold streak will be snapped or a hot streak be maintained.

So the next time a coach says that the pucks just didn’t bounce their way, it could actually be not far from the truth.

Puck luck is absolutely a thing, with randomness bouncing every which way in the game of hockey.