Hint: It isn't luck
We know that PDO is the sum of a team's shooting percentage and its save percentage (we're just going to focus on team PDO) and that over time a team's PDO will regress towards but not to the league average of 1.00 (or 100.00 or 1,000.00 depending on what power of ten you're using). We also know that PDO is a measure of both skill and luck, see earlier article, and we've already established that if you're looking to PDO as a measure of luck it isn't a very good one.
We also know that as a data set grows larger the influence of statistical anomalies/noise (luck) on the metric decreases. So it follows that if you could control all the other variables and got a dataset large enough team PDO would eventually be a measure without the influence of luck. What you’d be left with is the relative skill of each team during the moments in a hockey game when a SOG is recorded. For a very large dataset it would essentially measure the relative skill of each team if each team had the same amount of SOG’s.
Hockey games aren’t designed to be equal, most nights the SOG’s aren’t equal. That’s why possession metrics like Corsi and Fenwick are so much better and more useful.
Also, you can’t control the other team variables long enough to get a good, large dataset to derive true team PDO. Teams change, they lose players to injury, trade, retirement or free agency so there will always be some noise in the metric.
But at the very least, can we please, please, please stop telling people PDO measures luck? If it was a good measurement of luck it would do a better job of measuring luck the more data you gave it and/or giving it enough data would eventually get you to an equal measurement for all teams. Neither of those is true so at the very least let’s put this myth to bed.