PDO, Luck, and the Minnesota Wild

You’ve heard several of us mention the Wild’s high mark of 106.79 PDO and how unsustainable that is. And while some people might not be interested in the discussion because it involves #FancyStats, I’d like to shed a little light on what it means and help everyone understand that it is not all that fancy after all. It is a statistic that can be really useful for discussing both large and small sample sizes while still considering a minimum sample to make it meaningful. Hopefully by understanding what it measures, we can find a way to apply it usefully to what the Wild have done through their first ten games and understand what it does and doesn’t say about the future. If I could recommend one advanced statistic for anyone to start with it would be PDO.

PDO simply is the sum total of a team’s shooting percentage and save percentage times ten. It is sometimes referred to as SPSV% and is named after its creator’s handle, so it is not an acronym. The common adage is that it is a measurement of a team’s luck and can help evaluate whether a team is over- or under-performing, based on historical data almost every team trends to a PDO of 100 over a large sample size. If a team has a PDO above 100, the assertion is that the team is lucky. Lucky bounces, stanchions, and the like are contributing to a team over-performing on how many goals are going in for them and how many are not against them. Inversely, a team below the 100 mark is suffering from bad luck that is resulting in less shots going in for them and goals against.

100 PDO

You’ve probably heard the term “regressing to the mean” in reference to PDO. The mean, or average, PDO is 100. This number has been calculated over many years of historical data across the league - thousands upon thousands of games of full rosters of players shooting and save percentage. Over the course of a season or longer with this astronomical amount of data, it tends to pull the average to a clear-cut 100. That is the effect of honing in on a number, which gives us a reference point with which to judge any other result. The fact that PDO generally progresses or regresses to 100 over the course of a season illustrates how large the sample size is in a season. What this means is that even a small deviation from 100 has a significant impact on a team’s performance. For example, the Washington Capitals tallied 120 total points last year and closed the season with a PDO of 101.7, with a SH% of 9.9% (2nd in NHL) and and SV% of 91.8% (3rd in NHL).

Luck

Using the example of the Capitals, the basic application of the PDO rule would dictate that Alex Ovechkin, Braden Holtby and company had a really lucky season, seeing more pucks go in the net for them and less against, which led them to winning more games and earning more points. It seems pretty logical that more points in the standings would come from more goals for and less against, but the argument of luck as a singular premise is where it starts to unravel. Certainly a lucky bounce off a stanchion or a defender at the last second can contribute to a higher success rate at shooting. But so can a highly skilled player. Alex Ovechkin logged 398 shots at 12.6 SH% good for 50 goals. It is clear that with his volume of shots and high percentage of success, Ovi had a significant impact on the team’s final 9.9 SH%. Braden Holtby posted a 92.2 SV% en route to a 2.2 Goals Against Average. If we follow the logic of PDO being determined by luck, then we are stating that Ovechkin (a career 12.4 SH%) and Holtby (career 92.2 SV%) both had extremely lucky seasons in which the bounces went their way, and they over-performed expectations. But if you’ll recall, Ovechkin just happened to be the top goal-scorer in the league and Holtby won the Vezina trophy as best goaltender in the league. The premise of luck dictating PDO would state that these players’ (who posted career average numbers) success was driven by luck, and that is where the assumption is flawed. Their actual performance was dictated by skill, which is measurably repeatable in the fact that their performances were almost identical to their career average, and by definition a “lucky” stretch would be far above average.

106.79

The top of the hill for the Wild crested at 106.79 PDO following their victory over the Dallas Stars. In just one game, that number decreased to a bit more modest 105.5, which is still sky-high on the PDO chart. As has been pointed out on this blog many times lately, the underlying fancier stats indicate that the team is under-performing, yet the points have continued to pile up.

So it must be luck that’s driving the success then, right?

Not so fast. First things first, 105.5 PDO is a 10 game measurement of one team’s data. It’s a large enough sample size to try to grasp some meaning from the results, but is far too small to draw many definite conclusions about the team. Some would say that ultimately, the team is going through a really lucky streak right now and that the bottom will absolutely fall out as the team runs out of puck luck and plummets back down to the inevitable 100.

I tend to interpret it a bit differently. First, let’s quantify a lucky goal, and here’s a prime example from Tuesday’s game:

The puck touched 3 sticks on its way in and the third degree of deflection took away any chance Devan Dubnyk had to stop this shot. We can quantify that as a lucky bounce.

So here’s a question for the Wild faithful: how many of those type of goals have the Wild scored this year? Or EVER? I would argue relatively few, and the “lucky” effects have been negligible. The Wild currently boast a 12.6 SH% (2nd in NHL) and 92.9 SV% (2nd in NHL) for a total PDO of 105.5. The SV% isn’t completely out of reach for a longer period of time, but has obviously been bolstered by Dubnyk’s consecutive shutouts and will also be affected in the future by games started by Darcy Kuemper. The SH% is very high and generally won’t be sustainable over a season, as the best season on record is around 11%. However, it is incredible to believe that the elevated SH% is solely a product of luck, rather than a team using its skill to take advantage of opportunities. Luck is certainly a factor in PDO, but it isn’t the only one, and I believe that skill weighs just as heavily on the result.

What comes next?

Even though the eye test and fancier stats paint a picture that the team isn’t playing very well, it continues to put points on the board. The narrative has been fairly dark surrounding the team and its prospects for the future. They must be over-performing and the results aren’t maintainable. And that’s true - to a certain extent. The team cannot continue to play the way it is currently playing and expect its SH% and SV% to hold up for the entire season. But, we are just 10 games into a brand new coach’s system, which is measurably different than the previous regime and is being executed with more or less the same group of players. The players are growing into that system, and can’t have been expected to master it just 10 games in. Bruce Boudreau himself believed it would take until Christmas for the team to totally master his system, and it certainly isn’t there yet. The silver lining is that even though the results aren’t sustainable with the current level of play, the current level of play is well below the bar the team will set for itself this season. If the Wild can weather the storm of mastering the new system the way they have done so far, they can close the performance gap by executing the system and all of its finer points. And while their PDO won’t likely end up in the 105 range, it is certainly possible that the skilled success of the players can post a PDO over 100, just as the Capitals did last year.