I crunched some PDO numbers for an article I wrote over at Pensburgh and figured I'd re-post the big points here. I'm not sure if anyone has ever looked at individual PDO numbers in great detail, so I wanted to see how players' PDO's changed over the years; in essence, PDO's predictive validity. The model for my analysis was inspired in part by Tyler Dellow's work in this article.
In my analysis, I looked at the top 40 and bottom 40 player PDO's for the 2007-2008 season. I then compared how their PDO from that season compared to their average PDO over the next two years. I limited the sample to players who played in at least 50 regular season games and saw at least 10 min of ES ice time per game in 2007-2008. All of my data was pulled from the behind the net. I found that the top 40 players had an average PDO of 103.3% and the bottom 40 had an average PDO of 97.1% in 2007-2008. The average over the next two years for both groups was 100.2%, which meant that the top players dropped on average by 3.1% and the bottom players rose on average by 3.1%.
It seems that PDO is an incredibly strong predictive tool for individual players, as only two of the 80 players surveyed failed to move in the predicted direction. Adding all of the players into the mix would weaken the correlation, but I think the point still holds that PDO is a very good predictor of future performance, and especially so for those at the extremes.
While 100% looks like the baseline we should compare the vast majority of players to, there was some elite talent that had a naturally higher PDO over the period sampled (like Evgeni Malkin and Nicklas Lidstrom). I have the excel sheet I used for this analysis if anyone would like to see it.