SHOOTING PERCENTAGE, SAVE PERCENTAGE AND PDO
What it is: It's technically three different statistics, but they are all important in understanding why Cam Fowler gets a bad rap.
On-Ice Shooting Percentage (SH %) is the team's percentage of shots that result in goals while that player is on the ice. On-Ice Save Percentage (SV %) is the percentage of shots that are stopped by the goalie when the player is on the ice. PDO is the adding of these two stats together and might be accurately described as the crossroads of luck and sustainability.
Apparently, PDO (name is short for guy who found it) is a very important stat. The reason why it's important is that when you add shooting and save percentage together, the number will always progress/regress to 1000 or 1 depending on how you're writing your numbers. Knowing that allows us to factor in things like luck, or teammates because we will notice anomalies in the form of numbers that have a great distance from that 1000. The balance isn't limited to particular numbers. Defensive oriented people will see a shrink in SH% but compensate for an increase in SV% that will still lead to a balance of 1000.
Limits: It's tough to say what the limits are on shooting and save percentage. The primary one might be talent. Talented players will have numbers that are above the norm, and playing with those players can impact your numbers, good scorers affect on ice SH %, good goalies can affect on ice SV %. In terms of PDO, it's hard to say what a good number is, and therefore hard to determine how lucky a player was. However, it is pretty well understood that PDO does regress and that all players will find that shift to the mean. I've seen the window for PDO represented as +/- 20, and that seems the most reasonable to me.
Advantages: The advantages are the same as for anything that gets represented as a percentage; they eliminate anomalies and focus on overall trends. Instances of bad defense might stand out, but if those instances only result in a .910 SV%, then maybe those mistakes don't have a huge overall impact on whether or not your team wins. As for PDO, it calculates luck and sustainability. That's a huge advantage! Not to mention that by monitoring PDO we can sometimes clue in to wear shifts in play will occur, by looking at how a player or team's particular PDO is constructed.
What it means: I think SH% and SV% are pretty self explanatory and are most beneficial for looking at larger overall impacts of players that can be missed. For example, who would have known that Devante Smith-Pelly led us in SV% in 4 on 5 situations last season. I didn't. But it's worth noting, as maybe he should have been getting more time on the PK to see if the performance continued.
The important stat here is PDO. Because it's been shown that PDO regresses over time and balances at 1000, it's very useful for looking at things like sustainability, luck, and variances in performance. A high or low PDO can literally mean that a player was unlucky, that the bounces didn't go his way. A high PDO can mean that more pucks went off the pipe and in than out when he was on the ice. If a player has a high PDO chances are he can't sustain that output, unless he is an extraordinary player, like Teemu Selanne extraordinary.
Finally, when PDO is out of balance, you can look at which stat is out of balance and hopefully gain an idea of where things are going to shift. For example, Sheldon Souray is our second highest defender in terms of PDO right now at 1122. Francois Beauchemin is tops, but Souray is an example that's easier to follow. Right now Souray's PDO is a composition of a 19.6 SH% and a 92.6 SV %. That SV % isn't that unrealistic for a guy who can take care of his own end, but those shooting numbers are. Offense with Souray on the ice is going to regress, and probably regress pretty hard over the course of an 82 game season. PDO is important because it's the most accurate predictor of performance trends. Guys who finish with really high PDOs have usually produced results that can't be reproduced again. PDO has great value as a predictive stat.