The thing that no stat can measure is if a puck will go in the net or when it will not go in the net.

Throw pucks at the net all game & end up with a duck egg or throw a single shot & score! No one can measure or predict that no matter what you have - it's the unknown. How many teams out shoot the other & lose? How many times you see a team dominate possession but end up behind after the first goal is scored by the opposition - Pens have had a huge amount of that lately. If you want to call that luck then do but don't call it PDO if your goalie gets scored on in the first 2 or 3 shots after you shoot 15. Call that good goal tending by the opposition & bad D or goal tending by your team.

In saying that, goalies 'may' have weaknesses - isles threw 2 shots over Flowers left shoulder the other day - same shot same result GOAL. Now when we start to see stats guys give us credible information along with the stats ie players strengths & weaknesses, you will see a much larger data base of information that can correlate more with how a game WILL play out more than just match stats up to games already played.

You will still have the unknowns but more of an idea of what might be the better option - ie throw shots over Flowers left shoulder before anything else.

I believe the Rangers series will have little to do with possession numbers & more with D & goal tending. We will throw 25-35 shots to their 15-20 & it's possible to be behind 0-3. Possession means little if the goal tending is good/great. That is a part of the Rags that has been very good even without the King. Pens need to score & once they do try to score again if they don't just shut the game down & win games 1-0 if possible. He'll just put the caravan in front & stand the tent up & let them shoot all game at that.
- Aussiepenguin

I wasn’t suggesting that PDO and Fenwick (or Corsi) are mutually exclusive. The outcome of a game is however, interdependent upon them. One can slice and partition a game to uncover score effects, gauge the impact of special teams or 5-on-5 effectiveness, or to find any differences in home/road games. These are illustrative examples. Hockey is filled with randomness, and as such, the data collected contains a lot of noise. The task of highly trained analysts is to separate the signal from the noise. So depending on the problem, there are a variety of solutions that can be used to model the state/ outcome of a game. At the lighter end of the spectrum, we get correlations (of events) with winning. Correlation alone should not be used as the end-all answer. Often correlation is confused with causation. These are two wholly separate concepts, of which the second is extremely hard to uncover (outside of designed experiments). Just the other day, I read a Google headline (Health) about a rise in skin cancer cases in the UK due to increased demand for low-cost vacation packages. Really?! For more spurious correlation examples: http://tylervigen.com/
Further, there are probabilistic models that can be used to help predict outcomes, as well as machine learning algorithms that can uncover trends/patterns and/or latent variables associated with a given outcome (e.g. goal, win, lose). Bayesian methods allow an analyst to include uncertainty in their models, and are more adept to handling small sample sizes with the addition of simulation. There is even research being done to see whether officiating is a Markovian process (i.e. are penalty calls memory-less of previous ones). Whether one is trying to model the outcome of a hockey game, playoff series, or predicting the Presidential election, or even trying to beat the stock market, signal from noise is ever-present. All models are wrong… some are useful (George P. Box). I don’t think many analysts are running around claiming to have found a secret ‘formula’ to winning; they are however, finding trends, associations, correlations, and other patterns (signal) from heaps of game data that can be used to postulate strategy.

Back to hockey, Fenwick close and Fenwick 5-on-5 have been used as an indicator of future success. Intuitively, the concept of puck possession makes sense; how is the opponent supposed to score with out the puck? If my team dominates game flow (shot attempts) then the opponent is forced to expend more energy defending AND not attacking. Over time, from game-to-game, lagging in the possession department, can then lead to some pretty consistent outcomes. Hockey is a team sport – hence PDO. Sometimes, a goaltender can bailout his team, while other times leaving them high and dry. The probability of outcome can be leveraged by the interdependence of such metrics.

Personally, I’ve found the most information from stats/ metrics are found in differentials, along with current trend (from the past ## of games). Given RW’s possession vs. PDO paradigm, shot differential will most likely be the determining factor between winning and losing. Think about the interdependence between ‘D and goaltending’ and that of possession and PDO. NYR SH% = 9.60% and SV% 92.3%. Pens SH% = 8.37% and SV% = 91.2%. You can do the math to figure out the expected number of goals for each team, given SOG’s (25-35 and 15-20 range) + PDO. What will change (on the ice) as opposed to ‘on paper’?

Hint: an AV coached team with a min of 4, and at most times 5 guys collapsed down-low, protecting the house/ blocking SOG's. Yawn.