Please tell me you're just being facetious.
Not really. Even if Tinordi is already a better player, it is not necessarily best to leave him in the top-6 over Murray, if it is believed that Tinordi's development will benefit more by dominating 25 minutes a night in Hamilton.
Well, our winning percentage with Murray is 100%, however it's currently meaningless given that it's two games, just like Murray's fenwick is meaningless. Even then it would still be close to meaningless as Murray is not the only player coming and going. Desharnais is on his way out of the lineup, and Pacioretty and Emelin are on their way in, so it's conceivable the team will continue to play better with Murray in the lineup,
regardless of Murray. That doesn't negate the hypothesis that the team should play better with Murray in the lineup, even when he is off the ice.
Science works better than old wives tales and traditions when you have an actual science, like in physics or chemistry. When you're dealing with a situation with more variables, and applying less rigorous methods, such as in quantitative history, most hedge funds, political science, or hockey analytics, you don't really end up with science, just pseudoscience.
As an example, political scientists typically subscribe to "democratic peace theory", that no democracy can go to war with another democracy, they have all sorts of irrational explanations for this, and I've heard some of them tell me that it's the most fundamental law of political science, equivalent to gravity in physics. Of course, that's not actual scientific knowledge, that's just obfuscation, the type of obfuscation that is inevitable when you analyse a complex situation with ultra-simplistic methods and thinking.
What I would need to be convinced that Murray is a terrible player just from the stat sheets is the following (basically to know that the stat sheet is meaningful):
- For hockey Sabremetrics to become a genuinely more robust discipline, to have the effectiveness of physics as opposed to the arrogance of economics. That means numerical methods that cannot be replicated in mere minutes by a freshmen undergraduate majoring in statistics, and a predictive track record that beats pure chance. How good are you at predicting the standings at the start of the year, can you do better than merely assuming a replica of last year's standings? Can you do it 7 times out of 10, with your errors those 3 other times being consistent with your reported theoretical error? If not, you have nothing. Can you predict the art ross race better than merely copying and pasting last year's art ross race, even after giving yourself an injury allowance?
- For the actual underlying numbers hockey Sabremetrics numbers to be more rigorously counted. A missed shot or a hit or a takeaway should mean the same thing in every arena, as opposed to meaning something different, and then being treated as if it means the same.
- For the language from the Sabremetrics community to become more robust and accurate. State "We can find no evidence that size is linked to better play" rather than "there is no evidence linking size to better play", as a generic example.Just because you can't find something, doesn't mean it's not there. As an example from real science, this dark matter experiment couldn't find any dark matter:
http://arxiv.org/abs/1310.8214
So they state that they could not find anything, rather than that nothing exists. This simple and modest philosophical mindset is crucial to justifying confidence that the people involved know what they're doing, and is not demonstrated by your community.
Until then, hockey sabremetrics are minimally robust, and should be used as supporting rather than primary arguments.
Other than that, when I'm watching the games closely I can often tell where a player is going prior to the stats experts. For example, I could see that Desharnais was not the real McCoy as early as 2012, even during his miracle season, in spite of the fact that all of his stats indicated that he was in fact the real mccoy. I watched the games and I saw that he was weak on the puck and easily lost the puck. Yet he had good takeaway/giveaway numbers at the time, probably due to the stat being unreliable. There is also the fact that he benefited from having two huge wingers, with "size" usually showing no correlation with anything in most hockey stats. Similarly, if Murray is in the 1st or 2nd percentiles as you and mathman imply, then it should be obvious from pure viewing.
ETA: None of this is to argue that Sabremetrics is a waste of time. It's not, but hockey is a complex game with a lot of mutually-interactive parts, unlike baseball. If people keep pouring over data, it's entirely conceivable that the these methods could become as powerful as currently claimed within the next 30 years. There is no easy road to science, but the first step is almost always counting.
Kreider was pestering Gallagher. Murray saw he was possibly going to get the puck so Murray hit him. That's what a big guy protector player does. Chris Neil is this type of player who often takes penalties too. Yes one is a D & the other a forward but if someone is pestering Conacher & Neil takes care of business do Sens fans freak? We can't have the Kreider's of the world pushing around Gallagher! 
As such, Murray should converge to 21st percentile once a large sample size is reached, and if he does so he will be "average" for his role. A respectable basket of stats should be used, and we should recognise that the margin error itself will be a few percentage points.
