A lot of criticism of hockey analytics takes the form of skepticism about the utility of analytics for making predictions. I'll give some recent examples, just to prove I'm not fighting a straw man:
- Gord Miller wrote (in two tweets), "Was asked by a delegate at the #ptse2013 Sports Management why I am against hockey analytics, or 'advanced stats'. I'm not...what I said to him is that I haven't yet seen a 'magic bullet' that can show why things happened or predict what is about to happen."
- Dave Nonis said that poor record keeping by various arenas render analytic measures inaccurate or irrelevant.
- Steve Simmons has often said that he doesn't think stats will work in hockey because it's too fluid a game with too much that can't be quantified.
It's easy for me to understand why some people don't enjoy the analytic dissection of sports. But I can't even imagine why people who haven't dug into it in detail would assume that the stats aren't useful.
There are a lot of people who put a lot of time into developing and understanding hockey stats. When those people come to the conclusion that shot differential is a useful predictive measure, do the skeptics imagine that it just never occurred to anyone that some shots are more dangerous than others?
It's common to hear lectures about correlation and causation; do these skeptics really suppose that nobody's ever looked at causality?
Do they presume there are people who don't think wins and losses matter?
What do these guys imagine that a thriving community of deeply interested analysts does, exactly? Is it really believable that someone just decided to pay attention to shot attempt differential for no good reason and everyone else just jumped on board without thinking to test it?
Of course there's evidence of predictive utility behind any prominent approach. Some of that evidence is mathematical, showing analytically that (for example) winning percentage in future games correlates better to a team's current place in the shot differential standings than to their place in the actual standings. Some of it is empirical, as each year analysts make predictions based heavily on shot differential and have so far generally done well.
It's no skin off my back if people aren't that into stats. But the blind assertion that the stats don't have predictive value is just silly. Each new stat goes through some pretty rigorous scrutiny, and it's awfully hard for any new stat approach to gain widespread acceptance without strong evidence of predictive utility.
I can't fathom why people would assume otherwise.