Puckatron 3000
Glitchy Prototype
Thanks for posting Garret. I was hoping you'd chime in on this.
There's that word - "significant".
This argument does not dispute the thought experiment. It only questions how much better goalie 1's sv% would be than goalie 2. What's the threshold? 1% sv? 2% sv? 5% sv? With all the confounding factors in attempting to analyze this, is it hard to say? If so, perhaps it is significant, depending on whatever threshold we might agree on.
That's a fair point to consider. But then, what you are claiming is that a goalie with a strong D faces the same proportion of "good" vs "bad" quality shots as a goalie with a weak D (within the usual statistical variance). This seems possible, but improbable to me. And, improbable things should only be believed if proven.
I question whether some of the statistical studies on this which claim proof have rather come to incorrect conclusions due to the noise-to-signal ratio of other confounding factors. The analysis is only as good as the data, and this is a pretty tough data set to eek out. For starters, it requires we quantify good vs. bad defense (a hard thing to do), and if that isn't done properly, any conclusions are built on a shaky foundation.
1) Significance of effect. Just because a team can decrease quality shots, doesn't mean the difference between teams is large enough to impact goals. Goals are rare, even for high quality shots. It takes a large amount of high quality shots to make a significant difference in goals.
There's that word - "significant".
This argument does not dispute the thought experiment. It only questions how much better goalie 1's sv% would be than goalie 2. What's the threshold? 1% sv? 2% sv? 5% sv? With all the confounding factors in attempting to analyze this, is it hard to say? If so, perhaps it is significant, depending on whatever threshold we might agree on.
2) The likelihood of whether or not that same team would be more talented in other areas. Teams who are skilled and have good coaching to be able to decrease low-quality shots against which would then lessen the effect on the ratio.
That's a fair point to consider. But then, what you are claiming is that a goalie with a strong D faces the same proportion of "good" vs "bad" quality shots as a goalie with a weak D (within the usual statistical variance). This seems possible, but improbable to me. And, improbable things should only be believed if proven.
I question whether some of the statistical studies on this which claim proof have rather come to incorrect conclusions due to the noise-to-signal ratio of other confounding factors. The analysis is only as good as the data, and this is a pretty tough data set to eek out. For starters, it requires we quantify good vs. bad defense (a hard thing to do), and if that isn't done properly, any conclusions are built on a shaky foundation.