As I refer to the law of large numbers here it has to do with the number of traders, and the number of trades, if this is large enough then statistically you get fewer outliers from the actions of individual traders and individual trades. They don't disappear completely but they are less frequent. And if outliers are less frequent I think you can rely more on statistical measurements.
I think you're making a big mistake with this kind of reasoning, the returns are not normal (just download any course from yahoo and make a QQ plot). Assuming this is something that in my opinion will cost you money (or rather, you will severely underestimate the volatility). I believe outliers are important and provide a lot of information (and happens on the days which you read about in the newspapers....).
Btw, this rasied an interesting question: What is easier to predict, where 10 people are going or where 10 thousand people are going?
You are absolutely right. Outliers provide important data. When I am talking about eliminating outliers I am referring to the results of the projection compared to actual events, not eliminating them from the source data. Because, there is a certain path dependence that comes from extreme events and this has to be included on the analysis.
It is absolutely easier to predict where 10 thousand people are going...
I probably did not phrase that quite correctly either. It is easier to spot true outliers from a large data population than a small one. If you only have ten, then does a 10% event really count as an outlier? It is hard to tell. But if you have 1000, then it is pretty easy to tell when a handful of events are true outliers...
The fact that this level of conversation exists for Bitcoin says everything. If it's not Bitcoin, it will be something else. I think it'll be bitcoin.
