You said, "If the data distribution were approximately normal then about 68% of values would fall within the Bollinger Bands with the Std Dev set  to 1.00, 95% with SD set to 2.00 and 99.7% with SD set to 3.00."

1)  Does this apply to Bollinger Bands no matter how many bars they incorporate?

2)  If not, then is there a statistical formula (or an app) calculcates the Std Dev for a certain number of bars?

3)  As an example, let's say you see that more than 99.7% of bars are falling outside a Std Dev of 3.0 within a 3 month period, does this tell you that the data distribution is not normal?

4) Would it be overfitting to determine data distribution on a rolling basis for each separate stock or ETF?

5)  Do any of those who manage large assets determine data disribution on a rolling basis for each separate holding?

To understand data distribution, I found:

https://mathbitsnotebook.com/Algebra1/StatisticsData/STShapes.html

It applies no matter what the period.

Stock data quite simply is not normally distributed. Bollinger Bands can be a useful tool, but the lines are not statistically valid in the sense of predicting the percentage of values which should fall inside or outside of certain bands.

That said, 99.7% of bars falling outside a 3 standard deviation range of the average would definitely be a demonstration of this.

You can in fact (at least mathematically but not in TC2000) calculate a value at which a certain percentage of the data falls within that range. That is kind of what median and percentile rank type calculations are all about. It isn't standard deviation however and it may or may not be overfitting the data (depending rather specifically on the underlying data).

I have no idea what measures those who manage large assets do or do not use.

Thanks for your explanation, appreciated.

You're welcome.