Making a Python trading system for MT. - page 16

 
Alexander_K2:

If you calculate this measure correctly, you will find that in a sliding window, we are always inside the normal distribution and have the coveted Grail.

Logically, the shorter the measurement interval, the less likely the outlier. Of course, adaptive channels correct for these outliers. The question is how to determine this 'Measure', i.e. what should be the criteria for selecting channels so that they measure price changes well?

 
Aleksey Vyazmikin:

Logically, the shorter the measurement interval, the less likely the outliers. Of course, adaptive channels correct for these outliers. The question is how to determine this "Measure", i.e. what should be the criterion for selecting channels so that they measure price changes well?

Take a sample of 1,000,000 linear deviations of price from your moving average. Look at the histogram of these deviations. The closer the resulting distribution is to a normal distribution, the better.

 
Alexander_K2:

Take a sample of 1,000,000 linear price deviations from your moving average. Look at the histogram of these deviations. The closer to a normal distribution, the better.

The question is whether the automation needs a coefficient or something else that shows the dynamics and makes it possible to measure.

1kk is too much, I don't use ticks.

By the way, why did you decide to count the deviation from the average instead of dividing hai and loys into two parts and count hai at the upper boundary and loys at the lower boundary of the channel - it would be more equitable.

 
Aleksey Vyazmikin:

The question is about automation, is there a single coefficient or is there something else that shows the dynamics and makes it possible to measure.

1kk is too much, I don't use ticks.

By the way, why did you decide to count deviation from the average instead of dividing hai and loys into two parts and counting hai at the upper boundary and loys at the lower boundary of the channel - it would be more fair.

When the linear deviations of the price from the moving average form a normal (read: binomial) distribution, the grail is at hand. You just have to know about the binomial distribution. Yuri does :)

 
Alexander_K2:

When your linear price deviations from the moving average form a normal (read binomial) distribution, the grail is at hand. You just have to know about the binomial distribution. Vaughn, Yuri does :)

I do not like redundant theory. I need a method to determine whether the indicator is suitable for creating an artificial normal distribution or not, and then I will see what to make of it.

 
All right, since no one has shown any interest in the price chart with averages (posted above) I'll delete it.
 
Evgeniy Chumakov:
Well, since no one showed any interest in the price chart with averages (posted above) I will delete it.

You could not answer how you got this graph, I did not understand how you found the USD price separately, which is not expressed in anything...

 
Aleksey Vyazmikin:

You could not answer how you got this graph, I did not understand how you found the USD price separately, it was not expressed in anything...

You couldn't and didn't want to. And how can this separate price not be expressed in anything?
 
Evgeniy Chumakov:
Couldn't and won't are different things. And how this separate price can be not expressed in anything.

Price is the valuation of one asset in the equivalent of another, and this is what the other is not clear.

 
Conclusions: относительно не запаздывающей меры центральной тенденции we will always be inside the normal distribution. In fact - inside the grail. And if this measure is polynomial regression lines, which we need to check again and again, then that's it - the problem is solved.

Thank you for your attention.


And if price is the non-lagged measure of the central tendency?