Looking for patterns - page 104

 
Martingeil:

I wonder what the time span is?

Nah, it's all 20 years of information in the array. It's all compact, all the information.

 
You need to identify a pattern, for example a week past, and its impact on the present week, i.e. on each day. There is no need to scratch the whole story, a week completes the cycle of any. The price moves this way in myopinion.
 
This is exactly what you should find out. There are a lot of different conditions you can come up with and check. It could take a long time. I can't do it alone. Plug me in. I can give you an example of how to set the conditions in the code.
 
Martingeil:

I wonder what is the time span?

You're not going to scratch the whole story, there aren't enough resources, what's the interval?

I understand it is a week, there is no point in taking it higher than that, you will get into another market condition, which may never come back due to various circumstances, time of year, exporters' need for currency for a given period of time. The same coronavirus dictates its conditions at the moment, for oil futures for example, next year this interval will not be relevant.

every action has a counteraction

i.e. proportionally everything repeats itself

 
Aleksei Stepanenko:

No, there is information for all 20 years in the array. Everything is compact, all the information.

This information will overlap, you will get something in between. Roughly similar, I would look for a formula rather than similar cycles. Murad Ismayilov did something similar in 2012, check out his work.

 
Martingeil:

This information will overlap, you will get something in between.

The trends are stored sequentially in an array, alternating up and down. There is no need to average them. Suppose we build the logic: take the previous trend, look at its parameters, check what happened to the next one. Collect statistics for all cases. There is no averaging.

 
Aleksei Stepanenko:

The trends are stored consecutively in the array, alternating up and down. There is no need to average them. Suppose we build logic: take the previous trend, look at its parameters, check what happened to the next one. Collect statistics for all cases. There is no averaging.

It is then required to break them down by median, to determine the statistics of the trends.

 
Yes, just a huge layer of interesting work. Let's work together.
 

Let me tell you about the internals of the indicator.

The indicator has two arrays LocalExtremes and GlobalExtremes. Each element of them stores information about one trend. For a local fast trend and a global long-term trend respectively. There are more local trends than global trends. A global trend may consist of several local trends. In an array the trends alternate each other in direction. The time and price of the end of one trend is the beginning of the other.

In the zero element Extrmes[0], there is the oldest trend from 1905 :) In the last element Extremes[Finish] lies the latest trend, maybe even the current trend.

We record the trend when the price has passed a certain distance. Yes, this is later than the start date of the trend, but there is no other way, the future is unknown. When registering, we create a new array element and enter the current data. And the previous array contains the actual end date of the previous trend. That is, all information in the array is accurate. When an extremum is updated, data in the last element will also be updated.
 
Aleksei Stepanenko:
I will tell you about the inside of the indicator for the unprepared listener. It has arrays LocalExtremes and GlobalExtremes. Each element of it stores information about the trend. For LocalExtremes and GlobalExtremes respectively. There are more local trends than global trends. A global trend may consist of several local trends. In an array the trends alternate each other in direction. The time and price of the end of one trend is the beginning of another.
In the zero element Extrmes[0], there is the oldest trend from 1905 :) The last element Extremes[Finish] contains the latest trend, maybe even the current trend.
Now we register the trend when the price has travelled a certain distance. Yes, it is later than the trend start date, but there is no other way, the future is unknown. When registering, we create a new element of the array and record the current data. And the previous array contains the real end date. That is, all information in the array is accurate. When the extremum is updated, the data in the last element is also updated.

Let's take a simpler approach, and I think this information will be easier to understand.

Let's build a zigzag on the interval D1, and W1 on the chart H1, that data will give some food for thought.

D1 - we will consider it as a cycle, W1 - we will consider it as a global cycle.

1