From theory to practice - page 1980

 
Alexander_K2:

2. Market process for GBPUSD on Seconds TF S1

Beautiful!

 
Доктор:

There seems to be a higher probability of small increments and large increments being adjacent thanWiener's. That's interesting.

Alexander, which row did you take? What did you do with the series when there were no quotes in a given second? There is a version that the series contains many conditional "candlesticks" with the body of zero height. Then it would be a "square".

 

1) We should not compare with a Gaussian, but with a randomly shuffled sequence of GBPUSD increments.

2) At times of a second or so, the price discreteness effects may be present - it is necessary to look at a one-dimensional histogram of increments.

3) Something else, but I already forgot

 
Aleksey Nikolayev:

1) We should not compare with a Gaussian, but with a randomly shuffled sequence of GBPUSD increments

If randomly shuffled, there will be a circle.

 
Доктор:

Alexander, which row did you take? What did you do with the row when there were no quotes in a given second? There is a version that there are many conditional "candlesticks" with a body of zero height in the row. Then it would be a "square".

Values were read once every 6 seconds. If there was no new quote, nothing was written to the array. I.e. - good data for about a month last year. Interesting to see the same time period on M1 - what will change.

 
Доктор:

If you randomly mix it up, it will be a circle.

Well, yes, but it will be the right circle) This is important when assessing the significance of the information contained in the sequence of increments, which is lost when they are randomly shuffled.

Aleksey Nikolayev:

3) Something else, but I've forgotten already.

That reminds me - you also need to initially approximate the increments to remove diurnal volatility fluctuations.

 
Aleksey Nikolayev:

Well, yes, but it would be the right circle) This is important when assessing the significance of the information contained in the sequence of increments, which is lost when they are randomly shuffled.

That reminds me - you also need to initially approximate the increments to remove diurnal fluctuations in volatility.

This is a visualization. And if you investigate analytically, then yes, mixing and normalisation.

 
Alexander_K:

The values were read once every 6 seconds. If there was no new quote, nothing was written to the array. I.e. - good data for about a month last year. Interesting to see the same period of time on M1 - what will change.

There is an assumption that there will be no square on M1. Too obvious inefficiency.

 
Alexander_K2:

Made for:

1. Wiener process without drift with Gaussian distribution of increments



2. Market process for GBPUSD on Seconds TF S6

It's crazy...

Some kind of square...

That's how the dependence of the current increment on the previous one looks like on the phase plane.

The sample volume is the same.

Here's the thing, you see...minimum price change is 1 point, and to tick more or further, some conventions have to be followed ;-)
it works without allowing arbitrage situation in triangles and other cycles.

One pair as the tip of the iceberg, carries little information. In other words, we should not look at just one currency pair, but at the supply/demand of all currencies

 
Доктор:

There is an assumption that the M1 will no longer have a square. The inefficiencies are so pronounced.

Absolutely.

Downloaded OPEN M1 data from Dukascopy for EURUSD for all of 2019.

Got the following picture:


Conclusion 1: Working with OPEN/CLOSE data TF M1 and above is the Path to the Abyss.

Amen.