Discussion of article "A scientific approach to the development of trading algorithms" - page 5

New comment
 
Maxim Romanov:

you have to figure out what you're counting against. If it's relative to the previous price, I don't like it very much.

When we calculated the difference (block size), we did it relative to the price and didn't bother. What's different when you take the relative change instead of the absolute change?

[Deleted]  

To the question of volatility.

(I show in pictures to make it easier for people with undeveloped figurative perception to perceive the information)

Let the height of the block be set at 10 points. The price can pass these 10 points for any (not predetermined) time.

Depending on this, the volatility will be different.


1)

.


2)

.


3)

.


DIRECT !!!

 
fxsaber:

When we calculated the difference (block size), we did it relative to the price and didn't bother. What has changed when you take relative change instead of absolute change?

I have noticed a peculiarity that on stocks, when the block size grows, a proportional reduction of the lot can bring trading to a plus. I did not describe it in the article. I would like to understand why it works like that. So before you don't mess around, you first need to understand what affects what. Here I am thinking how it would be more correct or there is really no difference.

 
Igor Makanu:

count from the beginning of the day or month - get significant levels within a day/month

but this will be another scientific approach, but it will not bring you closer to developing a trading strategy, it will bring you closer to predicting reversals with a small probability of correct results ;)

I will try to calculate it from something later. Rather, I will count from profit.

I develop and work on trading strategies, and research is needed to improve parameters.

And I have posted the trading strategy, you can use it, I am not sorry. Those who need it will improve it.
 
It is possible to take not constant size of blocks, but depending on time, for example from 00:00 - 06:00 hours block = X points , from 06:00 - 12:00 block = Y points, etc.
 
Evgeniy Chumakov:
You can take not constant size of blocks, but depending on time, for example from 00:00 - 06:00 hours block = X points , from 06:00 - 12:00 block = Y points, etc.

it is better to take non-rectangular blocks.

and in general with blocks - it is better to go to the construction site :-)

 
Evgeniy Chumakov:
It is possible to take not constant size of blocks, but depending on time, for example from 00:00 - 06:00 hours block = X points , from 06:00 - 12:00 block = Y points, etc.
This is just about the sampling method from my first article. An algorithm for the "correct" discretisation can definitely be developed. It will not be correct to change from time to time, because it is only for forex. Of course, it is possible to calculate some regularity for funds, but it is better to invent other algorithms instead of time. My goal is universal algorithms working on any instruments.
 
Evgeniy Chumakov:

Off topic, I did a little research. I took ZigZag and calculated average values of price increment and time interval for a certain time (for example, if the beginning of the current knee fell on 13:25, I searched history when the beginning coincided, with an error of a few minutes). Then I built a trendline with average price increment and time interval.

Like this:



So what's interesting is that price tends to level off to the mean trajectory. In this context, a "return to the mean" is simply a sideways price movement.

Yes, there is such a feature. It's not completely off the mark. According to my research, the price travels vertically an average number of steps to the degree of 0.4-0.6, depending on the instrument. That is, it does tend to these values.
 
Maxim Kuznetsov:

it's better to go for non-rectangular blocks.

and in general, it is better to go to the construction site with blocks :-)

Yes, it is better to take rounded edges) to avoid cutting yourself)
 

This is how a regularity was found: markets tend to be trending on any scale, but as the scale increases, the trending decreases, i.e. after passing N points vertically, with a probability of more than 50%, the price will pass the same number of points in the same direction. This pattern is good because it allows you to use a simple trend strategy for trading, where you can open a Buy position after each upward move and open a Sell position after each downward move.


You do not take into account the situation when the price starts to run through several of your blocks in one tick? In this case your blocks are drawn post facto and you cannot open trades on them.
123456789
New comment