From theory to practice. Part 2 - page 27

New comment
 

Sifted Open M1 price by some algorithm from December to the current time, got about 1000 price data.

Got such 'pictures':

Indicator like Sasha's. Sample = 5 ;))) haha ha, if 10 just one trade. Range of interval by formula D = 3 * const * Sqrt(sample)

That's how to find such a key to filter out a false signal (or invert it)?

What to count, only understandable language for nerds ;)?


I have attached a data archive, there is a markup of when to buy and when to sell. There are a couple of (sort of) losing entries.

Files:
Tabs_Exel.zip  42 kb
 
Shurik's model does not assume the presence of false signals) It assumes that (almost) all signals are true)
 
secret:
Shurik's model does not assume the presence of false signals) It assumes that (almost) all signals are true)

Unfortunately (for Shurik), the market does not always agree with them (the model and Shurik). But that is not his (the market's) problem).

Personally, I wouldn't be interested in rebuilding the whole model from scratch, but just to try to supplement it a bit on the basis of the indicators it already contains - the sums of increments and their modules.

Let him do the complete rebuilding himself)

 
So where are the concrete proposals? So far just nothing about anything.
 
Evgeniy Chumakov:
So where are the concrete suggestions? Just nothing so far.
Increase the magic constant) Deals will be less, but they will be better (probably)
 

If you think in the spirit of diversification of the existing system, then you should try to catch strong movements, after which the price gets stuck at a new level, not rushing to return.

The first thing that comes to mind is an obvious entry in the direction of channel breakthrough (of a different width than the original system) and exit upon return.

 
Aleksey Nikolayev:

Personally, I would be interested not to rebuild this whole model from scratch, but just try to supplement it a bit based on the indicators it already has - the sums of increments and their modules.

Range-to-travel ratio is a long known thing (you can Google Kaufman's AMA).
You can play with this coefficient, but it seems to me it will be about the same. All such metrics lag a lot.
 
Aleksey Nikolayev:

You could probably also add a minimum dwell time inside the channel before breaking through as a filter.

I would add 'speed' of penetration as a filter, to both versions.
 
Aleksey Nikolayev:

The first thing that comes to mind is the obvious entry in the direction of the channel breakdown (of a different width than the original system) and exit on return)

I vanguard that there will be a very small distance between the return signal and the trend signal) Up to complete uncertainty)
In the bollinger variant this has all been explored by inquiring minds for a long time)
 
secret:
The range-to-travel ratio is a well-known thing (google Kaufman's AMA).
You could play around with this coefficient, but I think it would be about the same. All such metrics lag a lot.

I agree, the sum of these moduli appears quite often - in Matan, for example, it is called variation.

Alas, lagged indicators are all we have)

By the way, beautiful equities (whether from tests or from real) from the past are also just another lagging indicator)

1...202122232425262728293031323334...179
New comment