Once again, it's about the eternal: trend/flat. - page 17

 
Alexander Antoshkin:
Perhaps I'm not too old, but I remember as yesterday, that among the administrators of Tsarist Russia, not only in the time of Nicholas, but in the whole 19th century, there is probably no figure that caused so many reviews, assessments and characterizations, as Muravyov-Amursky ...
Everyone spoke of him as an autocrat, a stubborn narcissist, whose whole activity brings nothing but evil, and who does not want to accept the fact that as long as money will exist, so long will live commerce and advertising.
from which follows the fact that advertising is the eternal engine of commerce. And I'm sorry you never understood the benefit of the time series difference.
I'm just as much of an autocrat, by the way I'm in the advertising business myself, but that doesn't stop me from learning the intricacies of financial markets. You said that all this (talk about flat / trend) is nonsense and got the appropriate response. Think before you put your nickel in.
 
Комбинатор:
This one has a clinic, don't even try, it's a waste of time.
Too bad there are no likes, this post would get more than the whole top.
 
На данный момент поступаю элементарно - временные границы, скажем, 00:00-8:00 - флетовые движения, 08:00-22:00 трендовые (одно сплошное направленное движение или со сменой направления несколько раз), 22:00-00:00 - флет. Но такой the simplified method is very approximate, although it gives better performance of the TS, but does not allow it to be used on TFs older than H1.

So is it just the timeframe that is different for you? Or is the code for day/night also different?

As a rule, at night it is not flat, but just lower volatility and optimization will give you the difference not in TF, but rather an increased period, as it usually happens with lower volatility.

Here is an example of an Expert Advisor from one oscillator, but for night and day its TF

Testing on auto-optimizer by volking-forward method. Two months testing - one check. Net profit - by forwards, i.e. check.

One month Ch-profit Day Night
TF Period TF period
Oct 98,9 15 min 26 20min 40
Nov -96,2 15 min 30 20min 10
Dec 186,7 10min 20 H1 28
Jan 785,4 10min 10 H1 26
Feb -255,7 10min 12 H1 16
March -182,8 H2 10 H1 4
apn 424,9 H2 2 H3 2
May 327,7 H3 2 H3 2
June -249,8 H3 2 H3 2
July 697,7 H3 2 H4 40
Aug 195,5 H3 2 H4 40
September 91,2 H3 2 H4 12
Total:
2023.5


As you can see - Tamara and I go as a pair, the TF is almost the same.
 
Youri Tarshecki:

So is it just the timeframe that's different? Or is the code for day/night different as well?

At night, as a rule, it is not flat, but just lower volatility and optimisation will give you a difference not in TF, but rather a higher period, as it usually happens with lower volatility.

I used to have 2 different systems, a flat working in a channel and a trend working. I optimized the periods of the day in which the sisters perform best. This is how I obtained the previously announced working hours for the systems. After that I combined both systems in one EA that switches systems depending on time of day. The results have really inspired me to study more deeply the nature of t/f, to apply similar differences in the behaviour of these two systems, not only according to the time of day, but depending on the detection of t/f throughout the day.

I would not have been able to get such an improvement in the performance of the systems if I had used TFs above H1, because during the day on large TFs it would not have been possible to apply time filters. Of course, there are t/fs on longer TFs than H1, but they can be successfully detected only by means of series analysis (by such methods, which are presented in this branch), for them time filters are not applicable for obvious reasons. That's why this thread was born - to expand the sensitivity range of my systems and not to be limited to intraday trading.

 

So there is a difference in the code after all. What is it?

 
Youri Tarshecki:

So there is a difference in the code after all. What is it?

Different principles of determining the entry signal, different principles of position management, closing is also performed by different signals. Different systems - different code. But this is, of course, an intermediate stage. The plan is to merge the systems into one - to allow to accompany positions opened by another system if it does not contradict the current system, thus it is possible to reduce the costs of reopening positions by different systems.
 
We have focused a bit more on the flat and the trend topic has been left in the background. Shall we pay a little more attention to this part of the chart for completeness of the topic?
 
Uladzimir Izerski:
We have focused a bit more on the flat and the trend topic has been left in the background. Shall we pay a little more attention to this part of the chart for completeness of the topic?
Yes, of course, I don't mind. In terms of being interesting to study, both the flat and the trend are probably equally worthy.
 
Andrey Dik:
Yes, of course, I don't mind. In terms of being interesting to study, both the flat and the trend are probably worthy of study in equal measure.
Then we can discuss the transition from one state to the other otherwise blah blah.
 
Andrey Dik:
Different principles for determining the entry signal, different principles for maintaining positions, and closing is also based on different signals. Different systems - different code. But this is, of course, an intermediate stage. The plan is to unite the systems into one - to allow maintaining positions opened by another system if it does not contradict the current system, thus it is possible to reduce costs of reopening positions by different systems.
Then it is no longer possible to say that only TF has contributed to the improvement.