Don't tell me then that TA doesn't work - page 22
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
And I was really amused by the test result posted a few pages ago, where the start-up capital is 10 million.
:))))))
It seems to me that all this nonsense in the thread is not innocuous. At least from Reshetov's mouth. I support the provocation, because Reshetov should know that any conclusions about random variables must be accompanied by the amount of confidence in those conclusions. He has made three runs, has not calculated a value of confidence in these conclusions (which cannot be done on the basis of three realizations) and is pushing words instead of specifics.
A major shortcoming of TA is the lack of even an attempt to specify the confidence level of TA conclusions, so we have to refer to whiners, gundos and flunkies.
And I was really amused by the test result posted a few pages ago, where the start-up capital is 10 million.
:))))))
There's only one test I posted, there's an initial deposit of 1000 quid:
Gee!
I took Reshetov's Expert Advisor and processed/optimized it with Reshetov's trading robot for the period 2010.08.22 - 2011.02.22
I ran the Expert Advisor through the parameters obtained using this optimization on the period2009.01.05-2009-12.31, with an initial deposit of 1000 and 0.1 lot:
And I haven't inserted or corrected anything anywhere !
It seems to me that all the nonsense in the thread is not innocuous. At least from Reshetov's mouth. I support the provocation, because Reshetov should know that any conclusions to random variables must be accompanied by the value of the credibility of these conclusions. He has made three runs, has not calculated the value of confidence in these conclusions (which cannot be done on the basis of three realizations) and presses with words instead of specifics.
Wapcheta he has put together a system that doesn't limit you on the number of runs. Why don't you compile the statistics yourself already? :) I'd like to refute it with the numbers in hand. Your unsubstantiation looks even worse than Reshetov's. And vapche, to present him with any requirements to prove something to you, I think it is incorrect. The man put forward the idea, and even illustrated technically, is not enough for you to continue to roll up his sleeves (if you consider it promising)? You may even ask him for seed money. ;-)
The main disadvantage of TA is the lack of even an attempt to indicate a level of confidence in TA's conclusions, so you have to refer to whiners, gundos and flooders.
Fill in the gaps. Or at least try to.
// Otherwise you'll also have to be reckoned to the categories of "whiners, gobblers and flunkies" mentioned by you.
:)
Wapcheta has put together a system that does not limit you in the number of runs. Why don't you get your own stats? :) So you can refute it with the numbers in your hand. Your unsubstantiation looks even worse than Reshetov's. And vapche, to present him with any requirements to prove something to you, I think it is incorrect. The man put forward the idea, and even illustrated technically, is not enough for you to continue to roll up his sleeves (if you consider it promising)? You may even ask him for seed money. ;-)
Fill in the gaps. Or at least try to.
// Otherwise you'll also have to be lumped into the categories of "whiners, gobblers and flunkies" mentioned by you.
:)
Making up for it.
Took an ordinary TS, working on the crossing of two MAs, a fast and a slow one.
I did everything according to Reshetov, one to one, with his same three periods, the last of which ends today.
I ran the obtained EA for the whole year 2010. I attach this EA (T_Rest.mq4, Rest.mq4 - put it in ...expert\include\).
Here are the results:
==============================================================================================
The result is kind of a no-brainer. But on the other hand, I don't think anyone has managed to get this particular strategy
a year-long non-flush OOS test.
The result is sort of a no-brainer. But on the other hand, I don't think anyone has managed to get this particular strategy
non plum OOS - a year-long test.
From 1999 (11 years). Sample-periods are shown. There are three of them, as the expert with three perceptrons (attached in the trailer).
I am happy that the duration of the profitable OOS is more than 7 years, but it is tense that almost all of it is behind the course. On the other hand the TS itself is lousy, you can't expect much from it. I expect much better results with more reasonable TS (not with two wheels, of course). However, this result is not bad even for the "front" period of OOS - the system keeps the spread very satisfactorily.
Since 1999 (11 years). Sample-periods are shown. There are three, as the expert with three perceptrons (attached in the trailer).
I am happy that the duration of profitable OOS is more than 7 years, but it is straining that almost all of it is behind the course. On the other hand the TS itself is lousy, you can't expect much from it. I expect much better results with more reasonable TS (not with two wheels, of course). However, this result is not bad even for the "front" period of OOS - the system keeps the spread very satisfactorily.
Perceptrons are good, since 1999 (11 years) - impressive, if it is no secret, what is the duration of the Sample-periods shown ?
But let's take for a trend strategy, let's say 3.5, even 10 Sample-periods.
And what if even one Sample-period will not show a trend, but other Sample-periods show trend,
then this signal should be rejected ?
In short, there is still much to think about.
It is also disturbing just to have a qualitative explanation of the resulting effect.
1) if it is not a secret, what is the duration of the Sample periods given ?
2) Let's take for trend strategy, say 3.5, even 10 Sample-periods.
And what if even one Sample-period does not show a trend, while other Sample-periods show trend,
then this signal should be rejected?
1. three months each.
2. This is the key question. If you look at the attached Expert Advisor, you will find trigger parameter that allows you to change properties of the filter. When trigger<1, the system of 3 indicators operates in the mode of summing/voting, when trigger>1 it operates in the mode of "logical multiplication" of signals. During the experiments the ambiguity was detected. For the most part, the multiplication gives more profitable deals (but their number decreases, of course), but there are exceptions (though not many). In practice the reduction of signals quickly becomes a real problem when increasing the TC committee (at the speed of 2^n, where n is the committee size). Therefore I have planned to use some mix of addition and multiplication, realized by threshold cutoff of low-coordinated signals. Threshold value will be chosen experimentally, though at my leisure I'll make some theoretical considerations as well.
1. three months each.
2. This is the key question. If you look at the attached Expert Advisor, you will find the trigger parameter that allows you to change the filter properties. When trigger<1, the system of 3 indicators operates in the mode of summing/voting, when trigger>1 it operates in the mode of "logical multiplication" of signals. During the experiments the ambiguity was detected. For the most part, the multiplication gives more profitable deals (but their number decreases, of course), but there are exceptions (though not many). In practice, when increasing the TC committee, the decrease of signals quickly becomes a real problem (at the speed of 2^n, where n is the committee size). Therefore I have planned to use some mix of addition and multiplication, realized by threshold cutoff of low-coordinated signals. I will select the threshold value experimentally, although I will also make theoretical considerations at my leisure.
I am watching your Expert Advisor now. I will try to display the signal only when all the three TPs confirm it.
Although, I think I need to gather some statistics on more understandable TS in comparison with perceptrons,
for example, on the TS working inside/outside the channel - everything is clear here, support/resistance levels are easy to interpret,
so decisions will be easier and more natural to make.
I'm watching your EA now. I will try to give a signal only when all three TS confirm this signal.
Although, I think we need to get some statistics on more intelligible TS in comparison to perceptrons,
For example, on the TS working inside/outside the channel - everything is clear here, the support/resistance levels are easy to interpret,
so decisions will be easier and more natural.
Uh-huh. This is the result for signal addition (trigger=0)
But for the logical multiplication of signals (trigger=2)
Both results under otherwise equal conditions (pair, timeframe, optimisation periods etc). Same 11 years.