a trading strategy based on Elliott Wave Theory - page 14

 
Well, the fact that when testing the EA on M1 history (all ticks) the quality is only 25% - this issue has already been raised several times on this forum (read it). This is a reflection of the developers' approach on this issue, which contradicts the opinion of some people, including me ;o). As they say, I cannot answer for this indicator and all questions to the MT4 developers :o).

In no way did I want to claim in my strategy that I invented something original, which didn't exist before me! The approach I use is really a tweak to the historical data on the tester using the most trivial view of the market, which is already obvious and lies on the surface. The system does have an average of 10 trades a day in an attempt to take not the quality of trades but the quantity of them, as I've honestly given up on other methods that are widely used in forex :o( in terms of their application in MTS. And I hope only for statistical methods of work. If you do not know how the system will behave on real - I've given a clear answer - "I don't know, but I hope for a positive result and I'm just testing it on real account".

Thanks, I will definitely read the book. Surely there is a lot in it that I don't know. And I think it will be useful to me.

If you don't develop mechanical systems, but play by hand, of course we will never understand each other, because we speak different languages. I too, before I started programming MTS, had very different ideas about how it all should be. But then, when I started testing something, my understanding of the market and its methodologies has drastically changed (of course for the worse in relation to methodologies).

I have never claimed to be an excellent programmer (show me where you discovered that?)! Six months ago I asked a question on this same forum about how to identify an order from a comment ;o). I consider myself a pure amateur who learned to program some simple things from articles on this site and on www.mql4.com. And the vast majority of other programmers reading this forum surpass my programming knowledge by orders of magnitude!

Well, if you didn't like just a form in which I formulated questions ("childish naivety"), and it outraged you, then I apologize. I just firmly know the following - a person with some reliable knowledge will always answer any question in this area, no matter what form it is asked! And he can provide an answer both "on the fingers", as well as with more serious arguments. And Vladislav conducted himself in a very dignified manner in answering my questions, even though some of them were naive. A man who is confused by questions asked on his own strategy, most likely, simply does not know the subject, which he is trying to defend! And this simple psychological trick works almost without fail in real life as well.
 
Indeed, if you calculate where the positions of the system will be at its start a month earlier by history, it will probably be possible to find more convenient starting points for different branches of the system

solandr,
I don't think this is the right approach.
Trades should not be decided based on whether or not there are orders and where they are located.
If the situation is ripe for Sell, it should be placed and Buy, if any, should be closed.

It is clear that a certain strategy may use both lots and several one-directional orders of different values.
But I do not understand the trading principles based on the fact of orders presence or absence, while their random initial position with the possibility to correct the situation in a month is admitted.

This is a very peculiar approach.
If it is not a trade secret, could you please elaborate on this method with respect to the current state of orders?
Very interesting (no numerical figures are possible).
 
Действительно если подсчитать где будут находиться позиции системы при её старте месяцем ранее по истории, то наверное можно будет подыскать более удобные точки старта разных веток системы

solandr,
In my mind this is not the right approach.
Decisions on trades should not be made depending on whether or not orders exist and where they are located.
If the situation is ripe for a Sell, then you have to put it in and close the Buy, if there is one.

It is clear that a certain strategy may use both lots and several one-directional orders of different values.
But I do not understand the trading principles which are based on the fact of orders presence or absence, while their random initial position with a possibility to correct the situation in a month is admitted.

This is a very peculiar approach.
If it is not a trade secret, could you please elaborate on this method with respect to the current state of orders?
Very interesting (no numerical figures are possible).


Can you always tell exactly when to sell and when to buy? :o) I, for example, honestly say that I cannot do that! Of course it is possible to improve the system's chances of success in the first month after the start using additional methods, e.g. using Murray. But it is necessary to sit in front of the computer 24 hours a day and follow exactly when the situation is ripe to manually start each trading line on different timeframes! This is practically no different from playing manually. And moreover, no one can guarantee that this very entry point that will turn out to be profitable on the first day will be optimal in terms of getting a profit in further orders. Maybe, this was the last profitable trade in the series, and all the others will be unsuccessful? (After all, no one cancelled the banding of any strategy!) It seems to me that the probability of opening trades in the right direction manually using additional funds may turn out to be not much higher than 50%. Although I have not checked this method and I can only make unsupported guesses. And calculation of order positions in the Strategy Tester after a month to obtain the entry optimality is just a statistical fact which I also decided to share with the public (although nobody asked me to do it ;o)). I simply ran the strategy in the tester at different moments of time (but unfortunately on the same data sample, for which the optimization was performed :o( ) and usually the first month was less successful than the next ones. As a result, I concluded that the first month is a starting month, the task of which is to eliminate the consequences of a random start time of the system. I may be wrong in my assumptions and the system is ineffective in the first month of trading due to some other reasons. If someone has another opinion, I will be glad to listen to it.

To be honest, I did not quite understand your question about the details of the method concerning the current position of orders. If I did not describe the strategy's algorithm accurately enough, please tell me and I will try to explain it again.
 
Your question about the details of the method in relation to the current position of orders I honestly don't quite understand.

solandr,
What I don't understand is this.
How does trading in a future period differ from trading in the current period? From what you write, I can only assume that the only difference is the presence of orders. I do not see any other differences. Hence, I cannot but conclude that the success of trading somehow depends on the fact that there are orders. So I am asking if it is true or not. If it is true, please explain in what way. If not, why do you consider the first month to be more unprofitable (less profitable) than the next ones?

No one has cancelled the streak, that is true.
But in this case we are talking about something else. As far as I understood we are talking about full automatism. It means that the Expert Advisor has conditions for entering and exiting the market. These conditions must be equally applied to any historical period. In this sense, the first month's streak should not differ from the average streak predetermined according to the strategy.
 
.
 
SK, in general I have the following assumptions, which I am not entirely sure about. Let's take our strategy and reject using a lock, because as I said earlier, in principle it does not give anything, except increasing the volume of the used lot. Then the strategy comes to the following. For example, consider a period of H1. By the last closed candlestick, Sell and Buy Limit orders are set for the next hour period. If none of the orders triggered, then at the close of the next candle, the orders are modified according to the data just closed. This process continues until one of the orders is triggered and we enter for example into SELL. That is, we will have one SELL position and one BUYLIMIT position. Then we keep SELL until it reaches profit or loss. At the same time, of course, we continue to modify the BUYLIMIT position according to the above rules. Accordingly, the BUYLIMIT position will also have every right to close while we have a SELL position. That is, these trade directions are absolutely independent of each other. Now, an open BUY or SELL position can be held, for example, from a few minutes to several days until it is closed at profit or loss. So, at the time when we have an open SELL position, possible SELLLIMIT orders are not placed, and the same is true for buying. But we do know that while we are keeping SELL, a SELLLIMIT-order could have been opened, if we had not had an open SELL position, but only SELLLIMIT orders. And I have this purely speculative assumption that not every possible sequence of SELLIMIT-SELL trades is profitable during the very first trades. However, any such sequence will finally become more or less profitable after some time, for example 1 month. I do not know the reason why it happens. This is just my supposition that I cannot prove, except for simple calculation of profitability distribution by months from the start of the system. Anyway, that's what I want to do in the near future. I think I'll be able to post the results here a little later. Maybe they can even refute my assumption? And accordingly, if you sit near the monitor for a while and monitor the orders status (which we have SELL or SELLLIMIT) and at the right time, set them in accordance with the calculated position, you can probably increase the chance of the system success already in the first month. That is, we just need to enter the sequence of SELL-SELLLIMIT orders, which is more or less successful. So, according to my suggestion, such sequences in different trading strings will have already been formed (if we trade this month) or found in the tester (if we do not trade but calculate them by history) within a month.
 
As promised, here are the results of running the system by month at different times. For convenience, the system was started on the 1st of each month. The initial deposit was 1000. Increase in the lot size during the deposit growth has been disabled for the convenience of calculation. The strategy itself provides proportional increase of the lot size for each next thousand dollars of the deposit. That is 1000 is 0.1 lot, 2000 is 0.2 lot, 3000 is 0.3 lot, etc.
Indeed, based on the analysis of 11 system starts, the average profit in the first month was -12.7% less than in the second month of the system, if another system would have been launched a month earlier.
Also my assumption about arrival of a random sequence of SELLLIMIT-SELL and BUYLIMIT-BUY trades to the optimal successful one has been confirmed. If you look through the profit data by month shown in brackets in the table from the bottom left corner to the top right corner, you can see that the profit by month remains identical in many cases. Compare the green starts of the system. Also highlighted in red in the table is the system start that took 5 months to enter the optimal sequence of trades. But as a rule 1 month was enough within the available data sample.
 
solandr,
In general, the idea is clear.

But here...
..not every possible sequence of SELLIMIT-SELL trades is profitable during the very first trades.

Why do you make a distinction between initial and subsequent periods? What is the difference?
I understand that there can theoretically be edge effects in some technology. But I do not see them in this case.

I also do not understand the following. Why do we take bar characteristics as defining levels? Because the getting of a low-order candle to one or another candle of a high TF entirely depends on the time interval. It is enough to shift the countdown start, say, by 20 minutes, and the whole picture will look different. All the candlesticks will change. I think such a choice is random.

The idea itself is present and so am I.
But we should use something else rather than bar characteristics as levels for setting smart orders. For example, extremums of existing waves. And in further trading, we should proceed from the assumption that these extrema are the key ones and they will either be passed or the rate will not reach them.

Also, if possible, what are stop and profit values considered?
 
I think you can see in the table above the answer to your questions about the difference between the initial and subsequent periods for a given system. I am hardly able to explain in more detail and clearly.

The bar characteristics are based on the fact that they contain information about market movement and its direction from the point of view of the candle we have taken for the calculation. And we take only the last closed candle and do not take anything beyond it! Well, this is just a rule of the strategy and nothing more. You can take anything you want in another strategy :o). If you take something else for analysis than the last candle, it will be another strategy! I have not programmed this OTHER strategy, and I have no idea what it will be. If you build it, please, let us know the results.

Indeed, if you move the start time of a candle at some time shorter than the duration of the candle itself, we will need to recalculate all odds for setting start, stop and take profit points because the time pattern of the market changes, so the situation is completely similar to the fact that you have calculated the odds for one brokerage company and then started working for another one - all we have to recalculate anew. But the strategy works in this case as well. I already checked it.

Stops and profits are set on the basis of the same formula that is used for calculation of opening prices of limit orders but different multipliers are used. It seems the easiest to me. Although, you can experiment with other methods of profit and stop setting, for example, profit and stop loss fixed in pips. I am too lazy to check it, and I have no time, because I have other questions I want to check.
 
There is a rational basis in this methodology. I would change a few things though.

I have reread the latest posts, but have not found an answer to this question (if it is not a mystery).
How are the pending order levels calculated depending on OHLC?
Do I correctly understood that orders are simply placed by Open and Close, and stop and profit by High and Low respectively?
The speed formula is given at the beginning but it is not quite clear what is the use of it.