28 !!! currency pairs, 1 expert. Another grail, but this one I think no one has ever shown. + DEMO ACCOUNT - page 8

New comment
 
HIDDEN:
Again, this is the reasoning of someone who does not make full use of autotrading.

I do not use autotrading to its fullest extent, and sometimes I make manual adjustments, but very rarely, because I have come to the conclusion that manual adjustments should not be used, as it leads to bad consequences.
 
HIDDEN:


What's the benefit? That you spent time and ended up with "crap" in a beautiful package. If you spent time on a really working expert and achieved a similar result, then congratulate you, but it's all games. In fact, the developers should take care that we do not get such results.


Now it turns out that even if you have written a normal expert, you are not so sure about the tester. And you can show a lot of things, both pictures and statements.


Granit77 try to test this EA on the same timeframe you are taking the High and Low from, I think you will be surprised.
I didn't get a "ga...noy in a pretty package", but a nice toy. Actually, life, cynically speaking, is on the one hand a process of satisfying the necessities of life, which in sentient beings goes hand in hand with obtaining pleasure. Money allows you to buy pleasure, I got it directly.
On the other hand, life is a game in which the value of chips is determined by the player himself, and one can treat it without too much fanaticism. One simply lives and gets pleasure even from small joys.
I cannot test the Expert Advisor on a higher timeframe, because the whole point is to work on M1, getting indicator data from M1 and the higher timeframe and comparing them. I will free up a little, replace High and Low with Close and everything will become clear.
 
Yurixx:

Thank you, I was very interested in your answer to this question. It turns out that the real Grail is possible after all! That's comforting. :-)

If you don't mind a few more questions. 1. If profit reinvestment was used in this test, could you post the test with one fixed lot ? 2. How much difference between the test results obtained using quotes of a broker and a real broker (whatever type)? If you can lay out a test on real quotes, it would be very interesting. I think it will be enough the last 1-1. 5 years of history. 3. The results of this EA on the real account correspond to the test results ? To what extent?


2 conys

This post was addressed to you. Maybe you just didn't notice it.

 
granit77:
I did not get a "crap in a beautiful package", but a beautiful toy. Actually, life, cynically speaking, is on the one hand a process of satisfying the necessities of life, which in reasonable creatures is accompanied by pleasure. Money allows you to buy pleasure, while I got it directly.


I get pleasure from buying good and high-quality products, from my car, from household appliances, from the women who are near me. And all of this is really about finances, some of which come from the market. And I really don't enjoy looking at virtual profits.

There's a great joke.

One says to the other:

One: I know what virtual communication is!

Second: What is it?

P: It's when you know a lot of people you talk to, but you can't have sex with them.

 
Yurixx:
Yurixx:

Thank you, I was very interested in your answer to this question. It turns out that the real Grail is possible after all! That's comforting. :-)


If you don't mind a few more questions. 1. If profit reinvestment was used in this test, could you post the test with one fixed lot ? 2. How much difference between the test results obtained using quotes of a broker and a real broker (whatever type)? If you can lay out a test on real quotes, it would be very interesting. I think it will be enough the last 1-1. 5 years of history. 3. The results of this EA on the real account correspond to the test results ? To what extent?


2 conys

This post was addressed to you. Maybe you just did not notice it.


Sorry I missed it, I must have been quick to add messages. 1) Reinvestment has been added specially to avoid running the EA every month of history.
2.They are different, because of that:




I think the third question is the same as the second one.

It does not affect the overall result, the calculation was also based on that, moreover the internet shutdown, the broker's "Sunhour", etc.
 

The second question was about testing in the tester, only on different quotes. And the third was about working on the real. Often the results on the real account are fundamentally different from the results in the tester. If there is no such difference, the testing was correct and it proves the fact that proper use of the tester gives proper results.

I recalculated the results of your testing in the game with one lot. It turns out that the average profit on a deal of $150-170 or 15-17 points depending on spread. Is it true or do you have different results?

I also want to express my gratitude. Your results also confirm that successful trading depends not on the quality of testing, quotes, channel or broker, but on the quality of Expert Advisor, so questions like "what TF to play?", "should I use ticks or not?", "where can I get good quotes? You just have to write good experts. :-))

 
Yurixx:

And the third is real-time work. Often the real results arefundamentally different from the results in the tester. If you do not have this distinction, it means that the testing was quite correct.


In the tester I used demo quotes of 1 fig, but on the real 2 fig the values are sometimes different.
For the first one I have two trades, but the second one has only one, because of different candlestick configuration. But it closed better than on the demo those two together...
 
HIDDEN:
If the tester has generated ticks from the quotes history with bid == Close of all timeframes, which is logical. But if the indicators use High or Low or all of them together, and the Expert Advisor trades taking them into account, then we have a clear peek into the future.
I have simply replaced all High and Low with Close in my Expert Advisor on high TFs without paying attention to my initial idea and was a bit puzzled with the results.
First, HIDDEN is obviously right in the discussion between Rosh and HIDDEN. The behaviour of the EA in the Strategy Tester has significantly changed and became similar to its demo behaviour. There are a lot of false entries in the flat, the policy of prevention of which led to creation of the grail77, but the profitability of the trend is still high.
Second, total profitability of the Expert Advisor, of course, has significantly decreased but still remains acceptable. Here is the comparative data with other conditions being equal (slash "EA with High-Low"/"EA with Close") .

EURUSD symbol
Period M1 (2007.01.01 - 2007.07.31)
Model All ticks
Lot=1 (fixed)
Initial deposit 10000
Net profit 106293/94121$.
Trades amount 328/775
Expected payoff 32/12 pips
Profitable trades (% of all trades) 92.07/46.19%

Third, due to being busy and irresponsible I have not performed a serious test of the Expert Advisor on a demo. I ran it on a chart and was convinced for a couple of days that it behaves differently in flat conditions than in the tester and moved to the next variant without statistics.

In light of the new data, the next step should be an assault on the last bastion of tester authenticity defenders - Rosh's assertion that the current Bid and Close of all higher timeframes are correct.
For this I put an EA that does not use any timeframe, except for Close of the symbol under test, on a demo today and run it for a week or two. Then I run the same period on history in the tester and compare the results.

2 Rosh
May I have some comments on the testing methodology?

First chart - EA with High-Low, second - EA with Close
 
granit77, let us play with it - post the source code:)
 
sashken:
Granit77, let us "play" - post the source :)


Exactly!

Otherwise, we'll also start posting pictures of our grails here.

12345678910111213
New comment