[Deleted]  

Good day, colleagues.

I will take the liberty of demonstrating here a trading system that was previously discussed by some of the authors as a perfect build, but was not completed in practice.

It is a system that guarantees no losses and no profit.

Let me briefly remind you the essence of the trading system. Let's assume for clarity that spread = 0. And examine one currency pair. Everything happens independently for each currency pair, so the result for a pile is simply a superposition of results for certain pairs. So, we open a pile of trades with TP=SL. What do we have in the output? Right, 50% of trades closed on TP, 50% on SL.


What follows are important assertions.

Assertion 1. If you with all of your indicators and reasoning fail to show profit in this geometry of experiment - you won't succeed in any other geometry, the draw-off is guaranteed. It is not even worth trying.

Assertion 2: If the core, which forms the decisions to "buy" or "sell" in this geometry of experiment "guesses" in such a way, that the probability of a deal closing on TP is much higher than the probability of closing on SL (even by a couple of percent) - what more do you need, here it is - the Grail.

I think if people think about it, the validity of both statements is obvious. The only question is to have a "core" that makes decisions with a "probability override" of at least a little over 50%.

To begin with, I would ask for an answer to a question like this. Can anyone (say on a demo account) in such an experiment geometry make the probability of SL greater than 50% and the probability of TP less than 50%? No way. No way you can get SL with probability more than 0.5. This problem is equal to the problem of getting a TA with a probability of more than 50%.

So 50/50 super stability is guaranteed. Better yet ... All in your hands - if you solve the problem - then the probability of TA you will increase (to demonstrate the correctness of the solution you can increase the probability of SL on the demo). So: the worst case is 50/50 with a guarantee, or only better. It can't be worse (if you know the way to deviate the probability, why would you use it against yourself)?

[Deleted]  
Dr.Drain:

_ Couldn't be worse (if the way of deflecting probability is known, why would you use it against yourself)?

50/50 can take a very long time to materialise. What are your suggestions?
[Deleted]  

Suggestion to break the ratio. There are two issues here:

1. Whether it is possible to do so.

2. If so, how. Obviously, any violation will have a positive sign for the trader. For if we find an algorithm that stumbles on the SL more often, we just need to invert the trades, and there will be more frequent TPs.

So question 1: Is it possible? What do distinguished colleagues think? :-)

[Deleted]  

I raised the topic in a threadat http://forexforum.ru/showthread.php?t=2587

Then closed it due to the deadness of the forum and the lack of interest in writing in the void.

The algorithm is demonstrated on a demo account:

Everyone can calculate the ratio of SL to TP for any more or less noticeable piece of history of at least a couple of dozen trades. The probability of TP is obviously greater than 50%. While this is working in real spread conditions :-)

[Deleted]  
Dr.Drain:
No one is stopping you from entering the market after a sufficiently large series with a sufficiently large deviation. This is what almost all normal strategies operate on.
[Deleted]  

No. You're being silly. I can satisfactorily paraphrase you as follows: You hope that after 10 heads rolls, there's more than a 50% chance that the 11th time will be a tails roll. No more than that. You can arrange an experiment in any mathematical package if you don't believe in logic (I repent, although I knew it had to be 50%, I arranged it.)

Roughly speaking, are you suggesting to test the chart in the past by the rule as-if-you-open-transactions-and-see-how-many-completed-by-TP-and-how-many-by-SL and draw conclusions? No way.

[Deleted]  
Dr.Drain:
I didn't say that. Firstly, not after 10. The minimum series size should not be below the size of the maximum possible historical series, so 10 is small. Secondly, I didn't say anything about an 11th fallout.
I was referring to a subsequent series, not a single trade. One trade is for martingale lovers, I am not one of them.
[Deleted]  

Dr.Drain:

Assertion 2: If in this experiment geometry, the kernel that forms the "buy" or "sell" decisions "guesses" so that the probability of a trade closing at TP is noticeably higher than the probability of closing at SL (even by a couple of percent) - what more do you need, there it is - the Grail.

I gave an example of the test. https://forum.mql4.com/ru/38834/page417 (seventh post). I will put it on Demo, when I finish all researches and calculations.

Grail or not - I don't know, but the preponderance is obvious. The series doesn't count here.

[Deleted]  

It's encouraging to see that many people are thinking the same thing. About the Grail being obvious. And the only question is the algorithmic core. My construction of it boils down to the construction of a lag-free filter.

http://forexforum.ru/showthread.php?t=2610

If it's no secret, how long have you been working on your algorithmic kernel, which should make decisions in the geometry of "many trades with TP=SL" experiment?

[Deleted]  

Roman100 at https://forum.mql4.com/ru/38834/page418

I did something similar, but I rejected it in the trash. It is simply based on the fact that the price returns, and that the chart on the daily timeframe is more of a global flat and the risks are calculated based on this, but no one guarantees the absence of a catastrophic bounce.

I think the system is useless.

Yeah, my thoughts definitely go together :-))) I also thought about "global flat", but rejected it at once. If TP=FL=50 pips or 100 pips, it will never work.

[Deleted]  
Dr.Drain:

If it is no secret, how long have you been working on your algorithmic kernel, which should make decisions in the geometry of the "many trades with TP=SL" experiment?

Relatively recently. There's still a lot of work to be done.

And how are you getting on? I still haven't figured out if you managed to get a stable MO outperformance at TP=SL at least on the tests ?