Stable MTS - page 20

 
Yuriy Asaulenko:

What does over-sitting have to do with it? We have a clear stop. The loss on a losing trade is strictly limited.

Equal time periods have absolutely nothing to do with it either. Entry is random. Exit at stop or un-limited profit. No one is forcing you to close a trade after T minutes or hours. Closing profitable ones is the simplest and not the best example: trailing stops.

Spreads, yes, get in the way, but if you put a stop >(3-5) spread, the spread is compensated by the profit.

You tried it wrong.)) When working even with an accidental process, you have to take their statistical characteristics into account.

This is where the whole game is about the correct support of the trade. I already wrote that I modeled it on FORTS to work out algorithms of transaction support, and TS was in stable profit - about 16% in 3 months from the transaction. Of course, this is not for real.

In turbulence you can easily get more than 20 stops (at a stop in 3 spreads), it won't pay off, even with a trawl. If you increase the lot, an unexpected pullback will kill the deposit.

Specify your algorithm, the way I understand it, in my experience it does not work

 
Сергей:
not really. If you take the probability of 50%, which is not true, then you lose 50% - spread, and earn 50% - spread. Total adding up we have 0 profit - 2 spreads

The losses on the spread (broker's commission) are more than covered by profitable trades. Let's say the broker's commission for a buy-sell transaction in stocks = 0.1% of the transaction value. This is more than the spread on forex.

If we do such a strategy on shares, then from earned 16%, we will give ~8% (half) to broker as a commission.

 
Yuriy Asaulenko:

The losses on the spread (broker's commission) are more than covered by profitable trades. Let's say the broker's commission for a buy-sell transaction in stocks = 0.1% of the transaction value. This is more than the spread on forex.

If we do such a strategy on shares, from earned 16%, we will give ~ 8% (half) to broker as a commission.

Could you tell us the algorithm? I must have misunderstood something.
 
Сергей:
Would you specify the algorithm? I must have misunderstood something.

It's no secret).

The first thing to know is the statistical properties of the instrument. Otherwise you may regularly hit stops, and on this basis it is chosen, so that random quotation fluctuations do not knock you out of the deal. It depends on the instrument. It may be picked in the tester, but it may not be profitable.

1. Look at the correlation function of the instrument and determine the interval of correlation. Time between deals should be random, but not less than the correlation interval. The long-short is determined randomly.

2 When opening a trade, place a stop at the selected level. In the simplest case, immediately include trailing stop. Trailing stop should be set at the maximum of the instrument. In other words, the stop starts to shift immediately at any upward movement. Since the stop is below the level of noise, random breakdowns are unlikely.

If you complicate trailing, you will get my results, but some profit will be there anyway. Actually, I repeat - strategy was made exactly and only for working out support of transaction.

Note that with each run, the deal will be different, because time and direction are chosen at random.

 
Vladimir Suschenko:
There are some specific features (to use a 90's slang, when people forgotten how to use their native language, "know-how") in the algorithm ... The balance may, in principle, for a short time "fluctuate" and within large limits, but you are correct to point out that the capital does not decrease. The algorithm has another peculiarity: large trade volume, it may give good additional bonus of some percents on accounts with rebates. With a large number of trades the algorithm shows stability - at any time interval the profit (though insignificantly) exceeds the loss:


The algorithm is very fresh, it has not yet had time to work in real life (the option "real ticks" has recently begun to implement, and it is precisely sharpened on the real ticks). I will "file it up" a little more, add some "trinkets" for convenience, and then launch it in real. At the same time, perhaps, will put it on the market. But then I will not be interested in INVESTORS as a class (logical, based on the meaning of the word investor?). Those who pay for a ready-made product are BUYERS - feel the difference...

By investors I do not mean those who pay for the development, but those who give capital to the account managed by the robot. You, as the developer, remain the manager of the strategy and have a share in the profits from the management.

Can you, without revealing the essence, describe the algorithm in a bit more detail. I am not interested in averaging, martingale or any other kind of oversaturation. I am good at sitting sitting without the robot.

 
Сергей:
not really. If you take a 50% probability, which is not the case, then you lose 50% - spread, and earn 50% - spread. So you have 0 profit, 2 spreads.
That's exactly what I said.
 
Yuriy Asaulenko:

It's no secret).

The first thing to know is the statistical properties of the instrument. Otherwise you may regularly hit stops, and on this basis it is chosen, so that random quotation fluctuations do not knock you out of the deal. It depends on the instrument. It may be picked in the tester, but it may not be profitable.

1. Look at the correlation function of the instrument and determine the interval of correlation. Time between deals should be random, but not less than the correlation interval. The long-short is determined randomly.

2 When opening a trade, place a stop at the selected level. In the simplest case, immediately include trailing stop. Trailing stop should be set at the maximum of the instrument. In other words, the stop starts to shift immediately at any upward movement. Since the stop is below the level of noise, random breakdowns are unlikely.

If you complicate trailing, you will get my results, but some profit will be there anyway. Actually, I'll repeat - the strategy was made exactly and only for working out of trade support.

Did you mean autocorrelation?
 
Oleg Shenker:
Did you mean autocorrelation?
Yes. By the way, you can't find it stupidly). It's no secret, but you have to write an article about it, which I don't want to do at all.
 
Oleg Shenker:

...Can you, without revealing the essence, describe the algorithm in a little more detail...

You want to know more than I can tell you. The secret you tell is no longer a secret and loses its commercial value. In brief and without unnecessary "fog", the essence of the algorithm is to "correctly" open positions.
Oleg Shenker:

....I should tell you right away, I am not interested in averaging, martingale and other types of overshooting.....

But it's much easier to describe that, which is not in my algorithm, I can tell you without fear - all of those things are not listed there. You must have some idea what charts of Expert Advisors with oversleeping and averaging look like - the graph of testing presented by me does not contain such parts. (Although, hypothetically, as a non-standard situation, I admit the possibility that we may have short-term over-exposures but they must not have a systematic nature. I can even say more - hypothetically I allow for some unexpected factors in the market behavior that can have a significant effect on the algorithm's results.) As for the martingale - I've already written that the testing was performed in the mode without reinvestments, with a constant lot, so if desired, you can easily verify its absence (martingale) using the tester's report.

 
Vladimir Suschenko:
You want to know more than I can tell you. The secret told ceases to be a secret and loses its commercial value. In brief and without unnecessary "fog", the essence of the algorithm is the "correct" opening of positions.

It is much easier with this one. I can tell you about what is not contained in my algorithm without fear - all of the above mentioned things are not present there. You must have some idea what charts of Expert Advisors with overshooting and averaging look like - there are no such parts in my testing chart. (Although, hypothetically, as a non-standard situation, I admit the possibility that we may have short-term over-exposures but they must not have a systematic nature. I can even say more - hypothetically I allow for some unexpected factors in the market behavior that can have a significant effect on the algorithm's results.) Regarding the martingale - I've already written that the testing was performed in the mode without reinvestment, with a constant lot, so you can easily verify its absence (martingale) using the tester's report.

Nevertheless. I need a general understanding of how the algorithm makes money. I do not need your know-how. But, the basic principles of operation are a must, I won't agitate investors for a black box. And one more thing... I still want to understand why the balance sheet fluctuates, but the capital doesn't. It seems very strange to me. That is, the market value of assets remains constant, but the financial result of closed transactions can fluctuate greatly. To me, it is an obvious sign that the algorithm artificially levels out the equity line, closing profitable or losing positions. It is not clear to me why it is doing that, and if it is not clear to me, it means that I cannot use this algorithm.

I think it would be more appropriate to move the communication to a private area.