Martingale is Evil?!

 
While studying materials of this resource and one of English forums dedicated to automated TS development based on MT4, I noticed that there are discussions about the efficiency of this money management method in branches concerning application of Martingale method in money management. Naturally, as any inquisitive minded person should, I decided to find one of two opinions on this issue independently, without relying on arguments and reasoning given in arguments.

First of all, I tested existing Expert Advisors based on the Martingale principle. As expected, each of these EAs turned out to be ineffective in general, despite the difference between them in code and trading. However, we managed to find some common features among these Expert Advisors:

1. The use of the Martingale method (clear without studying it, but I will still note it).
2. Ineffectiveness in the interval of backtest from 01.07.1999 to 03.23.2007 for eight currency pairs without optimization of parameters, as a positive result obtained after optimization is not an indicator of effectiveness in most cases. Moreover, I performed optimization, but it did not increase effectiveness.
3. Trading only against the trend, or more precisely, doubling the lot in case of a price movement against the 1st level trade.
4. Absence of rules for determining the moment of entering the market (in terms of any analysis).

Everything is clear with the 1st and 2nd items. But points 3 and 4 make us think twice.

I decided to begin by writing a similar Expert Advisor, but one that trades by trend. Anyway, I'll explain further.
I'm not an expert in MQL4, in fact, I'm not familiar with it at all. But it was 5-6 years ago I was writing EAs in MetaQuotes program, which is an ancestor of MT4. It took me a couple of days to write it. Although, I had to pull some parts of the code out of other EAs.

The Expert Advisor works in the following way:
1. 2 stop orders, one to buy, the second one to sell with volume Lots at a distance Delta from the current price, with take profit Delta and stop loss Delta*2.
2. If one of the orders triggers, the second order is deleted. Here you should also place the order opposite to the triggered order, but at +/- Delta (depending on the direction) from the opening price of the triggered order, with the volume of Lots*3, with Take Profit Delta and Stop Loss Delta*2.
3. In case take profit is reached by the first (triggered) order, the second active order is deleted. 4.
4. In case of triggering of the second order, point 2 is repeated, but the volume becomes Lots*9.

The progression increasing the volume of the deal in this case is not exactly built according to the martingale method, but the principle is the same.
Besides, there are two variants of progression:
1. Increasing of the next value of the Lots parameter by three times: 1 - Lots*1; 2 - Lots*1*3; 3 - Lots*3*3; 3 - Lots*9*3; 4 - Lots*27*3, etc.
2. Increasing the next value of the parameter Lots to the number corresponding to the next odd Fibonacci value: Lots*1; Lots*3; Lots*8; Lots*21; Lots*55, etc.

As can be seen with the naked eye, this method of money management is even more dangerous than in previous Expert Advisors that use the Martingale method. However, due to the fact that, unlike existing Expert Advisors, this EA only needs the presence of movements without significant pullbacks or no pullbacks at all, its performance has surpassed that of the former ones. However, in the end the result was negative because of the periods of price movements in the killer range for any Delta parameter. It should be noted that during strong (or weak, but without strong pullbacks and corrections) movement, the Expert Advisor works regardless of the direction of the movement.

I think many people know that it is much easier to determine the moment to enter the market than to determine the direction of market entry at that moment. So, I decided to add rules to this EA according to which it should determine the moment to place orders. I investigated some indicators reacting to volatility changes but I put this idea aside because of my lack of knowledge of MQL4 I am not able to insert correctly the code of these indicators in the EA, which increases the testing time significantly as the backtest is performed by hands and eyes and there are a lot of indicators. Thus, I have stopped at trading on the news.

So I have modified the EA in such a way that it is now possible to specify the day of the week, hour and minute to start working. I.e., now the Expert Advisor is able to set orders before the news release. But this is half the trouble.
To determine the news release time and divide them into important and unimportant ones (my poor knowledge in the fundamental analysis and macroeconomics prevents me from doing it myself), I used the analytics of one brokerage company. Unfortunately the archive of this brokerage company has a calendar only for 2007.
However, after the backtest (performed on a deal per run) for 5 currency pairs in 2007, I found that during this period, there was not a single losing trade, and the maximum level of volume increase was the 3rd, ie (2nd version of volume increase) 8 lots at the start with 1 lot, and total - 1 + 3 + 8 = 12 lots. Moreover, the number of deals of 8 lots did not exceed 1/7 of the total number of deals.

Of course, I cannot speak about effectiveness of this EA and this method of using it. Firstly, it is because of the short testing period, and secondly, because of the pitfalls, one of which (and perhaps the main one) is the non-execution of orders by a broker exactly at prices set in them during strong moves.

Nevertheless, I will continue both backtest and forwardtest on demo. I will report the results here. I hope my investigations will help someone decide on whether to use the Martingale method without carrying out investigations and thus save a lot of time.
Maybe someone knows and prompts me the address of the economic calendar archive (with sorting by news that lead the market to some movement) for a period earlier than 2007. I would be grateful.
 

So what exactly is the problem? I didn't get it from your report.
I'm also experimenting with the Martingale method, it would be OK if I had 2-3 losses in a row, but sometimes I have 10.

 

1. In determining the exact timing of the news that will take the price out of the range, i.e. lead to a move greater than Delta*2, or not lead to a move greater than Delta.
2. In the availability of an archive (1999-2007) of calendars, solving problem 1.

 
DrawDown писал (а):

1. In determining the exact timing of the news that will take the price out of its range, i.e. lead to a move greater than Delta*2, or not lead to a move greater than Delta.

I think it is wrong to think that knowing the exact timing of news releases can improve your TS, so I suggest you read Another attempt to work on the news (posts by Lena). My opinion is based on the fact that the price reaction to news is very far from homogeneity, unambiguity and predictability. Fifty-fifty. The best case scenario is zero trading on the news. And that means that taking the news into account in other TSs won't do anything either.
 
Martingale + news trading = Road to Nowhere.
The impossibility of making money on martingale has been proven mathematically by Dub in the middle of the last century.
Just in case, you can check out this article that tries to explain it with your fingers.
'What is martingale?

After wrestling with this direction for a while you'll come to the same conclusions.

By the way here are examples of how the martingale principle works in CHAMPIONSHIP: https: //championship.mql5.com/2012/ru/news
https://www.mql5.com/ru/users/vixenme/
https://www.mql5.com/ru/users/foil/
 
KimIV писал (а):

... I justify my opinion by the fact that the price reaction to the news is very far from uniform, unambiguous and predictable. Fifty-fifty. ...fifty-fifty...
And I absolutely agree with you. BUT! For my research, or rather for my advisor, it is not important what the price reaction will be. What matters is the reaction or no reaction at all. I.e. the direction of price movement does not matter. Moreover, it does not matter that the price can make a so-called "Arrow", i.e., it can move in one direction, and then immediately in the other one. In that case, I have the advantage of the method. Even if the price makes two "Arrows", I still win regardless of the direction of the last price movement.

Yes, thanks for the link. The discussion in that thread is really what I need. I will continue the topic there. The reason we cannot have a constructive discussion here. The participants do not understand the essence of the idea, they just leave it without considering and discussing it due to the presence of the notion of "the Martingale method". Although the system is not based on the Martingale method but on statistics of price fluctuations on the news. I will post an analysis of price fluctuations, which are the market's reaction to news, conducted by FX Engines. It clearly shows that there is an opportunity to use these fluctuations for trading.

The only obstacle for trading using principles of this Expert Advisor is quality of brokerage companies. If there is no brokerage company that allows using an automated trading robot and executing orders at a given price at any time, we do not need such a company.

Now I am running the Expert Advisor on NFP. I do not need calendars at all times. So far the results are more than good, but it's still a backtest. When I am done with NFP I will let you know the results.
 
DrawDown:

...Moreover, it also doesn't matter that the price can create a so-called "Arrow", i.e. it can go in one direction and then immediately in the other. In that case I win by using the method . And even if the price makes two "Arrows", I still win regardless of the direction of the last move.



What if there are six or eight arrows?

Martingale can probably be used if the probability of profitable trades is > 90%. Or in other words, the probability of a series
of 3-4 losing trades is negligible. But if there is a TP with 90% of profitable trades - why use martingale?

The strategy described above is called a volatility breakout.
As a rule it is not related to the news or martingale.

From my experience in the market I figured out an iron rule - close all positions before
Close all positions before the news and wait for them on the fence.
 
:) News again... Yes, it's tempting, especially when you see how the pound makes 100-120 pips in 5 minutes, and you think, why not throw two pending ones a minute before the news...
Let's just say I even earned a little money with this method.

I made MTS for downloading calendar for a week and then I put two pending orders on 4 currencies (depending on what country's news) one minute before release. I also filled out history and adjusted such parameters as distance to price at the moment of placing order, order lifetime after news publication and takei stops and trailing. It was not too much, but it was enough for bread and butter. Then I placed it on a demo and started watching. This is what happened.



is attached.

So I drew the conclusion that it was a lost business, although to be fair I must say that I have only differentiated the news by importance and quantity of simultaneously published
news, but I didn't have desire to do it specifically by types of news.
I will keep the calendar for the whole 2006 as an additional help.
Files:
 
New:


What if there are six or eight shooters?

Yes. In that case there will not be enough funds to open another order with a progressively larger number of lots, which will lead to losing the deposit.

I was hoping that at least on the news the movements would be more or less constant, even with two or three "arrows". But alas.

From 1999 to 2001 it was like that on the non-farms. I used to open 4 orders at the most because of fluctuations, and using martingale allowed me to close the orders with profit. But from 2001 to 2007 I had two times of fluctuations in the form of several large differently directed candlesticks. Such a "high fence". Naturally, this is enough for a loss.

Also on the backtest of movements caused by the CPI data from 1999 to 2007 the possibility to withdraw the deposit occurred 3 times.
On FOMC policy news - 2 times.
On PPI - 4 times.
Once on Housing Starts - 3 times.
It has occurred 4 times in Trade Balance.

Only Initial Claims and Durable Orders increased the number of orders up to 4 at the most and allowed the deposit to grow over 8 years, but this fact did not encourage me and I did not waste my time investigating other 16 news that lead to significant price movements.

My conclusion is as follows: you can use the martingale method only when there is a possibility for false entries in the market provided that re-entry will be correct. I.e. increasing the volume to the 2nd (maybe 3rd at most) level of progression. Otherwise the efficiency will have practically no chance to meet the requirements of a successful trading system.
As for trading on the news, I realized upon close examination that the price reaction can be not only uncertain, but more than uncertain. And it is impossible to use this reaction for profitable trading even in the medium term (unless of course you have access to insiders).

Thank you all for sharing your opinions.

 

Leverage 1:500

Working lot 0.01

Spread 8 pips

Stop loss 23 pips

Take Profit 23 pips

1 pip centilot price $0.085

GDP/JPY

------------------------------------------------------------------------------------------------------------------------------

step 1 2 3 4 5 6 7 8 9 10 11

------------------------------------------------------------------------------------------------------------------------------

probability 1:2 1:4 1:8 1:16 1:32 1:64 1:128 1:256 1:512 1:1024 1:2048

-----------------------------------------------------------------------------------------------------------------------------

4$ 8 $ 16 $ 32 $ 64 $ 128 $ 256 $ 512 $ 1024 $ 2048 $ 4096

-----------------------------------------------------------------------------------------------------------------------------

step volume 0.01 0.02 0.04 0.08 0.16 0.32 0.64 1.28 2. 56 5. 12 10.24

-----------------------------------------------------------------------------------------------------------------------------

step loss 1.96 3.91 7.82 15. 64 31. 28 62. 56 125. 12 250. 24 500. 48 1000. 96 2001. 92

----------------------------------------------------------------------------------------------------------------------------

0.01 lot minimum initial deposit for trades* 381.12 762. 24 1524.48 3048.96 6097.92

--------------------------------------------------------------------------------------------------------------

*Margin call is not relevant in this situation when calculating the initial deposit.

Proof of "flush" based on probability theory:

Probability of falling out of the 11 steps is 1:2048, more precisely 2037 without the same 11 steps - ie for 2048 transactions account for one losing series of odineteen steps, consider the ideal case: Profit one profit(23 points) 1.96 $ ideal case is when all transactions except one losing series(11st) profitable 2037 * 1, 96 = 3992 $ Profit 3992 $. When the loss of a series of 11 steps is $ 4004. Total: -$12 (2048 trades)

Doctor-pound

 
evil...
I haven't delved much into ... but ...
The main thing is not to throw a lot of money at a trade, but don't go too far with it.
The main thing is not to invest too much money in a deal and not to chase super profits.
The main thing is not to throw too much money at a deal and not to make too much profit.
The trading strategy is a high-risk tactic, and even if there is a profit, in the medium term it will most likely be a zero-sum game (or possibly a negative one).
not to mention long-term work.
i.e., there is no imagination ))))