Reduce drawdowns to a minimum! Any ideas? - page 5

 
Itum:

Nefedov Kirill | 6 Apr 2010 at 21:25 RU
Parameters AbsoluteDrawdown, MaxDrawdown, MaxDrawdownPercent, RelDrawdownPercent, RelDrawdown are calculated incorrectly, they do not coincide with the build 226 tester report
Do a better math.
 
Itum:

Nefedov Kirill | 6 Apr 2010 at 21:25 RU
Parameters AbsoluteDrawdown, MaxDrawdown, MaxDrawdownPercent, RelDrawdownPercent are counted incorrectly, they do not coincide with the report of tester build 226
Why do you need to determine it yourself? You don't trust it?)
 

The drawdown of funds is best calculated independently, because if the position is in the plus and this plus is greater than the minus for all time (modulo), and the position did not close in the plus, but returned to the original and closed at breakeven, the maximum drawdown will be this, in fact a positive transaction.

Ideally we should separate positive drawdown and negative drawdown.

 
-Aleks-:

The drawdown of funds is best calculated independently, because if the position is in the plus and this plus is greater than the minus for all time (modulo), and the position did not close in the plus, but returned to the original and closed at no loss, then the maximum drawdown will be this, in fact a positive transaction.

Ideally we should separate positive drawdown and negative drawdown.

But why?

Why not in a simple way?

Out of the market - balance, the drawdown is zero.

When in the market - the drawdown, and only negative, because there is a drawdown called the stop-out and it is this drawdown value that is fatally critical.

We run in the tester at minimal lot and look for the maximum drawdown, which equals the maximum loss (again, at minimal lot, and there can be a lot of lots). This value of the loss is a characteristic of the TS with the name "drawdown". The longer the time period, the more positions were opened, the better. This results in an interesting nuance.

Suppose we start with 1000 quid, gain profit with a small drawdown and with a balance of 2000 quid gain 1001 quid. It would seem that the drawdown is 50%. But, no, wrong. We have a sinker - just lucky that 1001 quid drawdown did not occur at a balance of 1000 quid. The drawdown we found on the story can occur at any time on the real.

Not satisfied - we detail the reasons and try to buy it - this is the subject of the branch. And initially we need to define the meaning of the word "drawdown" and not make it up, because there is a drawdown hanging over everyone called "stopout".

PS. All these considerations are correct if we believe that what we got in the tester will be the same in the future, but this is the main problem. that is why signals, PAMMs and real accounts are lost. And only the one, who has proof, that the drawdown, obtained in the tester, will be the same in real trading, trades profitably.

 

Absolute drawdown is a drop of funds in relation to the initial value. For example, if you have deposited 10 000$, then any decrease of funds below this level will be considered as an absolute drawdown. If funds will always be above 10 000$, then absolute drawdown will be equal to zero.


Relative drawdown is the decrease in funds relative to the maximum value. It is usually calculated as a percentage. For example, you deposited your account for $10,000 and lost $2,000 within a month. In this case the relative drawdown equals 20% (2 000$ from 10 000$). If you funded your account with 20 000$ and lost 2 000$ during the month, the relative drawdown will be 10% (2 000$ of 20 000$).

 

About how drawdown is calculated:

Что означают цифры в отчёте тестирования эксперта
What the figures in the examiner's test report mean Introduction. Any Expert Advisor can be tested on historical data. Once the Expert Advisor has been tested, the 'Report' tab will display the Expert Advisor's generalized test results and some key figures. The reports allow you to quickly compare the performance of different EAs, as well as the performance of the same Expert Advisor with different input parameters. This article helps you learn to read such reports and interpret their results correctly. Example of a Test Results Report. As an example, let's consider the following report of test results: Bars in test, the number of bars in the history, shows the depth of history upon which the modeling was performed. Ticks modelled, the number of modelled ticks, shows the size of the modelled sequence. Each entry of a sequence represents a bar condition (OHLCV) at a certain moment of time. Depending on the timeframe, modelling method and availability of historical data...
Articles | 2005.12.21 10:43 | MetaQuotes Software Corp. | Tester | MetaTrader 4


 
СанСаныч Фоменко:

What's the point?

Why can't it be done in a simple way?

Outside the market - balance, the drawdown is zero

In the market - the drawdown, and only negative, as there is a drawdown called the stopout and it is this drawdown value that is fatally critical.

We run in the tester at minimal lot and look for the maximum drawdown, which equals the maximum loss (again, at minimal lot, and there can be a lot of lots). This value of the loss is a characteristic of the TS with the name "drawdown". The longer the time period, the more positions were opened, the better. This results in an interesting nuance.

Suppose we start with 1000 quid, gain profit with a small drawdown and with a balance of 2000 quid gain 1001 quid. It would seem that the drawdown is 50%. But, no, wrong. We have a sinker - just lucky that 1001 quid drawdown did not occur at a balance of 1000 quid. The drawdown we found on the story can occur at any time on the real.

Not satisfied - we detail the reasons and try to buy it - this is the subject of the branch. And initially we need to define the meaning of the word "drawdown" and not make it up, because there is a drawdown hanging over everyone called "stopout".

PS. All these considerations are correct if we believe that what we got in the tester will be the same in the future, but this is the main problem. that is why signals, PAMMs and simply real accounts are lost. And only the one, who has proof, that the drawdown, obtained in the tester, will be the same in real trading, trades profitably.

Very interesting comment) And how do you buy the drawdown and do you work with drawdowns?
 
Dmitriy Ermolaev:
Very interesting comment) And how do you buy the drawdown and do you work with losses?

With moose? No way.

Trading is trending, so I'm working on entry/exits to/from the market. I get the characteristics of the TS in the tester on the minimum constant lot. Many lots may be open at the same time. The main information on the results of testing for me is the profit factor and the maximum current loss that is calculated in pips. I make efforts so these figures do not change in the future the TS should be stable (robust). I place stops after the maximum loss. If I missed a stop, it means the TS is not working. On October 7 I got a loss for the TS, which was a year on the real account and a tester in three years. Now the TS is being reworked, i.e. there are no lots in normal trading.

That's me, but there is a TS that has stops as a working tool. So I have one of the possible options.

 
СанСаныч Фоменко:

Why?

In the terminal, the drawdowns of the two types are combined into a total drawdown, which in some TS leads to a significant distortion of the results, and does not allow making correct preliminary conclusions about the interaction of price and TS.

SanSanych Fomenko:

Why it cannot be done in a simple way.

Outside the market - balance, the drawdown is equal to zero

When in the market - the drawdown, and only negative, because there is a drawdown called the stop-out and it is this drawdown value that is fatally critical.

We run in the tester at minimal lot and look for the maximum drawdown, which equals the maximum loss (again, at minimal lot, and there can be a lot of lots). This value of the loss is a characteristic of the TS with the name "drawdown". The longer the time period, the more positions were opened, the better. This results in an interesting nuance.

Suppose we start with 1000 quid, gain profit with a small drawdown and with a balance of 2000 quid gain 1001 quid. It would seem that the drawdown is 50%. But, no, wrong. We have a sinker - just lucky that 1001 quid drawdown did not occur at a balance of 1000 quid. The drawdown we found on the history can occur at any time on the real.

The easiest variant is to determine the maximum allowable drawdown, and then mark it out, it can be determined in two ways for the tester:

1. The initial deposit equals the maximum drawdown allowed

2. Making a function that closes the position when the maximum allowable drawdown is reached

You should allowable drawdown to be determined before you start optimization.

If you simply search for the maximum loss, it cannot be guaranteed that it was actually maximal during the entire testing period. For example, in a trend Expert Advisor, I use the algorithm for closing a position in the wrong direction that is actually closed during rollback, i.e. there is a difference between the actual drawdown achieved and the drawdown fixed using your methods.

Conclusion - for different ATCs there may be different ways to determine the maximum drawdown, and they differ in accuracy and complexity of implementation.

Besides, it is very important to know the frequency of the maximal drawdowns (limited by the maximal drawdown allowed) for money management purposes - there is no sense to keep the deposit under the drawdown, which occurs once in a quarter - it makes more sense to predict this drawdown and increase the deposit for this strategy when the dangerous situation is approaching.

 
I liked the thesis....If it's out on the stop, then the TS is not workable....I agree!