[Archive] Learn how to make money villagers! - page 679
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
It seems that both (drawdown and profit) should be the same, limited only by the size of the deposit)
Nothing too extreme. However, we need to clarify how Igor calculates the Recovery Factor (RF) over 7 years, given the known annual RF (and probably drawdowns). If FS is calculated as the sum of the FS for each year (very roughly), it is about 4.3 per year. And this is prohibitive?
It is not very clear where the requirement of minimum reference to transaction history comes from.
Here - let's clarify the order of calculation, especially since he once wrote here that he reads the forum regularly, but posts rarely...:-)
No, you don't have to. This is the strength of the diversification effect. If the systems in the portfolio are statistically independent, it will be exactly as Igor wrote. Numerical modelling helps us understand this qualitatively. I'll post a picture when I find one.
And statistically independent is what is meant by that. I.e. if so, then we may come to conclusion, that even putting a loss-making system on as much uncorrelated pairs as possible - and here it is - a stone flower automatically).
Of course, every system in the portfolio must have a positive transaction m.o.s. But its drawdown can be significant.
P.S. I haven't found the picture yet. I'm looking for it.
Of course, every system in the portfolio must have positive transaction m.o.s. But its drawdown can be significant.
P.S. I haven't found the picture yet. Looking for one.
OK, I'm going to bed, I'll look at the picture later with interest)
This- instead of two-sided Ilan, which is ineffective, as a means to increase FS you can use exactly the diversification (the same Ilan))
Nothing too extreme. However, we need to clarify how Igor calculates the Recovery Factor (RF) over 7 years with known annual RFs (and probably drawdowns). If FS is calculated as the sum of the FS for each year (very roughly), it is about 4.3 per year. And this is prohibitive?
It is not quite clear where the requirement of minimum access to the trade history comes from.
Secondly, we have to ask about the FS calculation - if by year, as you write, with subsequent summation, then it is one figure, i.e. the longer the test period in years, the higher the FS. If it is a direct calculation over the entire test period, i.e. net profit/max drawdown = Recovery Factor, then it is a different figure...
For example, I, for myself the bar for TS for the real world is even higher, for combat owls on the real world FS > 35, clearly, otherwise TS is a waste...
Example, from my own research and development - Test combat owl to the full available depth of instrument history:
TC RECOVERY FACTOR = 36... :-)
P.S. And indeed, this point is interesting: "1. I'm writing an Expert Advisor with minimum access to trade history... - How should I interpret it?
The problem is not so extraordinary. True, we need to clarify how Igor calculates Recovery Factor (RF) for 7 years with known annual RF (and probably drawdowns). If FS is calculated as the sum of the FS for each year (very roughly), it is about 4.3 per year. And this is prohibitive?
Judging by his approach to the calculation of the total FC of trading systems, it seems that the FS of individual TS he also considers the sum of FS values of this TS for the year...
"OK... Here's a real-life example...
System 1 on the interval 10.01.2001-27.12.2007 has EF=42.88, instrument EURCHF
System 2 for the interval from 10.01.2001 to 27.12.2007 has FB=60,32, instrument EURGBP
Simultaneous use of these two systems for the interval 10.01.2001-27.12.2007 has FS=102.95".
This FS of the system number 2 = 60 - IMHO, generally a BEGINNING NUMBER!!..! if we get it by direct calculation of FS over the entire testing period ...
Although, if we take a pure optimization period and there to count, then, for example, net profit = 125 000, max drawdown 2, the FS = 62.5 - then it is something like a bare bones fit to the story ... I mean that on the real world this TS, it is not impossible that it will be unprofitable... I'm just saying it's possible that the system will fail in real time.
Well, I've looked at this grail and tested it. Figured out how it works. And let me say this is bullshit.
I had a similar grail, as I thought, only it was implemented differently, but the principle is the same.
This grail gradually accumulates an error that compensates the bank. It makes profit, but approximately one time in a month an error becomes critical and no more money. A full-blown disaster occurs.
When testing this grail they appear on November 17, December 17, and January 17. I do not know if it is a mechanism of self-liquidation or just a coincidence )))). Although the mechanism of self-liquidation I have not seen to the best of my knowledge.
And most importantly, I noticed that he brings prbl only not you but the broker even with 0.01 lot broker prbl reaches $25000 per month it when tested on M1.
The collapse did not happen on H1, for which it is better adapted, but that does not mean that it will not happen sooner or later. After all, on M1 it was, so anything is possible.
I tested it on euro/dollar.
Here is the picture:
The thin charts (5 pieces) are the balances of the systems in the portfolio. The trade volume in each system is always 1. A total of 330 trades.
The thin blue line is the result of diversification, i.e. averaging of all 5 trades at each moment of time. This is the sum of all thin lines divided by 5. The volume of each trade here is also 1.
The balance is conditionally shown as 0, though it would probably be better with a non-zero value.
We can see that relative drawdowns of the thick line are evidently smoothed in comparison to drawdowns of each "elementary" system of the portfolio.
The result of the trade was modelled as 1 + (rand()-0.5)*50. Here rand() is a random variable uniformly distributed from 0 to 1.
I did not make more lines, but the more there are, the better the result.