System Analysis for Successful Trading

Hello Everyone,

First, thank you very much for taking the time to read what I have written

I have been trading the forex market profitably for a few years (managing a private fund) and I have taken the personal challenge of training a few new traders myself. After having all this experience I have seen that all failing traders have one very important thing in common which can be easily fixed. They don't know what they are trading.

I was only able to become profitable in trading when I knew exactly what I was trading and what I could expect from the systems I was following. Trading doesn't work if you don't know what the long term targets and results of what you are doing are. Think about this, can you answer the following questions about your trading :

 

1. How much do you expect to make in average every year after 5 years ?
2. What is your expected maximum draw down ?
3. What is the expected average draw down period length ?
4. what is the maximum number of consecutive losses you will experience ?
5. Does the system have a positive mathematical expectancy ?

The fact is that traders who are already profitable will be able to answer this while traders who don't will have a very hard time doing so. The easiest way to become profitable is to trade in a way in which all the above variables are known. If you trade a mechanical or discretionary strategy for which the above characteristics for past performance are known then it will be MUCH easier to tackle the future.

If you trade your manual or discretionary strategy in a forward "live running" simulation amounting to ten years of results and you can then have an idea about what the system does, then obviously you will have a better chance at surviving profit and draw down cycles and getting to the other side. Every strategy has draw downs, every system has losing periods but the understanding of these is what makes a dramatic difference for a forex trader (speaking a lot for myself and my students here).

Time after time I have found that people go on and on in forums trading and modifying strategies that simply don't have an edge and will never work. How in the world can you trade something you don't understand ? How can you trade a system for which the expected profit and draw down values in the long term (years) are unknown ? I certainly couldn't do this and I believe that most people will also have a very hard time doing so. Traders usually cling to forum developed systems without proper evaluation for ages - usually when they are within profitable periods - but when the market exposure of these systems is cashed they quickly get forgotten and people lose interest. I always think that if they had only done a rigorous evaluation they would have found the strategy to be worthless from its very beginning.

The objective of this thread is to teach new traders the approach to trading by evidence and analysis. Design and do extensive historical tests of the discretionary or mechanical strategies you are using and then you will have a much better chance of reaching profitability. What do you think ? Would you join me fellow traders ?

Please feel free to leave any questions you may have

Best Regards,

Daniel

PS : My schedule doesn't allow me to get into forums very frequently but I will do my best to come in here and answer questions from time to time

I'm looking forward to reading what you have to say, I'm already firmly in the analysis/automation camp, but there are always new things to learn.

Hello Craig,

Thanks a lot for your follow-up I am glad you are interested in what I have to say. Definitely I want people to see that analysis is not something that is exclusive of automated trading and mechanical strategies but something that can be applied to discretionary trading as well.

I want people to see that analysis is NOT a "plus" but a basic part of success in trading, that knowing what trading a certain way is bound to achieve in terms of draw downs and profits is extremely important for success.

For me having a focus on analysis meant no more psychological pressures and the ability to trade with confidence. Thanks again for your contribution to the thread

Best Regards,

Daniel

I dont see how a discresionary method can be accurately forcasted to predict future results without relying on past data or performance.

Sounds a bit backwards to me

Anyway will be watching with interest

Hi CindyXXXX,

Thank you for your input Of course you would rely on past data to find performance, in the same way in which you rely on past data to get the past performance of a mechanical trading strategy. You would execute the strategy on a simulation environment (like the metatrader 4 visual strategy tester where you can get a live evolving chart) as you would in your regular manual trading, this way you can get an idea of the performance of your manual strategy under very long periods and large amounts of different market conditions.

What I meant is that people usually think that backtesting - especially extensive ones - are reserved to mechanical trading while the evaluation of a trading edge and the obtaining of draw down and profit targets through similar methods can be obtained for discretionary strategies. It is very important to do this for both types of trading styles as it gives the traders very significant information about their possibilities of future success. At least if no edge is found after extensive analysis on previous past performance then future profitability will be most likely nonexistent. While if such an edge is found the trader can have an idea about the potential profits and the depth and duration of the draw down periods as market conditions evolve.

Thanks again for your interest

Daniel

Good post. Hope it continues.
I believe a true edge comes from finding an enduring characteristic of the market you are trading and exploiting it to your advantage. This eliminates the need for backtesting. The enduring nature of the edge means that it was always there and will always be there as long as the market exists.

Most will not trade these edges because they cannot mentally handle the inevitable large drawdowns.

FYI, you are supposed to be in the commercial section of this forum.

http://www.asirikuy.com/

Well I have abstained from mentioning my website or my blog because I didn't want this to become anything commercial, as you see my intent here is beyond selling anything. I urge you to edit your post so that this thread can remain that way

Daniel

Thank you for your contribution The problem here is, how do you know there is a long term edge ? How do you know that your entries do forecast a certain movement in price with a higher probability during a long period of time ? The answer is through evaluation and in order to do this practically you need to evaluate your strategy across past data for the past 10 or better 20 years.

Of course, new traders here face a big challenge because without backtesting you are asking them to trade something, without knowing it has an edge and then get into deep draw down periods with the "hope" that the system will come out profitably. In my mind this is extremely difficult to do and the main reason why new traders simply keep on jumping between systems. The solution to this is in-depth evaluation through past performance analysis.

Figuring out the ten year average compounded yearly profit, maximum draw down and draw down period length of strategies allows new traders to get into a strategy knowing exactly what type of behavior they can expect in the future. Sure, the future might show worse results to some degree but they will generally know when this is the case and when the systems starts to become to risky to trade.

As an example if you run the simulations of your strategy and you find out that it has a 10% average compounded yearly profit and a 5% maximum draw down you will know that draw downs between 2-3% are regular, 5% are more rare and 6-7% are new maximum draw down figures. Knowing this gives you the power to trade without a great deal of emotion regarding the uncertainty of future results.

Thank you very much for your contribution

Best Regards,

Daniel

Sorry, but they are not my rules. The owner of this site has stated if you have a trading related business, regardless if you promote it or not, you belong in the Commercial section.

I would've never have known except your post in another thread set off some alarm bells for me.

IMHO there are two kind of questions here:

This question above can be (MUST BE) answered in advance by backtesting. Also the %win of trades, the ratio win/loss and blahblahblah

We could call them mathematic aspects of the system. Just numbers.

On the other hand, this kind of questions are more related to psychology than technicalities. Don't answer in advance.

A good system "by numbers" could be just impossible to deploy due to mental issues.

Will you stand the stress and distress of actually trading the system?
Will you really be able of sleep by supporting that drawdown?
Will you resist that draw down period length?
Will you resist that period without trading, just waiting?


For me:

step 1, find something with positive expectancy.

step 2, make sure you CAN trade it without going crazy. If not, go back to step 1.

step 3, then and only then, you can begin to answer all that questions and other related to MM or whatever.

Just my very personal point of view

Well I do understand your concern and I do understand what the forum rules say. However it is not my intent to make this or other threads commercial in any way. I would encourage you to edit your post to remove the link, PM the moderators about the issue and let them decide. After all I think that the rules are not that black & white and the moderators will likely see that my intent here is anything but commercial. You are welcome to stay here and check this to remain true for all the reminder of the thread

Thank you very much for your contribution and for guarding the forum rules

Daniel

I absolutely agree The measurement of mathematical expectancy is vital to know if your system has any potential and then the psychological aspects are also vital to know if you will be able to deal with a system and actually trade it in real life. To evaluate the psychological part I have created a quantitative measurement called the "pain index" which relates the 10 year maximum draw down depth and lengths and gives you an idea of how hard it will be to trade a given strategy (note that it gives no information regarding actual profitability)

MD = maximum 10 year draw down as a percentage of account equity

MP = maximum 10 year draw down period length in years

pain index = 2*(Log( MD^2 * MP))

As a matter of fact systems that have pain index values above 6 will be extremely hard to trade psychologically while systems with small pain values will be much easier to trade. Note here that such systems as martingales and scalpers have very low pain index values since they suffer sharp but short draw downs. This also explains a little bit why these systems are so popular regardless of their mathematical expectancy values and the risk they pose to account equity. It is dangerous when you have a system that is dangerous to account equity but easy to trade from a psychological perspective (low pain index).

What do you think about this ?

Thank you very much for your contribution

Daniel

It's good and useful to have a number to compare one system with another.

Nevertheless, in my experience, difficulty in follow a system must be tested "in your flesh" better than with mathematical formulas.

Why 6 or any other number indicates the limit where a system became hard or not? Persons are different, with different levels of stress resistance.

Anyway, another way to measure the pain:

sqrt((sum(drawdowns)^2)/period)

it measures the portfolio volatility

Summarizing: test it in your skin before continue spending time on it.

Telling a soldier "in the battlefield, you'll be under a pressure of 5.8 according our mathematical models" won't help him too much.

Thank you for your contribution SwissRaj, you are absolutely right about this. Definitely it depends on each particular trader and their actual personality to measure how hard it is to trade a particular system. Maybe a system I will be able to trade successfully will put too much psychological pressure on another.

I came up with the 6 value as - from my experience - most traders will find it extremely hard to follow such systems. However, as you say, there might be someone for who 6 systems might be easy to trade.

Nonetheless, I think that having some quantitative measurement is interesting since - even though you won't be able to say for whom it will be easier to trade which system - it does give you an idea of which system is easier from a psychological point of view than another. Maybe it is true that someone will find it easier to trade a 4 than another person but trading a 6 will be harder for both than a 4, regardless of their implicit perception of difficulty. That is, the higher the pain index, the harder a system is to trade. Maybe your threshold is lower - you might only be able to trade systems up to 3 or 4 - or higher - you can withstand 6 or 7 systems - but without a doubt trading a 6 will be way much harder than to trade a system that ranks as a 3.

All that said, it is just a tool and a way to give us an idea about psychological hardship. In your soldier example, it would be like saying to him/her : well, our mathematical models predict that you will be under a pressure of 5.8, while most soldiers can't handle 8 this might be easy for you since you succeeded when our mathematical model predicted a 6.5. I believe that this mathematical tool is very valuable

Thank you very much again for your contribution

Daniel

So, we agree and this point is perfectly clear

Next point...

Hi Daniel,

Nice article btw; thanks for sharing. A few points of observation...
1) I think (with respect to mikkom's comment) that the the article's system is still static, however, the system itself is 'kind' of adapting in the sense that it is dependent on a changing parameter (ATR nday avg).
2)NA
3) One thing I found a bit peculiar was that you investigated and developed your system over the most recent 4 yr. period and validated it backwards (on prior 6 years If I recall). Are you of the opinion that direction is not significant? Or, alternatively, do you feel it is better to begin developing your ideas on the freshest data, and working backwards to validate?
4) Along the same lines, I know many traders have different opinions about validation and robustness. What do you think is a sufficient period of data to draw conclusions over? I.e. in the article you use 6yr development testing period/4 yr validation. Rather than 1 period in sample, 1 period out, what is your opinion on several folded cross-validation?
5) What sort of measures do you specifically use to determine robustness?

tks. Jam

Refer to post #9: http://www.forexfactory.com/showthread.php?t=252484

I have sent a PM to Twee about the issue you have pointed out so that he can sort it out in the way he considers best. I am here to offer nothing but help and knowledge and I say - yet again - that I have made no move forward to attract any customers to any business or to sell people anything. Now Slack, I invite you to enjoy the thread which contains absolutely no hint of anything commercial besides that which you have provided

I however understand your concerns and I thank you for participating on the thread

Daniel

Hello JamJamJam,

Thank you for your post I will now try to answer your questions :


Well, yes, the system is static in the sense that its logic remains unchanged. A completely dynamic system - by definition - would be a system that changes its own logic, something that this system definitely does not do. However the system does provide adaptability - as you say - through the use of the ATR indicator. Such ways of providing adaptability are effective and many times much less complex and robust than dynamic techniques (like neural networks for example) as they run lesser risk of curve fitting strategies. You will find this is a very common approach in the way in which many professional positional traders build their systems.


The reason why I developed the system this way was because the reliable data I had available for the initial analysis which included the EUR/CHF, EUR/GBP and EUR/USD went back only till 2006. Due to this particular fact I developed the strategy in this "unusual way" first designing the strategy over the last four years and then validating it over the previous 6 (since the EUR/USD data was available from 2000).

I also think that direction is not extremely significant since what matters is the evaluation over varied market conditions. Since the market is cyclical - or so we assume - previous or like-previous market conditions will redevelop in the longer term and evaluating through previous years just serves this purpose as well as doing it the other way around as far as robustness is concerned.

As I pointed out in this case the development was done in this way merely due to the alignment with the previous study but what I usually do for system development involves an analysis of at least a 5 year period.


I believe that at least 5 years are needed to draw valid conclusions since this is the amount of time the market needs to go through what I consider statistically significant changes in long term volatility which I relate to changes in market conditions.

Regarding cross validation, I believe it might be a valid approach to system development but it is not what I do for my systems. I am what you would consider "old school" when it comes to system creation and evaluation. I believe that systems with simple adaptive criteria that use simple entries that have a statistical edge over periods larger than 5 years can remain profitable for very long periods of time. I follow a path to system development similar to that used by Richard Denis and similar traders in the early 1980s.


This is a very interesting and important question since it puts the spot light on the potential for long term profits under future market conditions. I personally determine the robustness of strategies by the following factors :

 

1. The ability to make profits during a 10 year testing period with no optimizations
2. The ability of many different settings to achieve profitability during a 10 year period
3. The "clustering" of profitable settings around the ones used. That is, if I am using a system with a Moving Average which is set to a 100 period I want the system to be profitable with values from 90 to 110.
4. Normal distribution of monthly return. When I build a system I want the monthly returns to be distributed as normally as possible over a 10 year period. Although "fat sides" always exist the degree of normality around monthly returns is a good hint of robustness.

I believe they might be a lot of other criteria I use to check for robustness which I don't recall right now but we will get to them as the thread evolves.

I hope I have answered your questions, thank you very much again for your contribution Jamjamjam

Daniel