Monday, April 26, 2010

The Draw Down Obsession... Why Stubborn Systems are VERY Dangerous

Through the past few days I have written a few posts concerning system development and why most automated trading strategies developed are not designed with the right statistical care and validity. Today I want to talk about a different aspect of system design which concerns a current problem which has obsessed EA sellers and free EA developers and which seems to be the core of most system development being done. This problem, which is the obsession of programmers with the elimination of draw down periods will be the subject of today's post. Within the next few paragraphs I will talk about this problem and how it afffects the trading system community by generating a series of "stubborn systems" which try to eliminate all market exposure - making the problem in fact far worse - than for regular systems.

We all know that draw down is not a good thing. You wake up, you turn on your computer and you see that your account is down some percent of its initial equity. You are losing money and losing is not a good sensation, you feel like a downright loser, you feel that you are doing things wrong and you don't want to run this system anymore because you fear it may drain your account even more. The answer to this instinct is to look for systems which have the smallest periods of draw down and the longest periods of profitability.

However you have to understand that the market exposure of a given system cannot be eliminated magically mainly because the future cannot be predicted. So for any given system you make, you will eventually have to pay the price of being wrong when you think the market will develop somehow and it develops in another way. People then try to make systems which are what I call "stubborn systems" which are reluctant to let the market cash their market exposure. These systems may avoid taking losing at all (no stop loss), increase their lot size after losing (D'Alemberts, Martingales, etc) or aim to win very little each time so that the odds of winning are extremely low (very unfavorable risk to reward ratio).

The truh is that these systems WILL have a high like hood of starting at a profitable place. It is very likely that people trading these systems for the first month or two will have a very high probability of getting into profitable trades. Draw down periods will almost never happen or - in the case of systems which do not take an exit or SL - they will only happen once, when the account is wiped. Moreover, systems with very small TP and very large SL values with very unfavorable risk to reward ratios are likely going to give unrealistic profit targets in backtesting due to the fact that one minute interpolation errors and the lack of execution problems - which do exist in live trading - will embelish their performance significantly.

Draw downs are not something that can be eliminated or made "very small". All the successful systems I have known in my life have extensive and sometimes deep draw down periods which are extremely difficult to handle for anyone trading them, reason why confidence and understanding becomes VITAL when handling these systems. In fact, holding a small draw down for a long period of time may be hard for most people. Imagine waking up and looking at a 5-15% draw down every day for one year, how would you feel ? Most people would say "this doesn't work" and they would archive the system when they probably don't even understand the system they are actually trading and its normal draw down / profit cycles.

I have found that this is one of the great defenses the market has towards everybody living from or using automated trading systems successfuly. Stubborn systems do not work but they are the systems most people are most likely to use while long term profitable systems are extremely hard to use and very non-rewarding (from a psychological point of view) and difficult to trade reason why only very few people actually use them. In the end, the people who search for stubborn systems and neglect to accept the fact that draw downs - sometimes even extensive and deep - are necessary to achieve long term profits and sustained growth will not be profitable while the few people who can accept this fact will in the end achieve success (not easily by the way !). What did they say about that person who laughed last ? :o)

If you would like to learn more about my journey in automated trading and how you too can design and program your own systems to achieve long term success in automated trading please consider buying my ebook on automated trading or joining Asirikuy to receive all ebook purchase benefits, weekly updates, check the live accounts I am running with several expert advisors and get in the road towards long term success in the forex market using automated trading systems. I hope you enjoyed the article !


Tcxmon said...


I agree with you regarding Martingale systems. I recently tested a commercial EA called "Forex Steal Pips" that traded USD/CAD on H4 timeframe.

The system had some decent test results, but I noticed everytime equity took a dip, the trade size would ramp up. Before long, I realized that it was a Martingale and that pretty much ended my evaluation of the EA. I agree that Martigales will eventually wipe your account.

That said, I do want to mention the concept of a "Martingale Lite". Suppose you had a system that had a very high % of winning trades. And suppose you programmed that system that after a loss, the system would double the position size (only once) until the system reached a new equity high, at which point the position size would return to normal levels.

That strategy would seem to improve the results of the system without all the risk of a full martingale.

I mention this because some popular commercial EA's use this technique and it definitely boosts their returns.

Let us know what you think,


Daniel said...

Hello Chris -

Thank you for your email :o) As a matter of fact ANY progressive money management technique ends up in wipeouts, exponential techniques like martingales do it faster while linear or other techniques take longer but eventually reach the same fate.

I wrote a post a long time ago about why Martingales or progressive money management systems can never be succesful despite the fact that they may have a "very high" profit rate. Usually these "high rates" of profitability are the consequence of the exploitation of backtesting faults (which make them appear larger than what they are) or they are simply a consequence of a very unfavorable risk to reward ratio which makes progressive money management even MORE dangerous.

Increasing lot sizes with loses is ALWAYS a way to "hide" market exposure and a poor alternative to sound ways of improving a trading system like developing a sound exit logic which may truly improve the strategy. EVERY system I have seen employing any of these techniques suffers the same fate, they eventually reach a number of consecutive loses which is high enough to wipe the account or, if the person wishes this number to be SO high that such a case is extremely improbable then profits turn to be minuscule. Always assume that you will get at least double or triple the number of consecutive loses in back testing as you get in live testing, specially for these systems. However, NEVER trust backtesting results until you have proved that live/back testing consistency is real as commercial systems -more often than not- try to sell systems which are very prone to changes in profitability by virtue of one minute interpolation erros and differences with live execution.

I hope this answers your question :o) Thank you very much for your comment !

Best Regards,


Tcxmon said...


Okay thanks.


townjet said...

Hey guys,

Good post Daniel. I think with any martingale strategy there is that looming disaster chance sitting there waiting. I have yet to see a system out there that can achieve a 92% or more accuracy without having a very skewed reward to risk. Taking a martingale strategy based on this type of system, you never know when you will hit two losses in a row...
One thing I never understood about claims to an EA being a martingale strategy is I always thought that with martingale, if you loose, you keep increasing exponentially until it wins... When a win finally hits you make a small gain, or your initial bet (Can do this on a roulette table for example.) This does not seem to correlate into an EA correctly when thinking about reward to risk as a factor. If I am 92% accurate with an average gain of 10 pips and a stop of 100, this being a 1 to 10 for RtoR. If I double up my lots on my first failure in regards to martingale strategy, in essence I am only going for 20 pips on the gain with a possible loss of 200 pips now. For true martingale results, I would need to increase my lots by 10 + 1, "not double", because I need to makeup the lost 100 pips and grab the 10 pips I am expected to grab. Doing this however would give me the possibility of risking 1,100 pips on this second trade.
If EA's really do implore this type of martingale strategy in this manner, I hope nobody uses it. In the long term... you may be happy with the current results but when that 2 in a row loss happens, look out. Being only 92% accurate you have 8 trades out of a hundred that will fail. You have 4 chances that you will get two in a row failure. I think the chances of 2 bad ones in a row may be small but it will happen eventually(may not even happen for a couple of years). With 100 trades you are already at 920 pips in wins minus -800 pips in loss. That +120 pips, even over a few years, will disappear fast with the martingale strategy as we noted above that true martingale would have lost -1,100 pips. You would be much better off only making the +120 pips a year and leaving it alone.


Daniel Fernandez's Expert Author Email Alerts
Sign up to receive email alerts of Daniel Fernandez’s latest articles from!

Email Address: