One of the most difficult decisions that every automated trader has to make is when to turn the system off because its performance is starting to be questionable and when to turn the system back on because it is getting back to profits. In this article, I will try to describe the way I see it.
First of all, I need to say that this is one of the most difficult questions in automated trading. In the past, I made a lot of mistakes by turning the systems too early off or by turning them too early back on. To make things even more complicated, out of many ways that I have tried, there isn’t one rule that would stand out (negatively or positively) among others. Therefore, it is important to pick one and never break it.
TURNING THE SYSTEM OFF
1. Turn the system off when it exceeds 1.5 times of the drawdown of your backtesting equity
I set this rule in my early beginnings. There are several important facts about it that I need to point out.
First of all, this rule is good and bad at the same time. It depends on the backtest equity you use. In the past, I preferred to pick one optimization parameter set and apply it to the whole data history. More recently, I have started using regular reoptimization, when I combine several out of sample periods (each with different parameter set) and create one out of sample equity.
Retrospectively, I must admit that in the case of one parameter set applied to the whole data history, this rule of 1.5 times of the drawdown wasn’t really the optimal solution. The equity of one parameter set was too “in-sample” - i.e. the backtested history was almost always better than live results (which is usual). Therefore I turned the systems off too early and experienced losses quite often - should I have the system turned on a little longer, the system would have, in most cases, recover.
But you get completely different results when you use equity curve composed of several out of sample periods - as part of regular reoptimization. This equity is far more realistic in terms of what future results you should expect. So far it seems that this equity, composed of several out of sample intervals, is really realistic and the rule of 1.5 times the max. historical drawdown works very well in this case.
2. To determine the moment when to turn it off, use Monte Carlo drawdown
Despite the simplicity of the concept described above, I prefer the second method - using Monte Carlo analysis.
Again, you need to consider if you work with equity that uses just a simple parameter set, or if you work with equity curve composed of several out of sample intervals.
If we use a single parameter set for the whole history, then I find the Monte Carlo method better than the rule of 1.5 times the drawdown. When using Market System Analyzer for Monte Carlo calculation, you will get drawdown much bigger than 1.5x the drawdown and you don’t turn off the system too early. Moreover, what is really important here is that Monte Carlo really makes sense as the distribution of your future profits will be every time unique and different from the previous ones. So I consider Monte Carlo as a fundamental (and for me a primary) tool.
Recently, I have started to incline more to using Monte Carlo, even on the equity composed of several out of sample periods. I agree that drawdowns that you will get using this method are not very nice. On the other side, the numbers will prepare you for the worst possible scenario, so that you can create your portfolio wisely and capitalize properly. This is the method I currently use. Though it is conservative, it matches my trading style.
Most of the time I use equity curve composed of out of sample intervals, I run the Monte Carlo Analysis, note the 95% confidence level and the maximum drawdown that I get there is the point when I turn my system off - in case it is exceeded.
This is the approach that makes the most sense to me.
TURNING THE SYSTEM BACK ON
1. Turn the system back on when the equity gets above the point when it was turned off
When can I turn the system back on? It is even more difficult question then when to turn it off - at least for me. Many systems come back to life and start being profitable again. I have experienced this many times. One of the rules you can follow is to note the point when you have turned the system off and turn the system back on when the system gets above this point. Usually, the strategy continues in the drawdown for some time after you turn it off, but then it starts growing up again and quickly gets to the point when you turned it off. This approach I consider pretty aggressive, so let me get to the modification of this method that I prefer.
2. Turn the system back on when it is “fully recovered”
For a long time, I have used a rule to turn the system back on when it is fully recovered and makes new equity high. This rule works pretty well, even though the recovery sometimes can take up a year, or even longer. Still, I brought back several systems back to live trading using this rule and I consider it acceptable.What bothers me on this approach is that is too “binary” and also the fact that the recovery is sometimes so fast and so profitable that you miss some really nice profits. But on the other side, there is the previous method, which is really too aggressive for me.So, what I find to be the best approach is the combination of both.
3. Combination of both using progressive position sizing
The rule is to turn the system back on as soon as it reaches the point when it was turned off (method #1), but start trading it with a minimum number of contracts. As the system recovers, we start adding some more contracts.
Let’s say we have traded this system with three contracts. As soon as the system gets above the point when we have turned it off (or some acceptable level above this point), we start trading it with 1 contract. If the system recovers to the half of the drawdown, we add the second contract. And if the system gets fully recovered, we add the third contract as well.
At the moment, I find this method to be the best one. Currently, it is my preferred way as it uses the best of both methods.
THE RULE OF THUMB
Whatever rule you decide to follow, the most important is to keep using just one rule. Be absolutely conscientious. I have a lot of students who lost a lot of money just because they didn’t turn the system off at the pre-defined point. They switched themselves to so-called “hope mode” and they started hoping that the strategy will turn up and start growing again. But this moment never came and their loss got bigger and bigger.
You must be uncompromising in keeping of these rules and comply with them to 110%. It is painful to turn off the system, we have spent a lot of time on. But this is why we have a portfolio - we will always have systems that will fail, despite all our effort. We are not in a secure business, we are in the business with risks that we need rationally and professionally manage and control. The good news is that it is possible.