LTOptions at a 33% discount during the Year End Holidays.
Tell me More

Tuesday, March 13, 2012

Tips for reviewing Expert Advisors

Yes, we all know the vast majority of them are scams, or don't work in the end. However, successful automated trading is indeed possible with the right proven methods, that use sound trading principles and are backed up by serious statistical studies. One of the most common examples is that of the Turtles of Richard Dennis. Plus, most trading volume nowadays is done by algorithmic trading (I don't have the sources for specific numbers but I've read anywhere from 60 to 80% on the daily volumes on the NYSE).

Anyways, you might find the gem out there. It is tough, but possible. And here are my few cents on how to detect the potential of a Forex Expert Advisor, things you should expect and obvious red flag you must run away from.

1- Unrealistic claims
This is obviously the first red flag. Any claims promising "guaranteed" returns are a huge concern. There is no possible "guarantee" here.

Huge returns are the other aspect to look at. It is not uncommon to see experts promising you can double your money every month or every three or 6 months.These returns are unheard of in real Forex trading and in order to achieve them you have to expose your account to very high risks levels which are not sustainable over the long term. But what is a realistic return after all?

Looking at the Barclays Currency Traders Index you can see that the compounded yearly return of Hedge Fund Forex traders over the last 25 years is a non-breathtaking 7.33% as of this writing. And here's the interesting part: in order to achieve this, they had to put their capital under a risk threshold that led to a worst draw-down of -15.26%. This means that the relation between Yearly returns and worst draw-down is of around 1 to 2 for professional traders. In other words, in order to achieve a 1% return, professional successful Forex traders need to put their capital under a long term possible worst draw down of twice that value.

Simple math tells us that for a 50% yearly return, they could face a potential worst draw-down of 100% which would mean account wipe out. There are some traders and automated systems that have achieved a historical ratio of yearly returns to worst draw-down of 2 to 1. Which is by all means excellent and elite. And it implies that for a 50% yearly return you will have a worst draw-down of 25%. In this case, which is the most optimistic of all, such a trader/system would achieve a 200% yearly return as the higher cap, which would eventually result in a wiped out account when the worst draw-down is reached. And this is the most optimistic scenario of a system with an excellent 2 to 1 ratio of yearly compounded return vs Worst Drawdown, which is 4 times better than the average Professional Forex trader.

So, from now on, whenever you see an expert advisor promising returns north of 100% a year I would be really skeptical.

2- Lack of 10 year back-test
The backtest needs to be long enough so that the system is tested in all possible market conditions. A 1, 2, 3 year backtest is not enough in statistical terms given how much a market changes during a decade.

3- Lack of live trading proof verified by a third party
Live trading needs to be existing. It needs to be on a real money account. The vendor of the system carries the burden of demonstrating the viability of the system, and should trust his strategy enough to risk his own money with it. A Third party like MyFxBook is preferable to verify that the owner of the system is indeed using his strategy on his own money. Investor access to the account, with public credentials (for read only purposes) are also necessary in order to achieve total transparency.

4- Use of unsound trading principles (Analysis of the profit curve)
Systems that use progressive position sizes (Martingales) should not be used at all. They will invariably destroy your account. Systems that use some sort of Grid trading techniques aren't solid either, as they are based on absolute assumptions about the market that can't be made. Scalpers should be also ruled out as they use very unfavorable risk/reward set ups and oftentimes their backtests simulations are not reliable due to the exploitation of interpolation errors in the smaller time frames in Metatrader 4.

By looking at the profit curves you can quickly realize if the proposed Expert Advisor uses some sort of unsound trading strategy:



Grid traders

5- Unreliable testing.
This is another important aspect to look at. The problem here happens due to the fact that brokers don't save tick by tick data. They save minute opens and minute closes, and everything that happens in the middle is just generated at back test time through interpolation mechanisms. Obviously these "fake" ticks don't necessarily reflect the reality of what happened. (In fact almost never). Due to this reason and the mechanic way in which tick data is simulated, the developer of an expert advisor might exploit these interpolation errors (sometimes inadvertently) and obtain unrealistic profit curves promising exceptionally good returns. And this happens mainly on smaller time-frames such as those lower than H1 (1 hour).

Also for systems with small profit targets and small stop losses this can be a problem, as suddenly the spread plays a strong role in the outcome of the system. So, for a system that takes profit at 10 pips, the bascktesting results can be really good, as backtests usually adopt smaller demo spreads. The same system taken to real money trading under a spread of 3 pips would lose 30% of the potential profit which compounded means a drastic difference. Broker dependency and spread dependency becomes a huge issue.

Due to these reasons, I would say, stay away from systems trading under the H1 timeframe as their backtests are not reliable, and even if they were (by using Dukascopy real tick data), your system becomes spread / broker dependent anyways which is not good at all.

6- Lack of flexibility (adaptability)
If you see on the track record things such as fixed lot sizes, fixed stop loss distances and fixed TPs, run away. That is a rigid system which is unable to adapt to different volatility environments and is doomed to failure in the future. Solid systems adjust their stop losses distances and position sizes accordingly so they are in sync with current volatility environment while at the same time always risk the same amount of capital. This is usually done by using some concept related to the ATR (Average True Range).

7- Lack of statistical studies
For a serious system trader it is vital to know possible worst case scenarios that go beyond what backtesting shows. Profitable mechanical traders symply exploit a market inefficiency which could become non-existent in the future.  How can you determine the inefficiency you exploit is not there anymore? That's why you need to perform statistical studies. One of the most important being Monte-Carlo simulations that establish how bad things can go till you determine that the system is not profitable anymore.

Once this threshold is determined, you have to make sure that this threshold is reached before the wipe out of the account. This is by the way one of the problems of Martingales, when the worst case Monte Carlo simulation is reached it is already too late and there is no account anymore. You need to trade a system with a clear worst case scenario that indicates you have to stop using it, and on top of that you obviously need to have a worst case scenario that, when/if it takes places, your account is still far from a wipe out.

If you find a system that covers all 7 points mentioned, then you are in the presence of a potentially long term profitable automated strategy. Obviously, the vast majority of commercial $97 expert advisors out there don't comply with many of these rules.

Go to the bottom of this page in order to see the Legal Stuff

No comments:

Post a Comment