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

BookingAlpha Option Trading Advisory

Sunday, August 11, 2013

Forex Rebellion doesn't work

If you came to this article it means you probably know about the Forex Rebellion system. It's nothing new. In fact it came out a few years ago. There are certainly several positive reviews out there, many of them by affiliates which doesn't really count (read this article I wrote in 2011 on how to spot affiliate reviews), but there are many other positive reviews by real users. The reviews generally tend to be positive because the system advocates the use of stop losses, protection of profits, letting winners run etc, which I agree are good lessons. But what about real results? Does this trading system make money or not?

I met Forex Rebellion and studied it last year but I hadn't had the time to automate it and see what it really gives. And to me that is important because hind sight will always have you believe most past trades are winners. I wanted to know more. Keep in mind that the goal of this article is not to explain all the details and rules of the system,  but just to find out whether it is profitable or not. I assume you know the rules, and I just go from there.

The challenge to judge whether Rebellion is profitable or not lies in the fact that the system has some discretionary elements. For example you can choose your target to be the 20 Donchian channel, or simply 1.5x the size of your stop loss, but you can just not have a target, and trail your stop loss indefinitely until it is eventually hit, giving you opportunity for greater winners. That is three totally different and personal ways to decide your target. There are also some discretionary decisions regarding stop losses. You can place it below the most recent swing low (for a long trade) or above the most recent swing high, but you can also place it below the 5 EMA (for a long trade) or above it for a short play. With this discretionary rules it is difficult to judge the system and say "hey it doesn't work!" or "hey it does". But what if you create an EA that allows you to test all possible discretionary combinations? If the 1.5x target doesn't work for a whole decade we can say it is not a solid proposition. If the indefinite trailing also results in a negative result for a decade of back-testing it means it is not good either. And same thing with the stop losses. We can test each one of these rules by separate and test them for a whole decade to see its effect in the system profitability. And what is more, we can also do a mathematical expectancy analysis, to find out if the entry rules have a positive edge over a random entry. Read this article I wrote on  how to measure a system's edge. So, let's get down to business.

I will be using EURUSD in all the tests. The data is AlpariUK data obtained by direct request to the broker (not the crap downloaded from MetaQuotes servers)

Measuring Rebellion entries' edge

I will perform the math expectancy analysis form January 1, 2000 to January 1, 2010. 10 years worth of data.

Long entry: 4 EMA crosses above 5 EMA. QQE Adv is above the 50 line and blue line is above the red one
Short entry: 4 EMA crosses bellow 5 EMA. QQE Adv is below the 50 line and blue line is below the red one

I'm going to measure the Maximum Favorable Excursion (MFE) and Maximum Adverse Excursion (MAE) for timeframes H1, H4, D1 and I will measure them 10 bars after the entry, 15 bars and 20 bars after the entries.

Again read this article to understand what Maximum Favorable and Adverse Excursion mean in a mathematical expectancy analysis. What you want to see is all the MFEs higher than the MAEs, and not just higher, but higher by a decent margin, because you also need to mitigate trading costs.

Time frame: H1
Long entries: 11595
Short entries: 10852
DirectionBars after entryMFEMAE

As you can see, the favorable excursions are worse than the adverse excursions. This means the entry rules DO NOT HAVE a positive edge and are no better than a random entry. If anything, they are worse than a random entry.

Time frame: H4
Long entries: 3350
Short entries: 3046
DirectionBars after entryMFEMAE

Time frame: D1
Long entries: 624
Short entries: 493
DirectionBars after entryMFEMAE

It's pretty obvious that the entries do not offer any significant edge. Although the edge seems to improve a little in the higher timeframes, but still, it is not across the board. Although there is some hope in the D1 timeframe, the short entries reveal that there's really no intrinsic edge, possibly cancelling out what is obtained through the long entries.

Here's a backtest in H1 between 2008 and 2010 using the 1.5x rule as my target (that is Targets are 1.5 times the size of stop losses) and for the Stop Loss I use the 5 EMA rule. That is placing the stop loss 3 pips below the 5 EMA or above it depending on whether it is a long or short play. Risk is 1% of the account per trade. Stop loss is cut in half once profit in pips is half the distance that exists from entry to stop loss. There is also a trailing stop applied following the 5 EMA shifted three periods.

(Click on images to enlarge)

Folks, there is no hope in H1. Just as forecasted by the mathematical expectancy analysis for this timeframe. You can try any combination of stop loss rule with target profit, it is all the same. It's all downhill from the beginning.

Now let's do a backtest of H4 using no target, that is following the price with a trailing stop until kicked out. Hoping to maximize the winners. January 1, 2007 to January 1, 2009.
(Click on images to enlarge)

Ooopppssss,.....this doesn't look like what is advertised. 3.31% growth in 2 years with a 13.50% draw down.

Finally another test, this one in D1 using the 1.5x rule as my target (that is Targets at 1.5 times the size of stop losses) and for the Stop Loss rule this back test is done using the 5 EMA, placing the stop loss 3 pips below the 5 EMA or above it depending on whether it is a long or short play. Risk is 1% of the account per trade. Note that the number of trades in the backtest is not the same as in the expectancy analysis and that is because sometimes you get a signal when you are already in a position, and the system doesn't play it, nor does it add to the existing position.

(Click on images to enlarge)
Things are slightly better in D1, as there seems to be a very small edge. The 86% quality of the backtest is only because the last trade is not stopped out nor exited with profit but interrupted by the backtest. That means the last trade was still  open at the end of the backtest period.

Profit factor 1.24, Portfolio growth 29% in 10 years with a maximum drawdown of 13.67%. At least it is a profitable result, although not what you have been promised. 29% growth in 10 years is too much of a torture. But hey at least this is a profitable result. You could increase the risk per trade to 2% and then achieve 58% growth in ten years, but at the expense of a 27.34% draw down.

My conclusion is that Forex Rebellion has good points, such as protecting profits, attempting to let winners run and cutting losses short. But it doesn't deliver in terms of results. The claims that this system works on any currency pair and any time frame are not valid.

If you want to experiment with the Forex Rebellion Expert Advisor or modify the code, here is the EA
Download Forex Rebellion Expert Advisor (EA)

Related Articles:
LT Trend Sniper - a Forex strategy that works
The Turtles Trading system automated (Expert Advisor for download)

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

No comments:

Post a Comment