Blog Item

Black Belt EA - Expert Advisor in Forward Test

Black Belt EAI'm happy to admit that I don't know the first thing about either Judo or Karate, but I get the impression that the Black Belt EA developer might be a martial arts expert.

I've therefore removed my witches hat to write this short article, and I'm going to try to say some nice things about the EA, through fear that I might get a visit from Bruce Lee or Jackie Chan to 'sort me out'. :roll:

Seriously though, this brand new Expert Advisor seems to have already captured a lot of interest, and the developer has kindly provided a copy to put into forward test at MellyForex so that our readers can monitor its performance.

The Black Belt website states that the EA is a non-grid, non-Martingale high frequency trader that works on the EURUSD symbol on the 30-minute chart timeframe. It claims to use price action rather than indicators for its trade entry signals, it's supposedly ECN and NFA compliant and, although it is meant to work best with brokers whose EURUSD spread is under 2 pips, spread is not meant to be that big an issue.

Now that the standard marketing spiel has been dealt with, what else can I tell you about the EA?

Well, the first thing I noticed from the backtests on the Black Belt website was that those backtests were performed using the 'Open prices only' model. The suggestion here is that the EA would only ever open and close its trades on the first tick of a new price bar, and this is something that the developer confirmed to me. In addition, the EA doesn't declare its stop-losses or take-profits which means that, if the EA gets a signal to close a trade shortly after a price bar has opened, the user will have to wait patiently at the mercy of the markets while the current price bar completes its formation and a new bar opens before the exits the trade. Obviously the market could go either way during that time, and there's an obvious risk that an extra 100 pips or more could be lost while the EA is waiting to close its trade.

My next observation is that the EA uses a form of 'pyramiding' as part of its strategy. Pyramiding is the process of using the free margin generated by an open, profitable trade to open additional new trades. In its purest form, the lot sizes of the continuation pyramid trades are calculated to use the free margin which has been generated by the other trade(s). This approach frequently means that the continuation trades will be of a smaller size than the initial trade, although Black Belt's additional positions are most likely to all be of the same size as the initial trade. As I scrolled down to view one of the backtest statements, I didn't get very far before I noticed that the EA had opened no fewer than 9 concurrent trades. When I enquired, the developer advised me that the default was for the EA to open as many as 16 open concurrent trades. This might seem excessive, but remember that the additional positions are only opened on the basis that a continuation entry signal has been received and the previous trades are in profit. It most certainly shouldn't be confused with a typical grid system which adds to losers.

My biggest concern here is that, if you've got umpteen open pyramid trades, and the market turns against you, and you've then got to wait for a new bar to start before the EA closes out, you could potentially find yourself losing umpteen hundred pips while you're hiding behind the sofa overdosing on Valium. Pyramid trading can be both high risk and high gain, so widows and orphans should take note!

Although I've put the Black Belt EA straight into forward test without running any strategy tests of my own, I did spend a bit of time analysing the developer's own backtests. In particular, I examined a statement from 2000 through to the end of 2011 and can confirm that the win rate is approximately 53% with an average winning trade size of 51.8 pips and an average loss trade size of 38.4 pips.

The total number of pips won in those 12 years was 75,286 which, in 4,379 days of test, equates to 523 pips won each month, achieved from an average of around 12 trades each week. Given that this pyramid style of trading is similar to that employed by Forex Growth Bot, which I also have in test at MellyForex, obvious comparisons are going to be drawn between the two EAs.

Also of importance from the developer's backtests, the trade expectancy is 10.0 pips per trade which should seemingly support the developer's assertions that the EA isn't broker dependent.

Sounds great so far? Unfortunately, when things look like they're too good to be true, they very often are.

In the case of of Black Belt EA, there was something unusual that I could see in the 12 year fixed lot equity curve which was sticking out like a sore thumb. When an EA trades variable lot sizes, which increase as the account balance also increases, the equity curve tends to have an exponential appearance and often rises steeply in the latter part of the curve. With a fixed lot equity curve, however, the line should be nearly straight and rise at an even rate.

A quick look at the Black Belt fixed lot curve clearly shows a steep jump in the curve in the middle section as I've highlighted below.

Black Box EA Equity Curve

The rate of incline is reasonably constant from 2000 through 2007, but the balance curve rises steeply in 2008 and 2009, before resuming at a slope angle very similar to its previous incline from 2010 through to the end of the backtest.

It's common knowledge that the global financial crisis affected markets dramatically during both 2008 and 2009. Fiber quickly fell from its perch at $1.60 to a low of $1.41 in less than 4 months during 2008, before subsequently rallying back up to $1.51 before the end of 2009. Fiber's longer term average monthly range is around 500 pips, and it very quickly jumped to three times that amount back in 2008. Those were clearly exceptional circumstances, and it's therefore not surprising that a pyramid trend-following EA like Black Belt would have been able to latch onto some of the trends and profit much more than it would have been able to whilst operating during normal, more sedate market conditions.

When I looked at both 2008 and 2009 as individual years, I found that Black Box would have made over 25,000 pips in 2008 and nearly 13,000 pips in 2009. Removing these two years from the 12-year backtest results leaves us with 37,218 pips generated from 6,046 trades over 3,648 days. In other words, over half of all the pips gained in the 12-year backtest would have been won in just two of those twelve years. The win rate was unaffected by removing those two years and stays close to 53%, but the average winning trade size reduces to 44.7 pips whilst the average loss size only changes marginally to 36.9 pips.

We're now trading in 2012, markets have returned to a much more orderly fashion with average monthly ranges having fallen back to 600 or 700 pips, and the crazy events of 2008 and 2009 have become little more than a distant memory. I don't really feel, therefore, that it's fair to include Black Box's extraordinary performance of those two years within a current analysis of the EA.

Despite this, the results for the other 10 years of the backtests are still quite impressive, although we're now talking about an EA which looks to be gaining 310 pips a month instead of the 523 pips suggested previously, and whose trade expectancy is only 6.2 pips per trade instead of 10.0 pips. The reduced expectancy means that the EA is getting into the realms of possibly being broker dependent, as an extra pip on the spread, or a pip or two of slippage here and there, is likely to impair results noticeably. Unfortunately, I don't know what spread was used in the developer's backtests.

Remember also that the developer's backtests appear to be using MetaQuotes history data which has some notorious gaps of a month or more where history data is missing completely. I've not looked closely enough to say whether or not the MetaQuotes gaps are impacting 2010 and 2011 backtest performance for better or worse.

The worst case drawdown when trading 0.1 lots in the backtests was $2,294. Ordinarily, a risk simulation on the test results would give an idea of the minimum balance needed to run an EA but, in the case of Black Box, I'm uncertain as to precisely how much correlation there is between the pyramid trades. Risk simulations involve replaying a set of trades in a random order to get an idea of the worst thing that could happen, but I'm not so sure that it's appropriate to replay Black Box's trades in a random order because the continuation trades are only ever opened on the basis that the initial open trade is in profit and the baskets of open trades should remain in order.

Nonetheless, I did conduct a risk simulation to get an idea of the minimum balance required to run Black Box, and the simulation suggested a minimum deposit of around $4,300 for an account trading 0.1 lots. This roughly concurs with the developer's assertion on his website which suggests that $500 would suffice for a microlot account. My only reservation here is that using such a small account small could potentially spook users into turning the EA off once they witness it first hand opening multiple pyramid trades.

Bruce LeeReaders should also note that the largest single winning trade in the backtests was over 600 pips and the largest losing trade was over 500 pips. Admittedly, these worst case scenarios both happened during 2008/2009, but the biggest loss during the other 10 years was still as high as 473 pips which might leave some users running scared.

Despite the various caveats above, the performance is still likely to be better than that of most EAs, and I certainly hold the belief that this EA could prove a winner in the weeks and months ahead. In all seriousness, if Black Box can achieve an average return of over 300 pips each month, it will quickly find itself sitting very close to the top of my leaderboard.

Besides which, I really don't want to receive a visit from Bruce Lee!

For now, I've put the Black Belt EA into forward test on a $5,000 FXCC demo account. FXCC have recently changed their account structure and are now offering micro lot sizes.

The EA does not contain much by way of user settings to adjust, which is the way I think things should be. There is an option to either trade fixed or variable lots, and I've chosen the variable lot option for my forward test and have set the 'balance_percent' parameter to 2.0.

Readers should note that this EA handles risk in the opposite way to most commercial EAs, and users need to INCREASE this parameter in order to REDUCE the risk. I'm expecting to see the EA take trades of 0.05 lot size until the account balance has increased sufficiently for the EA to risk larger lots.

As with all my EAs in forward test, I'm publishing the Black Box EA's trades to MellyForex, and its performance can be viewed in detail by clicking here.

Discussion (0 comments)

Be the first to leave a comment!

Add a Comment

;-) :-) :-D :-( :-o :-O B-) :oops: :-[] :-P