EA Reviews

The end of the week is here, so we have a little time to analyze and reflect on how the robots did this week in testing. Here is a quick breakdown of where we stand with our demos.

PIP ANDROID

Pip Android took only four trades this week – two wins and two losses for a very small gain on the week. Our demo has been running since February 11, and the EA has lost $2.03 or 0.1% since that time. There are no positions open currently. So we’re basically at breakeven so far. I’ll give it a few more weeks at least before drawing any hard and fast conclusions.

FUSION V

As mentioned in my previous update, the account statement that the FRWC showed for their live test of Fusion V clearly shows a lot size of 0.1 lots. Thus, in an effort to test the robot on identical settings, I advocated making the change to a 0.1 lot size. Well, I immediately saw that this wasn’t going to work. Fusion proceeded to open up ten HiRIDER trades, and suddenly my available margin was in jeopardy. It got down to around 105% before the EUR/USD turned the opposite way and allowed me to close out of the trades with small gains before it damaged the account. Note that I am using an identical FXCM account as the FRWC folks ($1,000 balance, 1:200 leverage). Here’s the thing. If you get several wins in a row right out of the gate, you might be able to keep trading at such a high lot size, but you are risking a margin call, especially if the first few trades go against you at all. Therefore, I do not think it is even remotely reasonable to expect the kind of gains that FRWC showed on their 19-day Fusion V demo (345%). It is entirely too risky to trade such high lot sizes on such little starting capital, and frankly I think it’s misleading that they would advertise results that used such a risky setup but then in their manual say this: “Our opinion is that a lot size equal to 0.01 for each $1,000 of available equity is reasonable.” If that is their opinion and what they think their clients should use, then that is what they should have used in their testing. That effectively means the expected results should be about a tenth of what they showed, or about 34.5% over 19 days — still very respectable and worth using but a far cry from over 345%. These guys talk a lot about transparency. Where is the transparency in that?

Interestingly, in the manual they state (and it is highlighted): “In our opinion, it is far safer to turn off the automatic Money Management feature and set a fixed lot size of 0.01. This is because certain strategies, i.e. Straasha, involve a modified Martingale system while other strategies (like HiRIDER) can open up to 10 trades at a time! You must therefore set a “safe” lot size to avoid unnecessary risks.” This is a warning that shouldn’t go unheeded, in my opinion. Thus going forward, I have changed my lot size back to 0.01 lots with money management turned off. I will follow the guide of 0.01 lots per $1,000 of equity, as recommended. If you do likewise, just be aware of the fact that you will nowhere approach the ridiculous return that they reported during their 19-day test, but you should be able to rest a bit easier knowing your money is safer.

The good news is the robot did make it into positive territory by the end of the week. The account balance after closed trades is $1,016.84. One trade remains open and is currently profitable, so the account equity currently stands at $1,025.05, or a gain of about 2.5% over the first three days of the test. That’s a pretty good start.

LMD-MULTICURRENCY

Well, unfortunately the LMD-Multicurrency EA continued to dig itself a deeper and deeper hole and at one point was in drawdown over 50% – a complete train wreck, frankly. I’m not sure about you, but I’m not willing to even consider putting real money in a robot that has drawdowns of that magnitude. Therefore I am going to be starting a new demo with this EA with the following changes to the settings to see if we can get much smaller drawdowns with the expectation that the possible account increases will also be tempered:

Lots: 0.05
Microlots: true
Money Management: false

I will use the user guide’s recommendation of 0.05 lots per $1,000 of equity.

The current demo will be stopped and its results will be available on the Results page on Monday.

HI-RIDER

Thus far, this robot has been the most consistent performer of the bunch.  It has registered only one loss in the first three days of testing. Closed trades have resulted in an account gain of $62.77, or 6.3%. However, on Friday, two trades came just short of hitting their take profit level before completely reversing and going the opposite direction. These two trades together are down $40.10, so the account equity is at $1,022.67 at the close of the week, for a 2.3% gain for the first three days of trading. Again, this is a pretty impressive start. Let’s see if it can recover from the two stranded trades early next week.

In rereading the manual and after looking carefully at the backtests, I think I probably have the risk settings too high on this robot as well. There were at least three occasions in 2008 when all ten positions hit a full stoploss at the same time. That could potentially be account debilitating if your risk levels are too high and if it happens several times in a short period of time. Thus, going forward, I will be changing the LotRiskReductor setting from 5 to 1 with money management turned ON. Admittedly, this will result in slower growth than what was achieved during the two month competition, but it will be a lot safer. Account protection is the name of the game, folks.

SUPER VOLCANO

Not much going on with this robot. It took one trade, which it lost. Account equity is at $958.80, so the total loss for the first three days in the market was 4.1%. All we can do is keep the demo running and see what happens going forward.

Closing note

I’m going to be honest here. I still love the idea behind the FRWC competition, and I think there are some good robots here, but I feel they did everyone a disservice by allowing the robots to operate with risk levels that simply won’t be sustainable over the long term. Yes, you might get an occasional two-month period where some of these robots do really well, but that does not mean they are viable for a year or even six months at those settings. And the problem is that no one ever knows when a bad period is coming. Thus it’s not realistic to think you can just go forward with those settings. My guess is that you could run the competition again with the same robots and same settings, and you would probably get vastly different results, both in terms of percentage gains/losses as well as the finishing order.

As I said, I think some of these robots may indeed be winners, but I think we need to rein in our expectations a bit with them. I think the risk settings need to be lowered to what the FRWC suggests in its manuals so as to make the robots’ performance sustainable for the long-term. It does no good to have great gains for several months only to lose all of it along with a portion of your principal in one or two bad trades.