The Forex MegaDroid Backtesting Blues – Track 1
It's now three month's since "Forex MegaDroid landed on the Forex industry like a BOMB". The time has come to try and tie together all the backtesting, forward testing and live testing, and to decide whether to keep the Droid running on our virtual private server for a while longer. Unfortunately this is not an easy task!
We'll start by recounting the various trials and tribulations we have encountered whilst trying to reproduce the so called "Live Updates" shown on the Forex MegaDroid website.
On April 8th, shortly after we bought the Droid, we received an email from Forex MegaDroid support. This told us that:
After literally hundreds of support tickets on this subject…. Users have experienced a lot of confusion and frustration while attempting to produce reasonable back-tests…We are therefore requesting that all users follow the instructions in the manual to download and install MetaTrader from the ForexMeta website for back-testing purposes.
I'm sure that you can imagine our surprise when we subsequently got to read version 1.10 of the Forex MegaDroid manual, dated April 20th. This said that:
Matching results shown on the website – We used the FXDD version of MetaTrader for back-testing with the following settings: All default, except RiskLevel (generally 0.3) and GmtOffset (+3)
This news did however come as welcome relief to us, because having followed the instructions in version 1.00 of the manual we couldn't download any historical data from where we finally ended up, which was Forex.com. Leaving historical issues aside, we're sticking with FXDD for our current tests. Last weekend we went to the FXDD website and downloaded their demo version of MetaTrader 4.00, build 224, dated May 15th 2009. Then we tried to download historical one minute bars for the EUR/USD pair. As we did this we received a warning message that said:
You are going to download data from the "MetaQuotes Software Corp." history centre, but your account is opened on the "FXDirectDealer" trade server.
Please do not forget that the downloaded history data can differ from the data stored on the trade server. This is why these data cannot be used in disputes over historical prices with "FXDirectDealer". If you still want to continue, press OK.
We wanted to continue, so we clicked OK. We haven't forgotten.
Then we backtested ForexMegadroid over the first half of 2009 using the data we had just downloaded and the input settings recommended in the Forex Megadroid manual. This is the equity curve that the FXDD MetaTrader 4.00 build 224 strategy tester produced for us:
As you can see for yourselves, for some strange reason this is not the same graph that is shown on the Forex MegaDroid website.
When we inspected the trade by trade results carefully we noticed that although the first trade in our backtest occurred at the same time as the first trade shown on the Forex MegaDroid website the size was different. We got out our calculator and then tried another backtest using a Risk Level setting of 0.3403. This is the equity curve that the FXDD MetaTrader 4.00 build 224 strategy tester produced for us:
Eureka! We had finally managed to get a backtest result in which the first trade was identical to the first trade shown in the "Live Updates" produced by Albert and John and their team. Nonetheless this small success raised far more questions than it answered. In particular, why does the Forex MegaDroid sales page proclaim in a big yellow and black headline "No Losing Trade Since 20th March 2009!" when if you perform a backtest following the exact instructions in their manual two successive losing trades in May are immediately obvious?