Launching in ten days...

It appears that there is a bug with MBTrading demo accounts, which was related to my problem. They SDK support guy told me how to bypass it and it works. The robot is working. I added a Skype plug-in to my robot, in order to send me SMS every time an order is closed, or being opened.

I decided to use Alpari at the moment as my signal supplier, and an optimization is being run in order to find the best set.

Next week will be used to final tests and checks, and afterwards the robot will be launched for the first time.

Metatrader platform difficulties

In the last 3 days I was busy debugging the MBTrading robot application. I had a problem related to closing the trades, and I am still in touch with their support team. Except that, the robot is ready.

I noticed a problem related to back test with Metatrader platforms - The Metatrader platform from Metaquotes is very fast, and very comfort to back test, unlike the Alpari platform. I compared them and Alpari is much slower. The second problem is that Alpari has different performance with the same data.

I need a signal supplier which is reliable, and it's platform will be able to conduct my optimizations rapidly. I am examining different brokers in order to find a good signal supplier for my robot.

Version 6

This week started with two successful trades, and no more. While observing the expert advisor's performance, an idea for a filter came to me, and I decided to check it out. It seems that the new version (from now on it will be version 6) makes the same net profit, with half the trades, and with half the drawdown as well. The back test's results are available here.

The robot that trades with MBTrading is robust to logic changes, and will continue working with any version that I will choose. So, it seems that we will continue with version 6.

Week's summary

The robot software (which trades via MBTrading) worked well today, made 2 trades according to the expert advisor's orders. The software is stable and simple.

I had a problem this week that the forward testing of version 5 did not match exactly the back test's. I suspected that since I ran two different versions in parallel it might be the reason. So since yesterday only version 5 is being tested on that platform, and the results match at the moment - there were 3 trades since then.

What's next? I have to make a small addition to the expert advisor that will store global variables in case of Metatrdare platform shutdown. I will continue the robot's testing, and if it will be fine, maybe I'll start live trading two weeks from now.

Progress report

The robot started working, and it's stable. I had a few trades executed, but it needed a manual confirmation. I fixed it. I will have to check carefully the orders execution, and compare it to my expert advisor, since MBTrading works differently. In order to make a buy order, you have to place a position, and against it an open order with the limits. It's more complicated, and therefore it requires extra care.

I noticed that with the Alpari platofm, the forward testing did not match exactly the back test's results of this week. I suspect that since I ran versions 4 and 5 together on the same platform might be the cause. I left only version 5 for the forward testing.

Robot in demo mode

Well, the settings of the new version is set for forward test in this week. I still need to check the risks again, since it was changed due to logic changes. The back test of the new version is available here. The results are much better: the P.F changed to 2.31, 77% profit trades, with a little bit smaller drawdown.

I finished fixing the robot application. This application embeds the MBTrading API, and receives orders from the Metatrader expert advisor. I used a DLL I wrote in the past for communication between the two different applications.

Right now two versions are in forward testing: MYP4 and the last one - MYP5.

The robot is triggered from MYP5.

New version - MYP5

After watching my expert advisor in live, I tried a few ideas to minimize the loss trades. I checked a mechanism of rescue, when the trade is profitable, to set the stop loss to a profitable level. This idea did not achieve better results. Afterwards I tried a few conventional filters, which did not do any better. 

At last,  I had an idea for a filter, with three levels:

1. High risk trades

2. Low risk trades

3. No trading at all

I tried it, and it had better results.

The results are for the period 19/9/07-18/9/08 GBPJPY H1:

Profit	P.F	Drawdown (%)	Trades
19697	1.96	26.95	128
23061	2.21	26.67	109
23660	2.83	26.17	87

The risk is as it was the same (4% for the first and 6% for the second) with a small modification: the filter determines the risk of the first trade: 4% is considered as the high risk and 2% for the low risk.

What's next? I am going to continue with the third result and make a small risk optimization around that setting since the drawdown settings is changed.