New Version 2.0.0.0 Test Results And Graphs

I finally finished the new version 2.0 and tests results for last 3 years are done.
all values are from 2010.1.1 to 2012.11.1:
Profit: 107.6%
Drawdown: 9.3%
Max leverage: 6.57
Avg leverage: 0.732953595174
Sharp ratio: 1.89336547786

Max Drawdown per symbol is very low 6% only

Test Results per Symbol:

 All Symbols combined graph:

 

Drawdown combined with each symbol with a different color.

 This version has been running already for 3 weeks on a demo account, I will soon move it to a real account.
anyone interested in detailed reports just write a comment and I will reply.

New Version on the way 2.0.0.0

I am working on the a new version which should be much more attracting to clients with 1% per week and drawdown of 10% total.
I haven't recover yet from FXCM damages and My System is currently at 10% drawdown but I put my system on closing state so no new positions are opened and when all positions will close I will move to some other Broker.
I already started running demo test on my new version and I hope next week I will have graphs to show.

FXCM - Server problems

Bad news, turns out that my broker server was disconnecting me for long time periods for two months now, I discovered it when they asked me if I want to switch a server, I went to check my logs and I found out that my system failed cause the server disconnected me.
After that I filed a complaint showing my application simulation - what was supposed to be the positions but they refused claiming it is hypothetical.(I files a second query, its under investigation)
I learned my lesson - I changed my system from running only on new ticks(which didn't arrive from the server) to an infinite loop that check to see it the server is online and I am now getting informed about it.
In my next broker I hope I will be able to catch and solve the problems in time.
Seems to me you can't build an automated system because no matter how fail safe is your system other factors can ruin it.(you have to add human to solve real life problems)
If there is a broker you had good experience with, let me know please, currently I am checking Oanda and Alpari UK.
My Biggest problem now is my track record which will have a big drawdown, I hope next month I will have better news, I am still improving my algo and adding new features.
Next algo is in final stages and I will run it few hours each day, very aggressive algo should make 1% per day.

Bad Month, 7.74% total.

This month seems like we won't reach the 2% target or the 4% optimistic target.
The app reached max drawdown of 14% and we did only 1.7% this month reaching total of 7.74%.
I am now considering to create preset specially for periods like this month when the volumes are very low and we don't get signals on the current presets.

I found another Symbol I can trade on AUDCAD I will add it soon when upgrading to the new version that has more features for reducing the risk by filtering signals on very high volatility markets.
Test results of last version:

3 months passed, 6% profit and a bug

Another month has passed and the automated trading system did another 2%.
The system is now connected to MyFxBook which can show system graphs and statistics:




Below you can see the profit graph of my 3 systems:
HadarFx 10 Symbols - this is the real account system.
Hadar High Risk - This is a high risk algo which should make 100% for a year.
Hadar Demo - this is the next version running for testing on a demo account, should make 1% per week.





Two days ago I discovered A bug, EURUSD opened a very big position instead of opening a small one, It caused the system a 10% drawdown and leverage reached x6 which is high, I closed the position and fixed the bug, the system opened a 0.06 lots position instead of 2.34 lots that were opened by the bug.

Version 1.1.2.72 - one percent every week

New version looks promising and the improvements in that version should make One percent every week with max drawdown of 25% and max leverage of 6.
We connected the system to MyFxBook so you can see the Live trades there - Hadar 10 Symbols.
The main feature I added was closing part of the loses when a big jump happens, it reduces the drawdown and shorten the average position open time to 4.5 days.


Test Results new version (1.1.2.71):

(in 2007 we reached 40% drawdown for one day caused by 14% drawdown of AUDJPY which moved 50% in 3 months that year)


We are trading on 10 pairs, each pair makes an average of 5.4% for a year with average drawdown of 6.74% and max leverage of 4.5.

4 pairs already started running the new version, when the other pairs will close their positions I will move them to the new version also, till then the profit will probably be around 2%.

Automated Testing - Another key for success

One of the Keys for having a good EA is an automated testing system, My system can deploy tests automatically on many computers at once and that way reduce the overall time.(it is deployed on Amazon EC2 and Google App Engine, it can also be used on any computer network) and it is all done through a simple web interface.

It can run on any EA (expert advisor), and it saved all the data in a database for later use.

Each test saves the Preset In the database no need to keep it at your file system.

The Table below shows the test results filtered for one pair EURUSD for 5 years, it can also filter per date.

by clicking on a row you will get into another page that shows all the details for that tests including all the positions that were made in the tests and a detailed summary that MT4 produced.

The graphs below shows the totals for each year tests, each year it ran 10 Pairs.

The Graph below shows the equity and leverage for every hour combining all tests for that year together:

If you have an EA that takes a long time to test contact me maybe we can save you a lot of time.

I don't see anyone succeeding on making a good EA without a very good testing system that can show you what are the limits of your EA.

Running an EA without thoroughly testing it is like flying a plane without enough fuel  - you will crash and burn.

Two months have passed - another %2 profit

My fully automated system has been running for two months now on 10 pairs (the previous version run on 6 pairs only) and seems like it has a rate of 2% per month which was the minimum rate I predicted.
you can see below the graph of the last two months, The balance growing is linear ,the draw-down max was  5.63% and the leverage max 3.2.
Total profit is  4.3%.
This graph is generated by my system Hadar(build 1.1.2.7) and it shows the balance/equity/lots for every hour:

The graph below is generated by MetaTrader 4(build 432):

As you can see the MetaTrader 4 Graph Drawdown is much smaller 0.78% instead of 5.63% and that is because it Relies only on positions closing time and not every hour.
If you have a Metatrader 4 account and you want to see the Real Drawdown just email me the account details and I will give you the one hour graph of your account.(every account has a read only password that you can change)

How to compare Forex Robots or Automated Trading Services

Many Forex companies offer today an automated trading robot, the big question is how can you decide which one is the best ? how much past performance numbers are worth ?
I will try to give only a partial answer but it should cover 90% of the robots and traders that declare they have a winning algorithm or robot.
I will be referring here only to Meta-Trader 4 and Forex markets and for all use cases I will assume the account that is used has 100,000$ ,leverage of 1:100 on EURUSD price 1.25592.

• Leverage - Most of the robots I checked failed on that, they suddenly open those big positions which in the good case cover the small positions or in a bad case lose more money, leverage of positions in one currency pairs should not be more then 5 and even that exposes the robot to a big lose.

here is an example:
If a robot trading EURUSD, current price is 1.25592 and it opened a "buy" position with 10 lots, A market move of 5% to 1.193124 will cost 62.79% of the account which in our case is 62790$.
Lets assume now that the Robot is using Stop-loss of 1% then in that case it will only need to be wrong 5 times in a row and you get again the same lose.
Good robot should have a very low average of lots open.

• Drawdown -  Usually with big leverage you will also see big drawdown, obviously smaller drawdown is better, but the problem is to know what is the real drawdown which that robot had in the past, since the Meta-Trader 4 graph shows you only closed positions you don't know if those positions had a bigger draw-down, in my previous blog, you can see there an example of how mt4 graph hide the real draw-down.
• Can Work At Any Time - If the robot is declared to work only in special market conditions then it is worthless since verifying a market condition is a hard thing to accomplish and you can't tell if the special market condition is over already or just starting, black swan can happen at any time, so your algo must be ready for that in all time, even if it works only in predefined hours or predefined conditions.
• Tested - A Robot should run or be tested at least  5 years back and you should compare those results, you should check how it is performing on bad years like 2008 or 2003 or how it performed in special dates like 9.11.
• Profit - Usually the rate that a robot can make money is the rate it can lose money so if a robot can make 10% a day that means that it can lose 10% a day or more, so don't be tempted to choose a Robot that makes 100% per month cause that will probably be the time it will take you to lose all your money.

This basic points are also true for any trader not just automatic if you are using large positions or have a big drawdown or you think you know "how the market is behaving" don't trade cause your luck will end eventually.
Remember that most of the traders in Forex lose their money and when someone finds a way that works he will not publish it or sell it otherwise we would have seen more people wining.

The real picture behind Forex MT4 graphs

Lately I am founding myself explaining to people that MetaTrader 4 History graphs are not accurate and don't show the real picture, well I decided to address it here in my blog.
The problem is that the MT4 graphs shows the account balance/equity only when positions are closed and you actualy miss the time between closing positions, so you are only seeing the "good" picture in case of algo that closes only when it is in profit.
You can see the difference below:
This is the mt4 graph, as you can see  almost no drawdown because most of the positions closed with profit, but you don't see what went on between closing.

This graph which is produced by my system shows the balance+Equity+lots on every hour not only when closing, Both graphs are using the same data.

If anyone is interested in the script that is making the second graph, let me know, if there will be enough demand I will try to make it available a part from my system.

Almost a month

Almost a month past and I wanted to share the new look of my system report+graph, it needs some fixing but I couldn't wait.
below you can see the graph for the last month, the system was down for a week during that graph.

The system is currently working with only 8 pairs since the EURCHF is not moving and the AUDUSD failed the last tests.
I am working now on 4 new pairs, I hope I will be able to add them soon.

Maybe I should raise the risk

A week passed, and it seems the markets are moving very slow and the robot is not doing enough money, so we decided to raise the risk a bit, below is the test results with higher risk, not sure yet if to use them.
current profit is 1.5%.

Fully automated trading machine

My last version(1.1.2.48) is finally closed and it will be up and running today at 24:00, it should make approx 30% a year with a drawdown of max 30%, in most of the time the drawdown is between 2% to 5% and the leverage max is 10 .

You can read about my first year experience by going to this link where my old blog was( did 13% in 10 months).

new version already ran for two weeks on 6 pairs, and did 1% with drawdown of 0.5%, today it will start trading on 10 pairs.

you can see the graph of test simulations with the new version below.

the graph below shows the position state for the last year (simulation).

Use  this link to go to my previous blog.