Are these too good test results to be credible?

Trading Systems Expert Advisors and Automated Trading
1

Hi. I have this ea which I have been building and optimizing about half a year now. For every test, no matter if its based on OHLC readings or tick-by-tick data, or time period, it has always given profit factor of at least 2, usually over 5. Estimated monthly growth would be 10-40 %, depending on the risk level.

The ea uses multiple timeframes and indicators to find most comfortable trading opportunities, but here’s the basic idea:
It buys when 5-min chart has been in long and powerful uptrend, and then it has a pullback down. When RSI is below 30 and then crosses above it, the buy order happens. Sell orders have the same principle reversed, the sell order happens when RSI crosses below 70.
Trades happen approximately 4-5 times a week.

The most interesting thing is that my ea has enormous stop-loss levels compared to tp: when a stop-loss happens, 52 take-profits are needed to cover it.
However, backtest result shows that the big sl does not ruin the ea. Failed trades happens 1-2 times a year, when only 2-3 months are needed to come back profitable.

I will attach 3 pictures: One of a test with 30% risk, one with a 0.05% and a chart that shows some example trades.

What do you think, is this ea really profitable or just misleading? Should I open a PAMM account and start making some change?

EDIT: Oh, the attach image quality is very bad :frowning:
I’ll attach them again in a .zip file

report 0.05.jpg382×700 35.1 KB

report 30%.jpg384×700 36.1 KB

chart.jpg700×249 40.5 KB

backtests.zip (706 KB)

2

Hi 5astelija,

Care to share your EA and have others test it?
I have a demo account and would like a chance to run this EA to see how it does.

Thanks.

3

Back testing can be unreliable. The best way to see if the EA is reliable is to run it on a demo account first. If it is profitable on demo after a number of weeks and comparable to back tests then try running it on a live account will a small deposit.

Hope it works out for you.

4

Sorry, but I won’t share the ea… After working hundreds of hours on it, I can’t just give it away :smiley:
I know that the best way is to test with demo.
The thing is that the stop-loss is so big that market needs to move one direction for weeks, and so I would have to wait months to get reasonable live testing results…

5

Ok, no problems. I agree with anc200, test with a demo with trailing stops, once you hit your profit targets, you can always manually stop it instead of letting it rides for weeks. Just a thought.

Good luck and keep us informed.

6

I have backtested it with trailings stops, break-evens, smaller stoplosses, etc… They aren’t good for the ea, the best profit is made with static stop-losses and take-profits.

I have used mt5 strategy optimization tool to make this “perfect”. The tool has tried all stop-losses from 1000 points to 20000, all trailing stop values from 20-600 points, while trailing starts when trade is profitable 0-1000 points (by point I mean the 5th decimal)… I have also found the right values for every indicator I use by the same method. Every single value is optimized for the best result.
That is what a computer can calculate in a few months, it would be impossible to make something like this by hand.

7

Started 24/7 demo account testing yesterday, so far no trades as there are no safe trends because of high volatility.

8

All right, first week of demo testing done. Four trades were opened and closed, all made profit.
[B]Total profit of the week 52.5 pips or 5.61 %[/B]

So far demo testing has given the same results as backtests, so this EA is starting to look very promising…
Of course I will continue the demo testing and inform you about the results here.

You can also check my progress on my Myfxbook account:
FxProfessor System by 5astelija | Myfxbook

Unfortunately the account’s track record is not verified. If somebody knows how to verify a mt5 account, your help would be appreciated :slight_smile:

9

I would say, BEWARE from EAs with straight line results. They usually have big stop loss… and practice and Murphy’s laws teach us that the biggest problems usually comes in crowds and that it can never be so bad that there are no more chances… to go worse. :wink:
I bought once similar EA, it had 150 pips SL, I tested it on few micro accounts with different risks… and very soon those SL were hitted very often. Imagine the rest. :open_mouth:

I think it’s always better to have EA which hits SL often, but at least you see how it deals with the losses. And if overall trading goes to bad, you have usually time to stop this, it’s not one hit and dead situation. Also, it’s hard to set nice risk, where you are not sure how often this SL can happen. It’s easier to control risk where you see many trades and many losses too, so you get the whole picture about drawdown and percents of losing trades easier.

Overall, I support your great work with your EA. Just my thought about safety. :wink:

10

Thanks for your thoughts, Zokimaster. When I half-accidentally tried the big stop-loss on my ea, I was very sceptical too, naturally. I already said in the opening post that this ea indeed uses big sl, BUT it doesn’t seem to be bad for the ea.

During those months optimizing my strategy, I have thought the SL a lot… And I have came up with 2 thoughts that supports the idea of large stop-loss:

Firstly: In theory, you always have a fifty-fifty chance of getting profit - the chart can go up, or down. So, in the long run, the big SL doesn’t mean that your equity is in constant danger to explode. A trade can sometimes go so badly that it triggers the SL, but sometimes it almost touches the SL but never triggers it, instead it changes direction and reaches profit eventually. You only have to trade small enough lots that a loss trade doesn’t blow your account. Conclusion: In the long run, you always have 50-50 chance, minus the spread. If your system has even a bit of “truth”, the win odds will eventually tilt to more than 50%, which makes the system profitable, no matter how big the sl is!

Secondly: In today’s globalized world the economies are in very close relation, and in currency pairs really long trends to only one direction are very rare. So, if you just “hop in” to a chart and the price goes to unfavorable direction, there is still a big chance of price coming back to favorable direction, because the chart uses to fluctuate.

I use even bigger stops than 150 pips. Actually, 150 pips is a bit too small for my ea. Under is a mt5 optimization run that shows how SL affects the profits. X-axis shows the SL, Y shows the profit. SL is in points, you got the pip value by reducing one “0”. Upper image is my EA, lower is a simple MACD-based EA, just for contrast. Both optimizations were run with the same TP, lot size (0.01), and the same range of different stop-losses. Time setting are also the same: M5 bars from 2009.01.01 to 2014.10.18.

SL differences.jpg700×501 46.3 KB

I am open for questions, conversation and critic… Please show me where I am wrong :wink:

11

I have been testing the EA with Alpari server history, which doesn’t have 5-min bars before 2009…

Today I opened a demo account at Fxtime and ran some test from year 2000 to 2014. I got very interesting results:

test.jpg700×227 30.1 KB


It shows clearly that the EA wouldn’t have worked at all before the 2008 economy collapse, but after it there is a clean profit all the way to present day. I am quite shocked about the significance that the long-term fundamentals have :34:

Now I will modify the EA to work also with times like pre-2008

EDIT: Added a combination of daily RSI and Stochastics to avoid too stable trends on long timeframe… Also gave trailing stop another chance to lower the risk. Here’s the result, seems quite profitable but there is still things left to improve :20:

Note that I use very low lot sizes during optimization to prevent margin calls which would mislead the results. Therefore only the form of the graph itself matters, the profit on the Y-axle is just a matter of lot size.

testt.jpg700×226 30 KB

12

This weeks gain 65.20 pips or +4.89%, one trade open losing 4.30 pips.
FxProfessor System by 5astelija | Myfxbook

I’m also worked hard with the ea. Removed the sl and tp from the logic, trades would now open and close freely with oscillators. MyFxbook has still the older version.

Here’s tick-by-tick, 98% quality backtest graph from 2000-2014. It needs some work because the current year hasn’t made much profit compared to before…

asd.jpg700×262 28.7 KB

13

If I can create a perfect and profitable EA, will I share it with others? Certainly not, imagine if I always win with my EA and I sell it worldwide, and the majority of traders use it and win. And, if there are only winners and no losers, the financial system will get stuck and break down. There’s no such thing as a profitable EA spread around; yet there are some profitable ones but kept in private.