EURUSD M30 1000% Strategie

Hi I found a good Strategie.
I generated this Strategie with data EURUSD_fhdb with Timeframe H1.
But the strategie works on M15 M30 H4 too.

I have checked the strategy in metatrader with data from activetrades, It works excelent on M30.

If you want you can test this strategy on a demoaccount and give me a response. I will do the same.

Please give me a feedback. Is this strategy good? or is were somethink what is not ok?
thomas

Hi,

this really seems like a nice strategy. You tested it on quite short data only from 2008, but when I retested it on my history data since 2000 it was profitable also before that. Nice example of out of sample test 🙂

The only problem is quite long period of stagnation in the middle, it took almost 3 years when the strategy profit was just oscillating around zero. Drawdown is just 13%, which is quite acceptable.
Also, all the profit was produced only by long side, short trades ended in loss.
Perhaps the strategy could be further improved by optimizing parameters in MT4 or strategy parts in GB.

Hi Marc,
yes I started the generation only from 2008. I read in the internet some strategies work only a short period of 2 or 3 years. So i want to look for
strategies who worked only the last 3 years.
I think the GB produrce will produce more profitable strategies if I choose a shorter timeperiod. For Timeframe 1H are 3 years enough I think?

Optimaziation is a problem. I don´t want to overoptimise the strategie. I optimized this strategie for the data of 80% of its time and check if the optimized
result is good enough for the last 20% of the data.

At the moment I test the version 058245 and 058245_optimized on a demoaccount at active trades, so I can compare this both versions.

I started to test the “walk forward analyser” from “http://www.easyexpertforex.com/walk-forward-metatrader.html”

But this tool is very slow. If you use timeframe M30 and every tick I can wait a long time (1->x Days) for the result.
I don´t understand exactly how to handle with the work-forward- analyser? The documentation of this tool give me
not enough information.

thomas

tnickel,
Have you figured out the Walk Forward Analyzer? I just got it working yesterday. I was using it to find optmized settings, but when I would use that set file and run normal backtest it became aparent that was not a good plan. So other than the Walk Forward Analyzer Number given at end, have you learned what else it is good for?

Hi stearno,
the walk forward analyser is not able to optimize an EA.
He is only able to say if it´s a good or bad EA.

A walkforward result >0.5 means that is a good EA.

You can use the last setting at the end of the testing period for your EA.

For example.
The walk-forward was from 1.1.2011-1.12.2012.

You can use the setting from 1.09-1.12.2012 (this is the last generated setting) for your ea. [It is only an example]

thomas

Okay, that helps. I will take what you said next time I test an EA. Have you found good settings to use in the WFA? like how many days to optimize and how many passes for each timeframe? For examlpe:

M1 – 30 days, 6 passes
M5 – 45 days, 9 passes
H1 – 6 months, 24 passes

-Stearno

Hi Sterno,
I don´t look after the days.

I look for timeperiods with more than 100 trades. If not enough trades I increase the days.
6 months for H1 is a good value for a H1 Strategy.

.
24 passes für H1 is a little high.
The calculation for this takes very long.
I use max of 15 passes.

thomas

THanks Thomas. I will tyr that next time.

Hi Sterno,
I don´t look after the days.

I look for timeperiods with more than 100 trades. If not enough trades I increase the days.
6 months for H1 is a good value for a H1 Strategy.

.
24 passes für H1 is a little high.
The calculation for this takes very long.
I use max of 15 passes.

thomas

Hi I tested the mql files between 2001 and 2013 and it looks realy nice. Unfortunatelly in your zip file it seem the strategy file (.str) is not the one related with the mql file !!! Also the strategy ID created by GB is different in the file name. I will appreciate a lot if you could publish the right .str file. Thanks

hi,

fcardix,

yes this was not the correct strategy in the zipfile.

I add the correct strategy in this posting.

I have done the backtest again, but in this test the strategy is not profitable.

I think this strategy was generated with an older version of GB. It is possible that in this older version was a bug.

thomas

Hi,

it is very silent here.

Are all people earning money with the GeneticBuilder?

Or have the most people given up to find good strategies?

My best strategies I found are in the M15 Timeframe (I found similarly good strategies in M30 and H1 timeframe)

I don´t found stable strategies in M1 or M5 timeframe.

The biggest problem was the curve fitting. It is very easy to find strategies with good equitycurves. But it is very difficult to find

strategies who work on real accounts.

I will show the portfolio of my best ten M15 strategies.

I hope this will motivate more people to find strategies in M15 Timeframe.

thomas

Glad to see you are having success with the strategies   I took a break for a little while to focus on manual trading.  

But basically, I still having difficulty with my evaluation process of the strategies produced by GB.

Can you share the evaluation process you have found that works for you to determine the strategies in your portfolio you showed above?  From GB generation until you put it in your portfolio?

-Stearno

tnickel,

Do you care to post some of your profitable strategies in here?

Hi,

stearno:

In the evaluation process  I use only a very short time for the strategy generation. I check all this found strategies in a longer timerange.

I hope this will prevent curve fitting.

john99:

No, I don´t share this strategies.

At the moment I check some strategies on an real account. And the rest on demoaccounts.

The portfolio looks good in the simmulation. But for forrex it is not good enough.

The period is 13 years. 80000 Euro in 13 years is too less.

I need strategies on M5 with more trades.

At the moment some questions are open:

1a) What data should I use for strategy finding?  If I use the realtickdata form a specific broker, are the results better If I trade this strategy on this broker?

2a) Should I test a strategie with tickdata of different brokers ? Or have every broker a specific charateristic, so it makes no sense to check

strategies with tickdata from different brokers?

3a)I found winning strategies who works good on Demoaccounts and Realaccounts.

… so it is fine….

In the next step I backtested this found strategies on GB AND Metatrader.

The backtestresults on GB and Metatrader are the same. (Strategie goes sideways)

Where comes this difference (Slippage ?)

Have someone made this type of tests:

1b) Backtest on GB

2b) Backtest on Metatrader

3b) Test on demoaccount

4b) Test on Realaccount

==> after 20 Trades compare this all.

thomas

Thomas,

Thanks for sharing.  Yes, I have come across these same questions.  

But to be honest, what I have come to the conclusion of is that I will not know these answers until I go through the entire process all the way to the Real Account and comparing those results to the backtest, walk forward test, and dmeo test results.  Then, and I think only then, will I know if these tests are actually predictive to the real world results.

I have taken time away from strategy development because my discretionary trading started to fall down.  I have gotten that back to be consistent, so I was looking back at making new automated strategies.  Before I took a break, I just had too many strategies fall apart as soon as I changed to a different timeframe or different currency pair (curve on created data set was beautiful, and then just went completely in opposite direction on new data set).  That is why I was asking you what you have found to work.

I think basically, we as a community can make this learning curve shorter for all of us if we worked together to get answers to the questions you posed.  If we all tried different backtesting processes and methods and then we took them to demo and then took them to live testing on different brokers, then we could report back onto this forum the results.  Then, we could quickly see what worked and what does not. We would get answers to your / our questions so much quicker than doing it alone.

What do you think?

-Stearno

Hi stearno,

It is very easy to generate good strategies (strategies with good equtiy curves)

But the most strategies are not running in real life.

Here is the generation process of this portfolio:

A69 EURUSD M15, Finfx

Idear1: short timeperiod with realtickdata from finFx
18.10.12-31.12.12-5.02.13
Spread 2, 0,5 Slippage, Tick Simmulation
Sl/Tp 5/120

1) Random Generation
2) Backwardtest with data from forextester 1.1.2001-31.03.2013
Spread2,0

(long period Test)

I installed this portfolio on a Demoaccount and run over a period 23.04-3.07

Here is the result.

Idear 2:

In the next idear I testet  strategies form step 1) with data from different brokers and take only the best strategies.

If I do this additional test the success will better.

From 15 strategies is one  ore two strategies good on demoaccount (I think real account will work too because this are not scalper, I have not testet this on real account)

What is to do in the future to find good strategies?

1) use the robustness test in the GB 3.0. This feature is genial, I can´t say more about this feature.

2) Install the good testet(with robustnesstest) strategies on demoaccounts

3) compare the trades of this strategies of demoaccounts with the backtester in Metatrader and Genetic Builder.

4) Check this for M15, M30, H1 (M1… M5 is a little difficult)

5) post your found strategies here in the forum 🙂 🙂

thomas