Topic: Curvefittingtests who have more Information about this topic

I have done the following things to prevent curvefitting.

a) different currency pairs
b) different timeframes
c) tests with long period (1.1.2001-now)
d) robustnesstests
e) Walkforward
f) Different dataseries (fhdb, duka, forextester, pepper, alpari, fxopen, finfx, armada)

This all is not enough to beat the curvefitting.

We need more curvefitting tests I think ?

A additional curvefittingtest is
Bootstrap
http://www.automated-trading-system.com/bootstrap-test/

thomas

http://www.tradessystem.com/

Re: Curvefittingtests who have more Information about this topic

you can consider these alternatives

do not use oscillators

use only price action indicators

use the higher time frames 4 hour and daily

Keep in mind that no system will run well forever...... they all have to be reworked regularly.

Even the Strategy Quant solutions that take forever to build have to be redone often

I am not aware of a system that does not need periodic revision, if you have seen one, please share it and we will ask the developer to have a long look.

I use MT5 a lot and I have tested many EA's in there for very long periods, they all have needed adjustments, even the ones with price action only. My best results have been with 8 hour or 12 hour or daily bars.

I feel that simplicity is a better approach to EA construction, the fewer the variables, the better and the longer it will operate.

Because each broker has different data and slippage a large slippage component may help your tests.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Curvefittingtests who have more Information about this topic

Blaiserboy wrote:

I am asking the administrator to move this to the correct thread (wishlist)

Better to keep this separated as this is a very important subject in its own right, and secondly the wishlist thread has grown into a clutter difficult to grasp.

Re: Curvefittingtests who have more Information about this topic

probably you are right Footon!

Actually I had started a few weeks ago, another thread, labelled 'CURVE FITTING' that this could be added to.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Curvefittingtests who have more Information about this topic

tnickel

How many indicators are you using in your strategies?

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Curvefittingtests who have more Information about this topic

Hi Blaiserboy,
I use the build in indicators in the forex Strategy Builder Pro.

In the old Version of  forex strategy builder I downloaded all fsb indicators from the web.

I found better strategies but most are curvefitted.

I think the FSBpro can find strategies in M5 timeframe too. The problematik with the curvefitting is the same in M5, H1, Daily etc.

If I search in M5 I have more bars for testing.

At the moment If I found a single strategy I have to switch to a different Timeframe, and make a new Test with a different dataseries. This have to be done hand by hand with one strategy after an other.

A solution which is in Strategy Quant will be good. I can make a new backtest with different settings on a list of strategies.
So I can generate 1000 Strategies for example and filter all strategies at the same time.

thomas

http://www.tradessystem.com/

Re: Curvefittingtests who have more Information about this topic

I found this information about strategy testing.
http://www.adaptrade.com/Newsletter/NL-GoodOrLucky.htm

thomas

http://www.tradessystem.com/

Re: Curvefittingtests who have more Information about this topic

tnickel wrote:

I found this information about strategy testing.
http://www.adaptrade.com/Newsletter/NL-GoodOrLucky.htm

thomas

Nice article, thanks

Re: Curvefittingtests who have more Information about this topic

Curve fitting is real, and also is the marketing of curve fitting...

Many writers fill pages about curve fitting with no concrete solution that is available for all to apply.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....