forex software

Create and Test Forex Strategies

forex software

Skip to forum content

Forex Forum

Forex Software, Forex Strategies, Expert Advisors Generator

You are not logged in. Please login or register.


Forex Forum → Forex Strategy Builder Professional → Monte Carlo - randomize history data

Pages 1

You must login or register to post a reply

RSS topic feed

Posts: 8

Topic: Monte Carlo - randomize history data

In the attached strategy, when using Monte Carlo (randomize history data ONLY), no matter I change only 1% or 99% of the bars nor the data range ATR, I always get the same horrible result. I find it strange that both parameters, regardless of their values, yield always the same result.

However the strategy behaves very well with differing indicator parameters.

Any clue why?

Thanks
Nuno

Post's attachments

USDCHF 1d.zip 344.72 kb, 3 downloads since 2014-11-06 

You don't have the permssions to download the attachments of this post.

Re: Monte Carlo - randomize history data

If you compare the strategy backtest with and without intermediate data, you'll see a big difference.
The strategy shows many ambiguous bars tested only on it's main data frame, which are rectified by using the M1 data. Monte Carlo - Randomize History Data test takes and modifies only the main data period. It does not use the shorter data. The result of Randomize Data test you see are near to the strategy result without using shorter periods.

Re: Monte Carlo - randomize history data

I see. But that limits the usefulness of the test. Do you think in the future we can have the monte carlo tests based on intraday scanned data as well?

Re: Monte Carlo - randomize history data

Do you think in the future we can have the monte carlo tests based on intraday scanned data as well?

I don't see what is the point of optimizing main data and using existing shorter periods. They will not fit in the new ranges. It is possible of course, to "adjust" them to fit in some way, but it is going out of the purpose of the test.

It is better to find strategies without ambiguous bars. There are many benefits to that - less data needed, faster backtest, less memory usage...

Re: Monte Carlo - randomize history data

Yes but the idea would be to update intraday data as well, and not only the strategy's timeframe. Or is this too intensive from a computation point of view?

Re: Monte Carlo - randomize history data

When the program randomizes a bar range , the original intrabar data will be useless because they will not correspond to the new range. A possible solution is to generate new intrabar data with the sole purpose of fitting in the bar. In spite of that it may work, it will not serve the main idea of the scanning - to provide more correct interpolation of the bar. Do you se the contradiction? We make a random bar in order to test the strategy on different data and at the same time we want to precise the price route in it.

Re: Monte Carlo - randomize history data

Yes, I see, and it's interesting.
Maybe it's good to leave it as it is and as a recommendation one should also check other interpolation methods besides the pessimistic in case of unexpectedly bad monte carlo results.

Re: Monte Carlo - randomize history data

one should also check other interpolation methods besides the pessimistic

Why not test Random method. I think it is very useful in such cases. The random method tends to go to a randomly chosen order in range. The chance of the nearer order is greater.

Posts: 8

Pages 1

You must login or register to post a reply

Forex Forum → Forex Strategy Builder Professional → Monte Carlo - randomize history data

Similar topics in this forum