Backtesting/Optimization - page 74

 

I know what you mean, that also happened to me. My friend said that maybe there's something wrong with the save data regarding the market. I hope you resolve this and post anything you find out. Good luck

 

simply setting them to zero should do.

 

If I set them to 0, no order is opened at all.

Before it was working. Is it due to a recent metatrader 4 update?

 

must be some problem with the logic then. normally, "0" means no sl/tp. you should try to debug the ea. try printing out the error codes and the parameters that you pass to OrderSend(), and you will spot the issue with certainty.

 

ES historical tick data

hello guys.

can some one help me to get ES historical tick data or 30 minutes bar data?

I'M LOOKING FOR DATA FROM 2000- TO 2007

i want to check a strategy and can not afford tp pay so much for it.

regards.

A.

 

Backtesting - Which period to use

Hi,

'm trying to find the best parameters for an EA, but I don't know which period to use for backtesting to generate good parameters for the future.

For example, my EA is taking into account the trade of an higher timeframe.

But during the backtest period, the higher timeframe market might be trading and then if I forward test during a trending period, the result might be bad.

I could also backtest during a period of trending market, and then it will perform poorly during forward testing in a ranging market.

If on the longer timeframe there was a permanent rise from 2000 until 2008, and then there is a sharp fall in 2009, then the parameters used for 2000-2008 could be very bad for 2009. When do we have to backtest again to find better parameters?

Anybody has an idea on how to backtest optimally? Which period should we use?

Daniel

 

Optimising the backtest period is complete madness (unless your trying to falsify results).

Test over as much data as you can find. Understand conditions in which the EA does well, and the conditions under which it doesnt do well and then use the EA as appropriate.

The approach should be a) does this trading idea actualy make any sense ?, b) does the idea backtest as expected, c) does the idea forward test as expected.

Fiddling around optimising parameters etc is a complete waste of time although there's no shortge of software that will allow you to do that

You want the EA to fail miserably, if it works, find additional data and keep testing until it doesnt work, if you cant break it, then maybe you have something worth pursuing.

Its also worth starting the backtest from different starting date's and looking at the variability in the results.

 

Tester small change:)

Hi All.

Who can do it?

Files:
t32.png  25 kb
t33.png  67 kb
 

people should never trust backtests, end of matter.

 

Hi,

I'm trying to optimize an EA, but I've some problem to understand how the strategy testers works when we use the "Open bar" model instead of the tick model.

I want to avoid using the tick model because it is really slow. I've 5 different parameters which have 3 or 4 possible values and using the tick model on a 15M chart for a one year period is so slow!!!

Are you guys using a shorter period for backtest?

Also my EA is always checking the previous bar. For example I check the close of the previous bar using "Close[1]", but it seems that the "Open price" model misses some bars. How is it possible? I've read that the tick model is mainly if we are opening trades for about 20pips and a 1 min bar can have a big impact, but I'm checking each bar only after it is closed on the 15min chart. How can the model miss some data?

Daniel