Backtesting/Optimization - page 27

 

Backtesting Corrupts Historical Data?

Hello guys,

A few weeks ago I've downloaded 1 minute data from: http://www.alpari-idc.com/en/dc/databank.html

And backtested lots of EAs on it.

Today I've found an EA that gave me these results:

I decided to speed up the backtesting process so I've installed metatrader on my new computer, downloaded the newest 1 minute data, uploaded my EA

and got these results:

The settings for the EA are the same, the only difference which I can see is that the Ticks modelled are much higher on the newest 1 minute data.

I've noticed that other EAs which have been generating good results on the old data stopped "working" when I use the newest data.

Does anyone know the reason for this?

Was my old 1 minute data corrupted?

Thanks In advance, Drue

 

EA used 14416 ticks in first case and 14391 ticks in second 90% case. So it may be the following:

- you selected more lobngher period for backtesting in first case;

or

- you kept first Metatrader live for trading and Metatrader downloaded new data and it was mixed. Import additional data in fooline mode (if you open the folder with data so you probable will see many files with the same timeframe and pairs so import it to the main file).

And if you are backtesting EAs very often so it may be good to keep one copy of Metatrader just for backtesting only and this Metatrader should be switch off from the broker permanently.

 

Strategy Optimizer using GPU

Hi. I have been using the strategy test optimizer heavily recently, and I have found that it takes a long time to otimize a strategy, and there is a limit on how many variables one can have it work with. I have found that there is now software to allow use of one's GPU on their video card to do computing work, and the GPU is hugely faster than the CPU. In some tests, my CPU has about 5 GFlops, and my GPU has about 550 GFlops. That is over a hundred times increase! SO I am interested in modifying Metatrader4 to use the GPU instead of the CPU, and I want to get rid of the variable limit so that it can compute based on many more variables. I found a company that will do the modifications here:

http://www.gpucomputing.eu/

So I am wondering if anyone here is interested in this? I am sure that I will not be able to afford this myself (I am an unemployed college student), so I am hoping we can do sort of group-buy on this. I emaileed them asking for a qoute, and when I get it I will post it in this thread.

 
triwebb1:
Hi. I have been using the strategy test optimizer heavily recently, and I have found that it takes a long time to otimize a strategy, and there is a limit on how many variables one can have it work with. I have found that there is now software to allow use of one's GPU on their video card to do computing work, and the GPU is hugely faster than the CPU. In some tests, my CPU has about 5 GFlops, and my GPU has about 550 GFlops. That is over a hundred times increase! SO I am interested in modifying Metatrader4 to use the GPU instead of the CPU, and I want to get rid of the variable limit so that it can compute based on many more variables. I found a company that will do the modifications here:

http://www.gpucomputing.eu/

So I am wondering if anyone here is interested in this? I am sure that I will not be able to afford this myself (I am an unemployed college student), so I am hoping we can do sort of group-buy on this. I emaileed them asking for a qoute, and when I get it I will post it in this thread.

Before i start commenting could you please elaborate more on this variable limit you are having ..

thanks.

 

Interesting idea, but...

One unfortunate restriction of MetaTrader4 is that there is no API available, so you are sort of limited in what you can do. The only possibility MT4 allows is calling external DLLs, but that wouldn't really help I think. Most of the time-consuming part of the strategy tester, for example, would stay resident within MT4.

 

Yeah, I know that it would be difficult, that is why I am looking at this company to do it :P I think they would have to decompile Metatrader, if possible, and re-write some of it.

About the variable limit: I have a multi-staged EA that I am testing, and I use multiple indicators with different values for each stage, and I want the optimizer to run through them all and find the best values. It seems to be limited to something like 20 variables with 10 steps each. I want it to support about 80 variables with 50 steps each... That many variables would take a long time to compute, and for that I want either the GPU thing to work, or the ability to use a cluster of PCs to do it so that it doesn't take 2 years on my pc. But Metatrader doesn't support either, so right now I am stuck with trying different methods for the EA's operation, but I really want to try the one with the huge amount of variables and see what it comes up with.

 

Hi, here is what i want to do:

I want to let my EA optimize himself (an optimization of the last 2 months), and it have to change his variables automatically if he found better variables during the optimization (it will be optimizing 24h/24, all the day, all the days).

Do you think it is possible to code something like this? maybe a script or something that i don't know, what do you think?

 

Ya

I think This mmmmmmmmm Oky Oky

 

Do you think it can be possible?

 
w4rn1ng:
Hi, here is what i want to do:

I want to let my EA optimize himself (an optimization of the last 2 months), and it have to change his variables automatically if he found better variables during the optimization (it will be optimizing 24h/24, all the day, all the days).

Do you think it is possible to code something like this? maybe a script or something that i don't know, what do you think?

I don't know. It is very good idea especially for ema crossing EAs. But I do not see any implementation. Codersguru i having the thread about self-learning EA or something like that but development did not finish yet.

Reason: