Backtesting/Optimization - page 8

New comment
 
Morpheus:
Sometimes it's fast and sometimes it goes way too slow. I don't know why. I found a 1.5 GB file in the logs and I deleted it but it still is slow. Is their a better way to backtest programs? I'm using Metatrader and I often only have 20 % modeling quality.

I am doing with 90 % modeling quality back to 2001 every tick and can you imagine how it is slow! Many hours! Optimization is going during the several days sometimes just for one pair/timeframe.

Sorry but it is MT4.

 

How did you get 90 % modeling quality? I used to have 70 % and then something happened and now only 20 % many times.

 

The modeling quality has to do with the quality of your data. If your just using the real time data there are a lot of holes and gaps. You need to download some quality historical data and load it in the history center.

Some free data for about a years worth of one minute can be found at

http://www.alpari-idc.com/en/dc/databank.php

You probably want more than a years worth. Maybe newdigital has some ideas where to get it.

The optimizer will always go slowly because "optomization" runs your system on every tick in your history using every combination of attributes you assign to it. This can quickly become millions, so it takes a while. Experience shows that if you're going to optomize you can reduce the number of runs by making the step value larger. Say you want to optomize your stoploss between 5 and 50. Rahter than step values of one use 5. You won't get much improvement at all doing every single one. They tend to run in patterns and you will quickly see the range that is innefective. As always remember not to overfit your system to the data you have, because the market will change and your system wont be robust enough to react to it.

Hope that helps reduce the time. Monte Carlo optomization can be a little faster but its not built in.

As for the other I'm actually surprised why the backtester is running so slow. Perhpahs you are recalculating every time you run the tester. You might try only doing this once. Good quality data should actually help speed it up as well, not sure how much though.

I don't know if that helped at all but feel free to shoot another shout out.

 

WOW newigital almost 4000 post. Thats impressive!!

 

It's depend on your computer too I guess. A newest dual core processor will speed anything up

 

Backtesting in MT

Hi,

I have heard that there has been some issues with with the MT backtester, has this been solved? I have written a EA and wanting to test it on 30 minute bars data, will results be reliable? And how far back will it test? I have data since 1986, but it is in ASCI, by default will MT backtest over the last 1 month, 2 months, 1 year??

Regards

Nick Frandsen

www.historicalfxdata.com

 

tester/live trading

Question 1 :

When i backtest my EA, in the report tab it says 0 trades were executed, in a 6 year period, wich i know should have been more, (obviously) as i have manually executed them before, the results and graph tabs are blank, and the journal says all it did was load my strategy,

Am i supposed to do something else i don't know about?

Question 2:

How come when i've attatched my ea to a chart and allow live trading, it still never executes any trades when it should?

 

maybe some parametors of your setting are not allow on your server like mine :

if take profit less than 5 and lot size less than 0.1 , the order will not be executed ...

check your setting then ^_^

 

well, the stop loss and take profit are greater than five pips away and i'm going at exactly .10 lots, and the strategy will attatch to the chart, so i'm cluelessly confizzled.

 

ok, well, i just learned the value of using the expert properties function of the tester, it turns out i was using .01 lots instead of .1, (even though i thought the coder i used only went as low as .1) thanks for the feedback!

now there's several strategies i may have found out how to make work on the tester.

& of thanks!

123456789101112131415...95
New comment