Backtesting Forex

[ilia king]

I'm not sure whether this should go in the software forum or this forum - as its a bit crossed in the middle. But here I go (Sharky, feel free to move this anywere you want.)

I would like to know were I could get a free trial of a downloader that would get me EOD forex data, maybe even historical data as moving from the Shares in the UKX to forex I would like test out my trading system.

Any replies would be greatly appreciated - thanks very much. :idea:

 

[JTrader]

Hi

I would like to gage other traders opinions of how much historical intraday forex data is required in order to gain reliable/representative results, or to be sure that my results are based on enough data history. If i intend to test 1 minute data on a 5 minute - 60 minute timescale, what is the minimum amount of data that you would recommend? 1,2,3,4,5,6,7,8,9,10,11,12 or more months?

Thanks

jtrader.

 

[SvenFoster]

illa,
There are a few but this is one that comes to mind.
http://www.dukascopy.com/english/cquotes/CSV/
your better off using the one with the list of symbols on it.

jtrader,
my view on this is that the more data the better (especially if your optimising) but I would be happy with 2-3 months if I was testing a system that would trade say 5 times a day.
Basically if I was confident that the data I had covered the majority of market conditions the system would meet. and on a 1 min chart that doesn't need 12 months IMHO.
As long as you don't fit your system to an unrepresentitive set of data then utimately it comes down to whether you have confidence in your backtested results.

 

[frugi]

Hi jtrader,

It depends somewhat on the system in question and how often it trades. A 1 min scalp system that trades 10 times a day would need less data than one that trades once a day, as Sven says.

As you know I use a system that trades once or twice a day.

Initially I test these systems with four months of data.
If the system does badly over four months then I discard it (that's four months with no income, after all).
If it does well then I test it on a year's worth of data, more if available. Three years would be ideal.

But a year will witness around 250 trades if you trade once a day and give a fairly good idea of viability. It is also a good idea to then test random quarters and individual months to see if there are prolonged periods of drawdown that weren't immediately apparent in the whole year.

 

[Newtron Bomb]

Ilia King

Check out Metatrader its free and you can do all teh back testing you care too.

While your at it you may wan to check out there yahoo BB and also moneytech as a lot of people use it there also.

Happy Back testing
NB

I'm not sure whether this should go in the software forum or this forum - as its a bit crossed in the middle. But here I go (Sharky, feel free to move this anywere you want.)

I would like to know were I could get a free trial of a downloader that would get me EOD forex data, maybe even historical data as moving from the Shares in the UKX to forex I would like test out my trading system.

Any replies would be greatly appreciated - thanks very much. :idea:

 

[JTrader]

Thanks

the EUR/USD & maybe GBP/USD systems i intend to backtest in TS2000i are likely to be on a 5 minute chart, I would expect them to trade half a dozen or so times a day, and have had 8 months historic intraday data in mind. Would this amount of data suffice for you?

Do you think that it could also be the case that it is not a good idea to backtest data too far back? as 1 years historical intraday may produce different results to 3 years intraday - which may be due to a fundamental change in the dynamics (volatiility, volume, daily range etc) of the instrument traded over the last 1 year period, which means that the previous 2 years data was outdated and didn't reflect the present market (as perhaps 1 years data may do)?


Thanks again

jtrader.