Download MetaTrader 5

EUR/USD back testing how far back???

To add comments, please log in or register
Write articles and earn money. We pay 200 USD for them!
Tim
128
Tim 2014.08.23 21:14 

I've recently been backtesting a system on the downloaded data from Metaquotes and a question arose that has been troubling me. Since the inception of the EURO was roughly speaking around 1995 and the subsequent pairing of EUR/USD in the early 2000's (please excuse any slight inaccuracies here), when would be a relevant point in history to back test a trading system to??? 1995??? 2000??? 2003 when the ratio of EUR/USD started consistently trading above 1:1??? 2008 post the bank collapse???

 I accept the FX to be quite random, but the people behind the trades must have changed considerably over the years ie from BIG banks, to Big Banks and Investment Companies to Big Banks Investment Companies and now retail traders on smart phones....

 So back to my original question if YOU were backtesting a EUR/USD EA, how far back would be relevant???

Simon Gniadkowski
Moderator
18001
Simon Gniadkowski 2014.08.23 22:22  
timdchambers:


 So back to my original question if YOU were backtesting a EUR/USD EA, how far back would be relevant???

3 years, assuming more than one trade a day is taken,  plus a variety of other symbols for the same time span . . .
Tim
128
Tim 2014.08.24 09:54  

Hi Raptor,

Thanks for the reply, maybe I am over testing but I've been running my test on the 15minute time frame from 2001! I'm generating just over 10,000 pips over that time, two issues though.

1. My system runs at a slightnet loss till 2003 (about 2,500pips), from then it picks up to finish at approximately 12,000pips net, clearly when I run the test from 2003 I achieve even high without the first two years drawdown....with this post I was seeking to find a reason why my system failed before 2003 and has succeeded ever since.....whether there was a significant change in the way the market was traded before 2003...or it could just be it didn't work then....

2. For some reason the back tester in metatrader seems to stop testing around Feb 2003 (when I'm running a test from 2001-2014) and I have to perform a second back test to complete the run to present day. I've just rerun from 2003-2014 and it stopped July 2013...weird.

 Anyway I i've attached the chart showing performance from 2003 to 2013, starting from 10,000 USD, trading at 1pip per $ it made just over 10,000pips with a further 2000 in 2013-2014. Is this any good??? 

 


SYED NAUFAL GADDAFI
1401
SYED NAUFAL GADDAFI 2014.08.24 10:10  
You need to do a forwardtest regardless.
Simon Gniadkowski
Moderator
18001
Simon Gniadkowski 2014.08.24 10:55  
timdchambers:

. Is this any good??? 


Show the results from the other pairs . . .
Tim
128
Tim 2014.08.24 15:41  

I am actually trading this system, ie forward testing and achieving excellent correlation with the back tested results.

As for other pairs GB/USD...shocking! The system only seems to work well on EUR/USD...I'm going to look into whether the day ranges that it trades between benefit from a 1hr shift to compensate for it being traded differently between Europe and London...but to be honest that's just a wild punt.

SackMyCook
39
SackMyCook 2014.08.24 22:57  
RaptorUK:
3 years, assuming more than one trade a day is taken,  plus a variety of other symbols for the same time span . . .



Why 3 years?
Tim
128
Tim 2014.08.25 08:34  

I was wondering the same thing about the 3 years. I've just completed the backtest for the GB/USD and as you can see from the attachment, its VERY different from the EUR/USD with exactly the same system. 

This chart shows performance from 2004-2014, net profit only 2000pips with significant drawdown. 

Tim
128
Tim 2014.08.25 10:18  

 

GB/USD chart showing the results 2005-2014 same system as the EUR/USD but with very different results.

Simon Gniadkowski
Moderator
18001
Simon Gniadkowski 2014.08.25 22:13  
timdchambers:

I was wondering the same thing about the 3 years. I've just completed the backtest for the GB/USD and as you can see from the attachment, its VERY different from the EUR/USD with exactly the same system. 

This chart shows performance from 2004-2014, net profit only 2000pips with significant drawdown. 


3 years, assuming more than one trade a day is taken

 IMO the time period is not critical, what is more important is the number of trades . . .  if your EA takes one trade a week you will need to use a longer time period.

 

Please post the Average win, Average loss, Number of wins, and number of losses 

Tim
128
Tim 2014.08.26 19:33  

Here are the results from my latest EUR/USD back test...nearly 2000 trades from 2005-2014

  

12
To add comments, please log in or register