FX: Backtesting with intraday data

[nielsend]

My question is in regards to back testing a ‘Swing Trading’ system on the AUD.USD where trades can range from a few minutes to a couple of days.

For those of you that back test systems, how far back do you prefer to go. 1 yr, 2ys, 10 yrs? This may seem bit of a strange question, but I’ve found when back testing systems previously (stocks) that the further back you go the less relevant the results are due to fundamental changes in the market, i.e. mergers, splits, takeovers etc. For EOD systems I would assume the majority would prefer to go back historically as far as possible?

With FX intraday would you consider 12 months data enough to validate a system? I only have about 1 yrs reliable data, and the system over this period took (approx) 280 trades at an average of between 5 – 6 trades weekly. Results are very positive.

 

[waza1960]

I like to backtest over 2 years for any system designed on an hourly chart.
My belief is that if my time frame was shorter then my backtesting period should also be shorter and the reverse for longer time frames.

 

[tech/a]

I like to backtest over 2 years for any system designed on an hourly chart.
My belief is that if my time frame was shorter then my backtesting period should also be shorter and the reverse for longer time frames.

Thats 12 yrs of EOD data---2 yrs of 1 hr data.

Im interested in the formula for identifying swing trading in a system
You wouldnt happen to be able to post it would you?
Neer been able to find one.That Works.

 

[sinner]

Thats 12 yrs of EOD data---2 yrs of 1 hr data.

Im interested in the formula for identifying swing trading in a system
You wouldnt happen to be able to post it would you?
Neer been able to find one.That Works.

See here: https://www.aussiestockforums.com/forums/showthread.php?t=20860

In regards to the original poster. Use the highest resolution data (M1 or tick) available for as long as you can regardless of timeframe. I usually test in financial quarter 3 month blocks. Testing daily forex systems I usually do the EURUSD all the way back to its inception, same with GBPUSD.

I am not interested in tests with less than 90% modelling quality.

 

[SmellyTerror]

You used all the data you had to test and (critically) optimise?

Very VERY important to have a good slab of data untouched by man or djinn- errr, that is, data you haven't used to optimise the system, so you can accurately test that it works.

Easiest thing in the world to make a system that's absolutely perfect on historical data. Unless you've kept data back, though, to test the final version - seriously, final final, no messing with it after the test - then you haven't actually tested it.

Sorry if that's obvious, but it's a very easy trap to fall into, and is much more important than how long your backtest / optimisation period is.