Duration of back test and forward test for relatively reliable results
Backtest on M1.
The bigger your backtest period the more you approach the generic median for settings which means that there will be down periods.
Example :
You backtest 10 years .The settings you get are optimal for 10 years .Whithin those 10 years there were definately losing periods.
So the question is , if your EA on forward test starts losing for 3 months straight , will you discontinue it?Why ? you have to wait 10 years.
Now if your backtest time is too small you are adjusting your EA for a specific set of traders trading at that specific time.
Example:
You backtest 1 day and apply the settings for the next day.It wont be as profitable as before and it may even lose.
The upside of short backtesting periods is that -theoretically- the trading entities that where in the market yesterday , might
still be in today with the difference that some may not be waiting to trade but to exit their trades (meaning the effect of their transactions
will have the reverse effect on the market ,than the one it had a day ago).
Then theres traders who havent participated in the market yesterday and enter today.Thats another offseting factor.
Broadly , and in theory , if you reverse yesterdays backtest settings -and all yesterdays participants close their positions today- you might
get something.But its rare .For instance the only time i can think of where a periods participants would exit is before NFP's report or Rate announcements.
If you use trailing , set it to look at closed candle close price and apply trailing distance there.
You avoid tick interpolation this way .
If you use zig zag , dont use zig zag.
It's really hard to say, since markets change a lot.
When asked about this question, I usually say at least two years. But you have to consider other factors, especially the volume of orders for a given period. I usually say at least 200 trades. I am not a Statistician, but speaking Statistically, you need to determine a minimum sample size to obtain a statistically significant result. It would be of not much use if you would backtest/optimize a system that only generated less than 10 trades in 5 years or so.
For forward testing, I guess it would mostly depend on how much confidence you have with your trading system.
Thanks a lot for all your replies.
you are welcome
It's really hard to say, since markets change a lot.
When asked about this question, I usually say at least two years. But you have to consider other factors, especially the volume of orders for a given period. I usually say at least 200 trades. I am not a Statistician, but speaking Statistically, you need to determine a minimum sample size to obtain a statistically significant result. It would be of not much use if you would backtest/optimize a system that only generated less than 10 trades in 5 years or so.
For forward testing, I guess it would mostly depend on how much confidence you have with your trading system.
Allow me to add this:
If your ea is restricted to a maximum number of trades then , when you are at that maximum (while testing) you are actually losing entry signals .
The fact you lose signals is not the issue here.The fact that you are by accident not able to evaluate every signal your strategy produces.
Run a full on in the beggining , no trades limiter smallest lot SL TP and evaluate your hit rate .
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
A newbie question ...
I will try to code my manual system in an EA.
It will work on 30 mins timeframe ...
How many weeks/months I should do backtest for relatively reliable results ...
And after optimization for the backtesting period, how many weeks/months I should do forward test for a relatively reliable EA?
Is there such a general rule?
Thanks a lot