Thread: Automation and backtesting

Originally Posted by ;

For my backtesting my dataset is 15 years. The system generated roughly 3000 pips from 7000 pips range. Once finished, I conducted 100 random test (random beginning date and random end date from inside the 15 years). Out of the 100 runs, 20 runs were negative (every run has its own positive and negative transactions but complete pips where negative). Now when I chose to test the system on a live account, how long the testing that is forward should be for?

Hello thats the question [about ahead testing and how much time it should be] which most people don't have any legal or definate answer or is going to have a kind of answer for example 'it atleast requires 5 months or even a year or 3 months' of ahead testing in real market conditions.It implies that for the majority of the people the time interval or the quantities of transactions will be relative.Because every one will probably be thinking differently based on their own belief system and outcomes that they got after analysing the backtested egy.

Now to answer your main question about

So 20% conducts are negative, though the initial run (15 years are favorable).

What's going wrong? Why a egy would work on a period but when attempting it on a random (sub) interval it neglects 20% of those random runs?

There might be two reasons
1] inaccurate price data [which I don't think that it could be but just a possibility]
2] The egy which you have developed may work well in certain market conditions and might fail in other conditions because markets are changing dynamic arrangement and it may have changed and thats why the system didnt work on the random runs which you performed.

Back-testing is for checking to see whether is viable. It should on no account make you believe it would function in a live market. Back-testing presumes you take every trade given by your advantage and gives you the benefit of hindsight. In a real-world environment that never occurs.

As for ahead testing, you should ALWAYS start in demonion. You are satisfied and once you completely understand your edge works, proceed to some tiny account. Over time you may either become more confidant in your trading, in which case you can gradually add funds in a way that does not place your overall financial health in danger, i.e. just use cash you can afford to drop, or you may eliminate confidence in which case you go back to the drawing board ans start again.

Great Luck

- Ric

Thank You for the input.

I guess it is based upon the egy a few have few transactions a day other have few transactions a week.


For 15 years I'd roughly 800 trades.

The other portion of your reply. Yes that is exactly my point, backtesting will never have a response if a egy has expectancy or not.

More stats however these are based on the day of the week, the system would trade only Tue, Wed, Thu..

For 15 years: Tue,3469.16,Wed,2707.15,Thu,917.55

For 100 runs (random start date, random conclusion date):

Here is how to perform the 100 random runs:

Focusing on Tue only:

In 100 runs: 34 were in the 48 runs (where complete was below 10000) we get 15.

So 30% approximately.

What do we get if we run 1000 times?

380 Outside 1000 runs are Unfavorable

I am not sure why a system will function for a period from X but if attempting periods either within Y and X or overlapping, it fails 30% percent of the time.


Any one has same conclusion?

Originally Posted by ;

I am not certain why a system would work for a period from X to Y but when attempting distinct phases either within X and Y or overlapping, it neglects 30% percent of the time. Any person has same judgment?

The market changes direction, from monetary policy, to world events, NFP or just the market changes instructions or the EA does not comprehend the change in direction.

Originally Posted by ;

quote The market changes management, from monetary policy, to world events, NFP or merely the market changes directions or your EA doesn't recognize the change in management.

So all backtestings are doomed to fail then?