How long should you backtest/forward test an EA?

Backtest - 0 time. The strategy tester is worthless in giving out any reliable results data. It is sad that the MetaOrg will not scrap it or fix it.

Forward - 1 month minimum. Even then, it will not guarantee long term total accuracy, but at least you will uncover many bugs in your program or your logic.

Do us all a favor, do not post any backtest results for they are not worth the time and effort in you doing the test.

There are forex systems developers. And there are fx programmers.

But there are EA testers. Means people who are doing it professionaly during the many years and who are well-know concerning this subject (i am testing but it is not me). According to their rules they may start to think about EA or to take some EA into consideration (and to give some information about EA to developer and programmer) after 3 month forward testing. And it may be some conslusion (preliminary conclusion) after 6 month forward testing only. It may be conclusion about how to improve it for example.

MT4 backtest vs forwardtest

hi,

I thought backtest result is only history data that can predict future data coz the most important is the same pattern appears over and over again..

=====================

PROVEN FOREX SYSTEM

In my personal view...

1. Backtest the EA over a long a period as possible.

2. Forward test the EA until you personally feel comfortable trading with it... In General this could range from 2-6 months (or perhaps even longer)

My own preference is before even backtesting the EA on MetaTrader is to actually manually backtest the actual rules. For example, printing off a chart showing a period of data, and then go with a pencil (to mark expected entry and exit points) along with a blank piece of paper (which you gradually slide across as you evaluate each bar) and pretend that I am actually trading for real. That way I seem to get more a feel of the characteristics of the rules within the EA and it's potential strengths and weaknesses.

With point 1. (Back testing the EA using the strategy tester) I feel the StrategyTester in it's current state is a rather blunt instrument, I still use it to identify periods where the EA demonstrates weaknesses, evaluate these weaknesses and work out how to best tackle this. At this point I may again even print off a chart showing a period where the E.A showed weakness and manually backtest to get a more detailed feel.

Once I am reasonably happy, I may start actually trading the EA in real time (Forward testing) whilse I continue to analyze the back tests. In my experience for about 97% + of E.A's and systems, this is about as far as it gets.

You may find that even after exhaustive back testing, that you may still have to apply a discretionary element along with Trades from your E.A. to get anything more than mediocre performance.

Interesting comments here. But has anyone noticed something really weird about backtesting/forwardtesting results?!

A backtest let's say from 3-28 July of a specific EA in a certain timeframe with a certain broker with 90%+ modeling quality and every tick setup will be way waydifferent than the actual forward test during the exact time with the exact circumstances and same broker.

If so, this means there is a major flaw in the programming of MT4 strategy tester, which inevitably raises the question of its usefulness at all. It is indeed a valid and legitimate question that should be addressed by MQ.