Stratgey Test Results Opposite of Demo Account Results

New comment
 
Anybody know why my EA performs well in strategy test but when I take it to live Demo account it performs poorly. This has happened twice and I am at wits end in trying to figure out whether running Eas on strategy tester is even worth the trouble because the results are so different when testing in Demo account.
 

It's not worth the trouble.

It's best when you test as close to reality as possible.

 
Differences between optimized test backwards and forward tests
 
Clifford Bryan:
Anybody know why my EA performs well in strategy test but when I take it to live Demo account it performs poorly. This has happened twice and I am at wits end in trying to figure out whether running Eas on strategy tester is even worth the trouble because the results are so different when testing in Demo account.

As Marco suggests and i advice to even get an external plugin to Mt4 where you'll be able to test including variable spreads and all other feautures as it was in a real environment or atleast as close to it as possible. I don't use MT5 but i think that one got feautures like that or a might be wrong......

 
Marco vd Heijden:

It's not worth the trouble.

It's best when you test as close to reality as possible.

Let me ask you something Marco. I saw an article you responded to on Applied Price. I use an EA builder to program my EA. There is a Price Cross of MA I have been using. It has applied price on it. I usually use applied price-close for my EA settings. It is opening trades if it is above or below the MA at the close of the candle. What would I use to open the order IMMEDIATELY when it crosses MA, not after close ?
 
Clifford Bryan:
Anybody know why my EA performs well in strategy test but when I take it to live Demo account it performs poorly. This has happened twice and I am at wits end in trying to figure out whether running Eas on strategy tester is even worth the trouble because the results are so different when testing in Demo account.

This could be anything but from the top of my head:-

1. Over Optimization

If you have 10-year data, you train your EA on the 10-year data. In this scenario, the optimization process will pick settings that fit perfectly (no pun) in the 10-year data. When it starts working with unseen data (demo/live) things go bad. Simple fix I do 5,3,2, Demo, Live. Means If I have 10 year data, I train on 5, optimize on 3, forward test on 2, then demo for 6-12 months finally live.

2. Time Zone

If the EA is time sensitive, always remember of the difference between the tester and broker. I once had a client who complained his EA would suffer negative swaps yet it was supposed to close all trades at 23:00.

3. Execution time, Slippage and Fixed Spread.

Well the tester doesn't suffer from the above.

4. Trust The Demo

As  @Marco vd Heijden said "keep it real" trust the demo more than the back test.

Also remember that when going live you are bound to suffer more slippage and slow execution than in the demo.

 
Clifford Bryan:
Let me ask you something Marco. I saw an article you responded to on Applied Price. I use an EA builder to program my EA. There is a Price Cross of MA I have been using. It has applied price on it. I usually use applied price-close for my EA settings. It is opening trades if it is above or below the MA at the close of the candle. What would I use to open the order IMMEDIATELY when it crosses MA, not after close ?

A real time monitor and flag to set when the order has been successfully opened otherwise you will get a cascade event of opening orders.

You can read the indicator values on every new bar but also every 250ms so that's up to you, but i would think that is hard to do when using EA Builder.

New comment
Reason: