EA and Backtesting

 

So I have read a few old posts here about backtesting and clearly I have learned that backtesting does not reflect reality! I made a research about that because I bought an EA and after a week the gain on my real account, it didnt reflect the gain in backtesting at all. All those videos and GIF on the MQL5 Market are not reliable then. Its a little bit confusing because people are selling EA based on fake result.

I also read that some EA that didnt do well in backtesting were doing really well on Live account. This confused me a lot because then how do you know the strategy is good? What are a rules good EA coder do to make there EA stable and profitable?
Let's just put away Martingale EA system, I'm talking about EA with stop lost and Take profit.

What are the best thing to do for backtesting your EA? Could you put your EA on demo account for 1 month then say thats it next month its going to be similar? with backtesting I have created good EA that are 120% profitable one years but then next year its not profitable at all. 

 
carl_carl101:

So I have read a few old posts here about backtesting and clearly I have learned that backtesting does not reflect reality! I made a research about that because I bought an EA and after a week the gain on my real account, it didnt reflect the gain in backtesting at all. All those videos and GIF on the MQL5 Market are not reliable then. Its a little bit confusing because people are selling EA based on fake result.

I also read that some EA that didnt do well in backtesting were doing really well on Live account. This confused me a lot because then how do you know the strategy is good? What are a rules good EA coder do to make there EA stable and profitable?
Let's just put away Martingale EA system, I'm talking about EA with stop lost and Take profit.

What are the best thing to do for backtesting your EA? Could you put your EA on demo account for 1 month then say thats it next month its going to be similar? with backtesting I have created good EA that are 120% profitable one years but then next year its not profitable at all. 

A backtest can be very similar to real live trading if :

1. The strategy doesn't depend on trading factor which are not well (or impossible) simulated with the Strategy Tester. For example slippage, volatility, bad connection, swaps, etc...

2. If the EA is well coded.

3. If the author is honest and didn't cheat in any way.

So when you see a backtest which is not the reflection of reality, that means one (or several) conditions above are not met.

Demo vs Real account results can also be different due to 2 first reasons above.

 
Alain Verleyen #:

A backtest can be very similar to real live trading if :

1. The strategy doesn't depend on trading factor which are not well (or impossible) simulated with the Strategy Tester. For example slippage, volatility, bad connection, swaps, etc...

2. If the EA is well coded.

3. If the author is honest and didn't cheat in any way.

So when you see a backtest which is not the reflection of reality, that means one (or several) conditions above are not met.

Demo vs Real account results can also be different due to 2 first reasons above.

Thanks for your answer!

I know there is slippage, volatility, bad connection, swaps, etc... but like for a simple EA with one Indicator like the MACD should it be almost the same gain if you were trading manually?

Generally an EA is coded so everyone can use it with different brokers,slippage and bad connection level with close to similar result.

You can even see certain EA on the market that have those: put 1000$ and get 900k$ in 5 years chart in GIF then you check the live signal and its making 20-40% a years, (which is really good) but clearly doesn't reflect the backtesting.

Any thought on that?

 
carl_carl101 #:

Thanks for your answer!

I know there is slippage, volatility, bad connection, swaps, etc... but like for a simple EA with one Indicator like the MACD should it be almost the same gain if you were trading manually?

Generally an EA is coded so everyone can use it with different brokers,slippage and bad connection level with close to similar result.

You can even see certain EA on the market that have those: put 1000$ and get 900k$ in 5 years chart in GIF then you check the live signal and its making 20-40% a years, (which is really good) but clearly doesn't reflect the backtesting.

Any thought on that?

What trading manually has to do with this ? We are talking about EA and backtesting.

Using this or that indicator is irrelevant.

There is no "generally an EA is coded...", a lot of EAs are not well coded that's a fact.

A backtest can be done in a lot of conditions which can lead to unrealistic results, without even talking about cheaters.

Going from 1000$ to 900K$ will never happen in real life, don't fool yourself, it just doesn't exist.

 
Alain Verleyen #:

What trading manually has to do with this ? We are talking about EA and backtesting.

Using this or that indicator is irrelevant.

There is no "generally an EA is coded...", a lot of EAs are not well coded that's a fact.

Already answered for the rest.

While trading manually The MACD indicator on the 30M chart gave me a 60% win rate with ratio 1:1 but when backtesting my account goes to zero.

 
carl_carl101 #:

While trading manually The MACD indicator on the 30M chart gave me a 60% win rate with ratio 1:1 but when backtesting my account goes to zero.

Because the EA is not coded the same way as what you are doing manually.
 

Alright!

How can we make sure the EA that we buy or create has a 90-95%-100% accuracy to the market when doing backtesting?

Is there something that people can read or check a video about it? This will help people to buy the right EA that they want and also help them to test the EA the right way.

 
carl_carl101 #:

Alright!

How can we make sure the EA that we buy or create has a 90-95%-100% accuracy to the market when doing backtesting?

Is there something that people can read or check a video about it? This will help people to buy the right EA that they want and also help them to test the EA the right way.

How to Test a Trading Robot Before Buying - MQL5 Articles
How to Test a Trading Robot Before Buying
How to Test a Trading Robot Before Buying
  • www.mql5.com
Buying a trading robot on MQL5 Market has a distinct benefit over all other similar options - an automated system offered can be thoroughly tested directly in the MetaTrader 5 terminal. Before buying, an Expert Advisor can and should be carefully run in all unfavorable modes in the built-in Strategy Tester to get a complete grasp of the system.