Hi all,
For traders who created expert advisors with very successful backtest results, how are your EAs doing in live trading and how long have you used them?
Please share your experiences!
Please note that any discussion concerning EAs must also have the source code for the EA attached.
Hi all,
For traders who created expert advisors with very successful backtest results, how are your EAs doing in live trading and how long have you used them?
Please share your experiences!
it really depends on how you do the strategy (concept of how to trade)
this is not technical, but it is a concept question
as a programmer, I have programmed several EAs, I can answer:
I have some EA which perform up to 14 days exactly like the forward testing.
and I have other EA which will perform wrongly just after 1 or 2 days
It depends on the concept/strategy of operation that you have implemented on the EA for it to trade, so this is not a technical/programming issue.
it really depends on how you do the strategy (concept of how to trade)
this is not technical, but it is a concept question
as a programmer, I have programmed several EAs, I can answer:
I have some EA which perform up to 14 days exactly like the forward testing.
and I have other EA which will perform wrongly just after 1 or 2 days
It depends on the concept/strategy of operation that you have implemented on the EA for it to trade, so this is not a technical/programming issue.
Not exactly. The tester has a lot of peculiarities one has to take into account in order to trust for back-testing results to match live trading. The whole market place is full of tester-grails.
-The tester has unrealistically high positive slippage. In order to eliminate this, test on a netting account on a custom symbol in exchange mode and use only limit orders. Then there will be zero slip - not positive not negatve.
-Every tick based on real ticks mode is the most realistic. In fact i ran tick strategies (so not using bars or time frames). When ran live for a week and back tested the same week apart from slippage the results where 100% identical.
-In M1 OHLC mode, the EA has to be coded as such (only open and close positions when the previous bar is closed.
-in Open price only mode, again the EA need to be coded as such. If you introduce bars from other symbols on different time frame then the tester time rame, it is possible to trade the introduced symbol at prices not even possible, resulting in tester grails. To avoid always use the M1 time frame and load a pseudo indicator running on M1 for all symbols used in the strategy. This forces the tester to load higher resolution prices and thus more accurate.
-With both open prices and M1 OHLC, the spread is used from bars. The spread from the bars is the lowest measured for that bar. Therefor it is not realistic at all. Avoid using spread filters. Depending on the symbol the result can be 10-50% less doing the same test in Every tick based on real ticks because in this mode the real spread is used. For symbol where spread is low (EURUSD) the result may be 10% less, for GBPJPY it can be as much as 50%.
There are more things, these are the major ones.
If you do not know what you are doing, testing/optimizing is a crap shoot regarding expecting the same results in live trading.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Hi all,
For traders who created expert advisors with very successful backtest results, how are your EAs doing in live trading and how long have you used them?
Please share your experiences!