Robot with great backtest result bombs on live demo?
-
Hi,
Why does a robot that is profitable on backtest fail so badly on live demo? Here is a picture of my equity growth on gold backtested over last year. The results in other years are not as impressive but still positive. Contrast this with losing 90% of my demo account in one night trading the robot live!
What are the variables I need to better control to ensure backtesting is more accurate? I used a 2 pip spread in the test. Could this alone account for the failure? Is backtesting just a really bad predictor of live trading? Is there a better alternative to testing robots on MT4?
Thanks for sharing your knowledge.
Mike
-
@mikehzzzz
I can't even see the equity green line, I suppose you are using very tight TP.
In this case the spread and latency have a lot of importance on the profitability of the trades. And you can set them, in the tester, smaller than in real conditions. -
Hi,
Thanks, I can make some adjustments for the spread and change to an account with lower spreads if necessary.
I am not using a TP just a trailing stop. I'm not sure how I would reduce the latency in MT4?
-
@mikehzzzz Probably you didn't test on a 'every tick' base?
-
@l-andorrà Thanks. I did test on every tick. If you enlarge, you can see this in the image.
-
@mikehzzzz show me the actual results, not the graph.
-
also, mt4 is a lie. that 10% history being incorrect is enough to totally change everything depeding on the time frame you utilize.
-
-
@mikehzzzz I had been making changes in an attempt to reduce the loss rate when it produced this result. It didn't reduce the loss rate from previous tests but the profit was significantly higher. I'll be interested in your thoughts. Thanks, Mike
-
Goodluck @mikehzzzz.
I can get result from backtest from 8 to 5M in 4 months. But its just too good to be true.
-
@mikehzzzz said in Robot with great backtest result bombs on live demo?:
Is there a better alternative to testing robots on MT4?
Mt5.
-
Thanks for the responses. It seems to be about the spread. Anything above 2 and the robots performance worsens. Adding a maximum 2 spread block reduces trades and worsens performance. Hmm?
-
build your EA's like training physically. Practice for worse case scenarios. 10 spread is where I like to set the bar for all EA's to be able to perform at.
-
@mikehzzzz did you get this sorted, mate?
Just building my first Goldbot myself and am curious. Thanks.
-
Hi Dave, kind of... My backtesting could not be repeated on demo or live trading with current spreads. The various bits of advice were probably spot on. The missing history, the low spread setting and the period of testing can all skew results. I've had lots of successful tests on different robots. Finding a robot that is consistently profitable over time on different pairs in all markets with larger spreads taken into account is looking impossible. That's what makes trying so enjoyable for me! Good luck with your robot!
