Discrepancy Between Backtesting and Live Trading Results

New comment
 

Hi !

I created an EA based on SAR indicator, I have excellent results in backtesting (over 3 years) but as soon as my EA runs on a real account, it's catastrophic!

Example: this Friday: loss -10$ in live in 18 trades, in backtest today (same EA): gain +50$ in 15 trades.

I'm lost, can't be only slippage and spread... I'm trading NASDAQ on CFD.

Is this something you encounter yourself, how to avoid it or take this bias into account in the EA?

Thanks for your help!

 
You may find this thread useful.
 

Based on how you created your system, you should understand how to perform an accurate backtest on it.

Ask and reply yourself to these questions:

  • Is the EA trading around rollover hour? If yes, have you ever considered that spread on these hours is huge and (especially on MT4) you can't reproduce it accurately?
  • Is the EA using functions that can be affected by a single tick (indicators checked on current candle shift 0, Ask/Bid prices used for calculations, etc)
  • Does your EA need an "every tick" backtest mode or can you rely on 1 minute OHLC?
  • Have you tried to do a forward test on a demo account?
  • What are discrepancies? Entries or exit? Both?
In few words, with informations your provided us, the only correct reply is "You performed backtest in a wrong way."
 
Fabio Cavalloni #:

Based on how you created your system, you should understand how to perform an accurate backtest on it.

Ask and reply yourself to these questions:

  • Is the EA trading around rollover hour? If yes, have you ever considered that spread on these hours is huge and (especially on MT4) you can't reproduce it accurately?
  • Is the EA using functions that can be affected by a single tick (indicators checked on current candle shift 0, Ask/Bid prices used for calculations, etc)
  • Does your EA need an "every tick" backtest mode or can you rely on 1 minute OHLC?
  • Have you tried to do a forward test on a demo account?
  • What are discrepancies? Entries or exit? Both?
In few words, with informations your provided us, the only correct reply is "You performed backtest in a wrong way."


Thanks Yashar and Fabio,

I'll read links you posted and try to answer questions you raised.

There is probably something around "every tick" backtest: OHLC and "every tick based on real ticks" seems less optimistic.

I should definitely try as well on a demo account and analyse discrepancies (they are huge: both entries and exits and indicators!).

New comment