Success on backtest but not in reality?

• 06/14/2016 at 11:21 AM #9376
  reihana9

  Hi Guys,

   

  Any reason why a backtested strategy would work great but fail when put into practice? Is this common or is it likely that, if backtest shows positive results over long period of time (3 years), there’s probably something worth pursuing?

   

  Thanks in advance.

  

  reihana9

  Participant

  Topics: 2

  Replies: 1

  Been thanked: 0 times
  06/14/2016 at 12:01 PM #9381
  K12AN

  Very common. It is to do with the shortcomings in the backtesting function – which shows that profits will always be achieved before loss is incurred.

  In real trading the losses can (and often do) occur before any profit targets are achieved.

  

  K12AN

  Participant

  Topics: 10

  Replies: 13

  Been thanked: 1 time
  07/19/2016 at 11:13 AM #10592
  Elsborgtrading

  Hi Reihana9

  I had similar experience, and also wrote a post on the matter. http://www.prorealcode.com/topic/the-holy-grail/

  I have some ideas to why that happened, however it’s not confirmed by any one yet.

  Br. Elsborgtrading

  

  Elsborgtrading

  Participant

  Topics: 37

  Replies: 359

  Been thanked: 154 times
  07/25/2016 at 9:22 AM #10866
  Stef

  Hi @reihana9,

  This is very common and may be caused by a number of reasons:

  1. The backtesting software does not work correctly. PRT, as an example, does not support look-inside-bar testing. If a target and stop is hit within  the same bar, PRT will assume that the target was hit first; which is not necessarily the case.
  2. You have not included spreads or commissions – or the correct spreads or commissions.
  3. Your strategy is curve-fitted. It was optimized/over-optimized for the data it was tested on. Read up on walkforward backtesting and monte carlo simulations for ideas on how to avoid curve fitting.
  4. The past is not a prediction of the future. Also, markets change. Try and test your strategy on different timeframes and different instruments/markets to see how robust the idea/strategy is.

  PS: A lot of strategies on this forum aren’t of much use, as they are curve fitted and a lot of the time do not include any costs/commissions. A strategy that backtests well without costs, quickly becomes unprofitable if costs are added.

  Hope this helps!

  Regards

  Stef

   

  

  Stef

  Participant

  Topics: 20

  Replies: 52

  Been thanked: 15 times
• Author
  Posts