Risk warning

All articles, strategies and indicators are just reflecting a single traders opinion and should be viewed as that.
I advice everybody to trade with a DEMO account!

Tuesday, May 3, 2011

Testing and experiences.



I found out that it takes a lot more time to back test and optimize a strategy than to write it. If you tried out any of the strategies on this blog please share your experiences. If you have any suggestions for improvement please share. Fellow readers will surely appreciated it.
Just put all experiences and suggestions in the blog post comments. In the one concerned of course....

Geplaatst door ctlprogrammer op 1:59 PM

11 comments:

  1. AnonymousMay 4, 2011 at 7:24 AM

    Hi,

    i admit that backtesting takes longer than writing the strategy. And in my opinion backtesting shows you finally what strategy should be written (e.g.) in ctl :-)

    From my experience backtesting with the old version of dealbook in ctl takes far too much time. The duration of one cycle over the data grows exponential with adding some more data. I don't know if it is different with the new version of dealbook. I think there is better backtesting software on the market than dealbook.

    I am a fulltime programmer and didn't want to learn the usage of a backtesting software which maybe don't entirely fit my demands. Therefore i programmed my own testing framework with an underlying database (the raw sql-data is downloadable on ctlforum.com). This gives me the possibility to backtest 10 years of data very fast. What i have learned from my backtesting is that i will never start a strategy without backtesting over at least 5 years of data :-) I could throw away most of my "great ideas" that should improve my basic strategy because they were based on special conditions in the last months. They were not profitable over the long run. My basic strategy is profitable but don't perform highly profitable in every backtesting year. Therefore i want to warn everybody not to backtest only the last months! If you get good results then you don't know if it works in the long run. If you get bad results (or not as profitable as you want) then you don't know if your idea is maybe the big winner over the long run.

    You must know the performance of your strategy (win/loss ratio, average winner/loser, max peak-valley...) in the long run to stay with your strategy. From my opinion it's the only way to get confidence. How will you have the guts to trade the next signals if you have a drawdown and you don't know if it is within the normal range of your strategy? You will certainly quit trading and looking for the next strategy.

    My advice:
    If you don't have good programming knowledge search for a good backtesting software and invest some money. Do the homework (backtesting + bugfixing the ctl-strategy), then start trading and you will see that you were never more relaxed at trading.

    ReplyDelete
  2. AnonymousMay 4, 2011 at 6:28 PM

    Does anyone think Microsoft Excel is a suitable platform for backtesting? Or have any experience using Excel for this purpose?

    ReplyDelete
  3. ctlprogrammerMay 4, 2011 at 11:06 PM

    Depends how good you are with Excel. The easiest way for back testing is to use Forex Strategy Builder. It is a free program.

    http://forexsb.com/wiki/start

    ReplyDelete
  4. AnonymousMay 4, 2011 at 11:09 PM

    Hi,

    yes it is possible to backtest with excel. There is an addon for excel at http://ta-lib.org/hdr_doc.html. With ta-lib you can calculate the common indicator values.

    ReplyDelete
  5. ctlprogrammerMay 5, 2011 at 1:04 AM

    If somebody has tried something tiny in excel like ema crossing or so and wants to share. Count me as one of the interested guys.

    I used to be pretty good with access but i haven't touched it in 10 years. Access can be used to but if you have to start from scratch it will take some time.

    ReplyDelete
  6. AnonymousMay 5, 2011 at 1:49 AM

    Hi,

    i have done my backtesting with java and mysql. BUT if you have knowledge with excel/access and vba then you can do it the same way like me. Make a new column for each ema. Translate the code of the indicator from ctl to vba (it's easy). Iterate over the data and calculate the values for the ema-columns and write them fix into the columns. Now you can write the backtesting logic and use each pre-calculated column to write entry points, exit points, profit/loss of each trade.

    ReplyDelete
  7. ctlprogrammerMay 18, 2011 at 10:21 AM

    Hello,

    Is it possible for you to send me some more information concerning testing with excel or access? Or if its possible send me some samples so i can see how it is done. I understand if you don't wanna share because you have been putting to much time and work in it. Anyway thanks for sharing your idea.

    If you want we can make a post out of it with detailed information. All credits to you of course.

    Thank you!

    ReplyDelete
  8. UnknownMay 21, 2011 at 3:52 PM

    Hi,

    i can send you the skeleton of my backtesting java program. It will contain the classes to prepare the database for backtesting. Also it will have the solution for the iteration over the data inclusively the synchronisation of different timeframes while iterating. Implentations for the indicators SMA and Bollinger Band will be included too. Of course i can't give you my strategy but the framework to create your own. ;-)

    ReplyDelete
  9. MikeJune 1, 2011 at 12:12 AM

    Hi,

    are you furhtermore interested in the backtesting java skeleton? I yes, how can i contact you?

    ReplyDelete
  10. ctlprogrammerJune 1, 2011 at 12:50 AM

    Mike,

    I don't know anything about java but the readers are probably interested. Can you write a post for me explaining and illustrating this?
    It doesn't need to be pages long but just the basics to get started. A kind of guest post. It totally up to you of course...no obligation.

    ReplyDelete
  11. ctlprogrammerJune 1, 2011 at 1:03 AM

    My email adress is ctlprogrammer@gmail.com if you want to contact me Mike.

    ReplyDelete

Subscribe to: Post Comments (Atom)