results: backtesting vs market replay vs live

Discussion in 'Strategy Development' started by calhawk01, Jun 20, 2013.

  1. ALL ARE DIFFERENT!

    backtesting 1 min data vs tick data= results are different by a LARGE amount

    backtesting second data vs tick data= results are diff by a LARGE amount

    testing on market replay vs backtesting ^^ above = results are diff by a LARGE amount

    then live trading is totally different

    wtf?? how are we suppose to test?
     
  2. Bob111

    Bob111

    and you are talking about WHAT?
    stocks,options on stocks,futures,options on futures,CFD's,bonds etc ..WHAT type of security are you babbling about?
     
  3. Well given that mostly all products are highly correlated with one another .. This applies to all products. But to answer ur question.. Emini to be exact
     
  4. Eight

    Eight

    You are testing something that is essentially random? How do in-sample results compare to out-of-sample results using the same type of data?
     
  5. Ninja Trader?

    1 min data: backtest broken.
    Tick data: backtest STILL broken, in other ways.

    Get it? Crappy software = unreliable results.

    Want to laugh? Backtest the same strat on 2 computers, same timeframe and parameters. Chances are you get different results ;)
     
  6. NInjatrader indeed......

    Is live testing the only solution? Spent months programming these strats.
    Any solution here or does this just mean that..
    Big hedge funds win bc their software is bryter
     
  7. No, more like NInja is crap front to end.

    Was looking for a solution for months. Now I am about 4 man years into my own platform. Could not get it to work reliable.

    Problem is for example they have a 1 second timestamp and store bid/ask and trades in separate files- NO way to get the original order back. Just not possible.

    The limitations are plenty and well documented. For example you must trade against a tick-1 chart to get reasonable fills, even better against tick-1 on bid and ask for executions.
     
  8. Bob111

    Bob111

    if you spend months programming those systems-spend a little more and create your own soft for backetesting.
    i don't trust anyone. everything i have is made by me from scratch. i mean-everything(except OS and program used to create my trading soft). even PC's.
    yes,take a lot of time, but on other hand-you can do whatever the f** you want with it. complete flexibility. you don't know,why write this ninja stuff and what's in it. i know every line of my code. it's longer,but safer path.
     
  9. Craig66

    Craig66

    +1, this is the only way to get complete control over your testing.
    In back testing you need to obsess over every detail and use tick data, otherwise you're just fooling yourself.
     
  10. is your software for sale ?
     
    #10     Jun 21, 2013