What is the best platform for backtesting?

Discussion in 'Automated Trading' started by jsingh1022, May 2, 2016.

  1. I have backtested strategies on both Tradestation and Multicharts. While the backtests produce results, if I comb through the results I find very obvious errors that the backtests made. Other people online have said that retail backtesting platforms like Tradestation and Multicharts are very prone to inaccuracy.

    Does anybody know what the most accurate form of backtesting is?

    Any thoughts on if cloud platforms like Quantopian are that much better?

    Thanks
     
  2. CBC

    CBC

    Ninjatrader. End of story.
     
    Patryk Jalmuzna likes this.
  3. userque

    userque

    Self-coded. End of story.
     
    Jeffro72, f2calv, VPhantom and 3 others like this.
  4. Thanks for the replies. I am familiar with Ninjatrader granted I've seen it named as one of the retail platforms that lacks backtesting accuracy. I don't have first-hand experience with it.

    For the person who said "self-coded": Where is a starting point to self-code a backtesting environment?
     
  5. userque

    userque

    You'd have to know/learn a programming language (CORRECTION: See below). I currently use Excel VBA. Soon to be using Python.

    As the saying suggests, if you want it done right.... :)

    Even so, I had no choice but to code it myself as my system uses custom machine learning algos.

    If your system is not that complicated, you could probably get away with just using Excel without the VBA part.?
     
    Jeffro72 likes this.
  6. d08

    d08

    The reason might be your code. I use an old version of Wealth-Lab and the results are consistent when the code is written properly.
     
    IAS_LLC and Xela like this.
  7. tommo

    tommo

    The software isnt as important as the data you use. Most platforms you can pick up for $100 a month has pretty unreliable data. Probably doesnt make too much difference on longer term systems. But intraday youre probably not going to replicate anything close to back test
     
  8. IAS_LLC

    IAS_LLC

    It also depends on the style of trading you're doing. If you are day trading, fill modelling becomes important. Just assuming you're going to get filled when a trade occurs at your ASK/BID isn't good enough. I assume that most of the backtesting for the masses apps just assume fills.
     
    userque likes this.
  9. tommo

    tommo

    Exactly. You need granular data, i.e can look "inside the bar" and only take limit fills on the back test if your price has traded through and not just printed there.. not to mention adding in additional slippage for stop orders
     
    CBC and IAS_LLC like this.
  10. lindq

    lindq

    All trading platforms are data dependent. Garbage in, garbage out.

    Before blaming your software, be suspicious of your data.
     
    #10     May 2, 2016
    ValueNGrowth likes this.