A challenger appears -
1 2

thread: A challenger appears

  1. #11
    Junior Member jesusdom's Avatar
    8
    Bi-weekly WTF is going on about time!

    Therefore, reading a CSV document is technically not CPU bound, but doing anything of value with it will generally take you there.
    However, my CSV binary signal format conversion procedure is hardly CPU bound, if that's something, and generally plenty fast today.
    No matter how optimized, a CSV parser is still no match for a mmap'd fixed document file, so that's likely to remain for the time being.

    I'm starting to have an item structure that's not completely terrible, together with indiors and egies that plug into an engine.
    Originally, I had been treating candlesticks as some sort of unique non invasive data format for everybody else to use, but that was my MT4 bias.
    Candlesticks are only one kind of indior, and there's really nothing special about them. My object model makes it effortless for indiors to be utilized as foundations for egies and indiors, so that seems like a joyful means to deal with this.
    I'm still a bit iffy about the account side of matters. I'm probably missing a variety of things required to properly represent orders and transactions. For instance, I really don't track drawdown and numerous other metrics that are very likely to be helpful. Arguably, a lot of these can be calculated on demand, and there's something to be said about not storing the information twice, so perhaps that's fine.

    I had been hoping a look at QuantLib or QuickFix would give me a few hints about how to best structure my inner code, but that didn't happen. Either I have not looked at them close enough, or they do not encode particularly strong remarks about code that would interface together.

    So overall, I have a backtester that pilots a dummy account using an engine plugged into a special forex instrument for that I have 7 decades of tick data, and I can run a coin-flipping egy in addition to candlesticks at a particular timeframe, and shed lots of pretend money.

    Things are going nicely.

    Next measures, in no particular orders:
    - stop using double kind to represent prices.
    - extend the indior framework a bit, mostly in order to get old values (and marginally old values efficiently).
    - write a few primary indiors. See how well indiors enjoy renko bars function with this.
    - hook zmq in, designing a plausibly impartial interface to trading engines, determine if we can blow off some mt4 demonion dollars with this.
    - think about threading egy a bit. Not a clear match for easy backtests, but if the cpu price of a intrie egy begins to dwarf the price of cross-thread data passing on each tick, it may make sense.

  2. #12
    Junior Member zstssa's Avatar
    1
    I recommend reading Martin Pring on Price Patterns? It's among the best books on technical analysis and price action I've come across. It made a big difference in how I approach trading. Just wanted to throw that out there. Good Luck!

  •