10 milliseconds ohlcv time-series

I am working on an algorithm based on a second order differential equation solved by 4th Order Cash Karp Runge Kutta method.
As you know this numerical method needs high resolution time-series for the best accuracy. So just need to build one day OHLCV time-series dataset of 10 20 30 40 50 60 70 80 90 100 milliseconds for some assets whatever they are, the purpose is to know how the numerical error increases in function of step size from 20 To 100 milliseconds. So i can define the maximum time-serie step size for which i can use this algorithm.

Let me know if your platform is connected to a reliable gapless raw tick database so I can build the ohlcv time series for the model validation process.

Thank you

Comments

  • ptunneyptunney Posts: 246

    The data on CloudQuant Elite is the same data that we receive direct from the markets and use to trade every day so it is accurate down to the millisecond.

    You can backtest and do analysis based on that data within the CloudQuant system.

    There are, not unreasonably, restrictions around removing that data from the system. Hence there are in built controls around how much data can be printed to the console / downloaded. There are also restrictions around how much CPU time a single script can utilize before the system will assume it is just a runaway script consuming resources and will kill it.

    But if all you want to do is write some python code to build your own micro bars on a reasonable number of symbols then you should totally be able to do that.

    I would suggest you run through some of the demo scripts and follow some of the introductory videos on the home page and you will soon receive an invite to upgrade to Elite.

  • bm18855bm18855 Posts: 3
    edited March 2019

    hi
    concerning the CPU usage i need a runge-kutta numerical calculation time of 15 microseconds for each 10 milliseconds ohlcv of the first time serie.

    let me know how many time i can run the python code for the model validation process.

    thank you

  • ptunneyptunney Posts: 246

    The simulator makes no attempt to simulate the passage of time in relation to the time your script takes.
    It will simply process your code then deliver the next event.
    However, obviously, you do not develop a model that it so resource hungry that it could never be taken live.
    All businesses operate or ROI (Return on Investment).
    If your model is spectacularly profitable then facilities could be made available.
    However, if your model is so demanding on resources that you cannot even begin to test without blowing out your allocation then it will be a non-starter.

  • bm18855bm18855 Posts: 3

    I am a researcher and just in the pre-development stage project. Thank you.

Sign In or Register to comment.