Algorithm trading: looking for project to join

Discussion in 'App Development' started by sptrader_79, Dec 16, 2019.

  1. Hi, guys

    I'm experienced C++/python dev, looking for some project to join for coding trading system on my free time. Have built so far Tensorflow/keras based signal app for scalping SPY, but would like to join some trading system development to learn more and contribute. I'm good with algos, C++/python. Ping me if have interesting project or idea and need someone to team up with..
    entropytrading, Fx-Game and fan27 like this.

  2. Here’s a project for you:

    Gather $SPY data 1 5 60 240 minute... create a database that stores it from Yahoo Finance (free).

    Use the Donchian Channel or Hi Lo of X bars back to create trendlines on those timeframes.
    TooEffingOld likes this.
  3. Thanks for good idea.

    I have all 1-min resolution data back from 1998, have temporal convolutional neural network trained on it which probably does better than most TA indicators on scalp range. Looking for smth more challenging, honestly... I would prefer joining some ongoing project to learn from others as well as contribute.

  4. traider


    How fast are you scalping? What is the average holding time of your trades? Short term horizon prediction can be done even without CNN but the trading costs are what kills the strategies.
    Also where are you based? Do you only work remote?
    931 likes this.
  5. Im in europe, my cnn targets price prediction for 15 min. I did this for learning keras/tensorflow and trying various architectures for time series prediction. I'm thinking of reinforcement learning for managing entries and exits, beyond vanilla NN. What are you guys up to?
  6. shatteredx


    Can you try running your neural network on live data to see how much money it could make?
  7. If actual trading counts for live testing then yes. I am more willing to share my experience and knowledge for teaming up with someone, not advertising own-developed trading system here
  8. I would really appreciate advice on how to become a 'proficient' algo-developer.
    I studied economics (useless) at college and worked in corporate finance (boring), so I kinda the typical 'traditional' finance type, but i really wanna become a solid algo-trader (independent).

    I have read those quantconnect/quantstart blogs and they do give good advice, and have read python and java books (up to Core Java/Effective Java). I know the syntax and programming concepts but I have no idea how to get started at all. I can't join any prop firm given where i reside and age, and no friends familiar with these quant stuff. My plan is as follows (any advice is appreciated):

    -Continue reading more programming books, including design patterns and latest Java constructs (lambda, streams, etc)
    -narrow in on understanding the network connection, IO streams, JDBC parts
    -read through the IB API documentation
    -practice first using Yahoo Finance API to get a hist DB, then move on to IB
    -read github IB project codes

    Can someone pls advise if there are other simpler/easier-to-understand broker APIs to play around first? I wanna get to know how to toy each aspect first. Thanks
  9. qlai


    Do you want to become a trader or quant/researcher/algo-developer? Learning how to code will not make you a trader, imho.
    nooby_mcnoob likes this.
  10. Be a good trader first, then code it. Many people think the coding is the magical thing that makes you profitable. It isn't.

    Check out these guys who are public with their progress and don't really do anything exceptional:

    The coding helps you scale but the problem that needs to be solved is you.
    #10     Jan 15, 2020
    rb7 and qlai like this.