Fully Automated Stocks Trading

Discussion in 'Journals' started by ValeryN, Jun 14, 2020.

Thread Status:
Not open for further replies.
  1. smallAxe

    smallAxe

    Sorry if this question already exist and has been asked before, but I haven't found it.

    How do you allocate capital per strategy? Is it the same amount? Or use optimal capital on a strategy that yields optimal returns?

    Thank you
     
    #541     Sep 13, 2022
  2. ValeryN

    ValeryN

    I think the idea of allocating by account or fixed fractions is limiting return potential. One example would be strategy using a lot of capital occasionally but mostly sitting dormant. Running it on a fixed fraction or separate account would make inefficient use of capital.

    In my case, all run on the same account and how much strategy is using is dictated by
    1. Associated risk, measured in 2x of historical max drawdown (<20%).
    2. Correlation = it's ability to play nicely with others. Generally don't want to see 2 strategies having DD at the same time, ideally - combined DD and days till recovery reduced after adding
    3. Then individual capital usage requirements. Average and peak usage, impact on margin.
     
    #542     Sep 13, 2022
    smallAxe likes this.
  3. ValeryN

    ValeryN

    Did some data crunching for the US interest rates last night. A teaser chart below.

    upload_2022-9-13_21-32-16.png
    Thread for discussion.

    Val
     
    #543     Sep 14, 2022
  4. smallAxe

    smallAxe

    Thank you very much. But somehow I'm still not so clear.

    By trading multiple strategies at once on one account, how do you set the position size for each stock, on each strategy, when all strategies are active?
     
    #544     Sep 14, 2022
    shuraver likes this.
  5. ValeryN

    ValeryN

    Position sizes are typically based on % of account size, eg 5%. Each strategy has max number of positions it can take per day & total.
     
    #545     Sep 14, 2022
  6. ValeryN

    ValeryN

    Those are projected ceilings for US Prime:

    6.8% by Mar 2023
    8.22% by March of 2024
    10.14% by March of 2025
    11.18% by March of 2027

    Unless fed pulls off something that never happened over the last 70 years.

    So +0.55% left from now till March before we hit historical maximum.
     
    #546     Sep 25, 2022
  7. KevinBB

    KevinBB

    Hi Val

    Those rates look quite favourable compared to Aussie interest rates around 1990. From memory they got to 18% or so.

    I'm not a borrower, but certainly hope that they don't go anywhere near their maximum during this rate increase cycle.

    KH
     
    #547     Sep 25, 2022
  8. ValeryN

    ValeryN

    Rates didn't go to 18% straight. In my analysis I used rolling 1,2,3 and 5 window, measured from min to max. Put simply - what was the max % change from 5y low to high starting from 1950.

    Hopefully this represents the worst case scenario since the system is way more leveraged now.

    And I agree on certainly not wanting to see anywhere near to those too. Fingers crossed.
     
    #548     Sep 25, 2022
    KevinBB likes this.
  9. d08

    d08

    Also ditched DigitalOcean not too long ago and went with Hetzner. I live in a place where neither electricity nor connectivity can be guaranteed, so cloud services are an absolute saviour. I also see it as being extra careful, both the VPS and my physical connection are unlikely to fail at once.
    The VPS is just an Ubuntu instance with 3 Epyc cores, 4GB of RAM which normally doesn't run a GUI. Everything had to be explicitly logged which took some effort over the years but it's been stable.

    I don't really write tests for my code so typically it's me who messes things up.
    Other times it's the updated Python packages which cause issues, especially if the changes are silent and cannot be discovered without reading the update log.
     
    #549     Sep 26, 2022
    ValeryN and shuraver like this.
  10. could you post a real rate vs mkt index chart?
     
    #550     Sep 26, 2022
Thread Status:
Not open for further replies.