Fully automated futures trading

Discussion in 'Journals' started by globalarbtrader, Feb 11, 2015.

  1. aukie1939

    aukie1939

    Hi Rob

    I had a question around the SOFR data provided with pysystemtrade. It dates back to 1984, but SOFR futures only started trading in 2018. Is this random data or did you use FedFunds Futures data prior to 2018 to create the SOFR adjusted prices?

    Thanks
     
    #3971     Feb 25, 2024
  2. No eurodollar with spread

    Rob
     
    #3972     Feb 26, 2024
  3. aukie1939

    aukie1939

    Thanks - assuming you used a fixed spread of 26.161 basis points?
     
    #3973     Feb 26, 2024
  4. Yup

    Rob
     
    #3974     Feb 26, 2024
  5. cholo

    cholo

    Hi Rob, how are you?

    Finally I had more time to implement some of the new strategies in AFTS. Very interesting as always, thank you so much for sharing! I started to work on `Safer fast mean reversion`. My plan is to dedicate some instruments to Mean Reversion strategies and configure the dynamic optimization to avoid trading in theses instruments in the TF+Carry portfolio, as per one of your suggestions in the book.

    I have started to implement the strategy and my results are very different from what I can see in the book, so I wanted to ask you if you had some code or intermediate results for any instrument I could check with. I have looked in pysystemtrade but I didn't find anything there, maybe is somewhere else? Sorry if you had answered this already.

    Probably it is related with my market data, I use minute data and my results are much worse. My gut feeling is that you need to hold the positions longer in order to make money, ie if you fill a buy limit order with an scaled forecast of say 10 maybe you shouldn't place the sell limit order for a forecast of 9 and place it for a forecast of 5 instead. I suppose using minute data instead of hourly data can be relevant here? But this is just a gut feeling and I just probably have a problem in my implementation.

    Thanks,
    Luis
     
    #3975     Feb 27, 2024
  6. I haven't got minute data, but I do find a big difference in performance between unlagged and one hour lagged returns, and also better returns with lower capital and lumpier positions, which suggests that waiting for the market to move more would be benefical (some kind of threshold on the forecast before taking a position is standard for mean reversion strategies).

    Rob
     
    #3976     Feb 27, 2024
  7. jiikoo

    jiikoo

    Has anyone experienced stack_handler crashing because of asyncio.exceptions.TimeoutError, and how to fix it? Client connects to IB Gateway, which pulls data from IB, but session ends in timeout. I wonder what's going on, yesterday evening everything was fine. Wrote longer description in github discussion.
     
    #3977     Feb 27, 2024
  8. wopr

    wopr

    Have you lagged your equilibrium calculation?

    I've found that to be the big difference in results, I posted the details here https://www.elitetrader.com/et/threads/fully-automated-futures-trading.289589/page-392#post-5906622
     
    #3978     Feb 27, 2024
  9. There is a general methodology for this type of things. Assuming your position size is proportional to the strength of the signal, you want some sort of min signal size from where you can grow your position linearly while capping it at some max position. Ie min(limit, max(0, signal - threshold)). There are some undesirable qualities to this approach, but the advantage is that it’s very simple
     
    #3979     Feb 28, 2024