Pair trading died - hello massive trading / Chapter II

Discussion in 'Automated Trading' started by elabunsky, Oct 4, 2022.

  1. Keezilla

    Keezilla


    First. Tick based. Then decision about limit order placed / cancel each second
     
    #11     Oct 4, 2022
  2. Keezilla

    Keezilla

    About the algo. Notice him we trade all symbols. Algo made positive and negative basket site. And one site trade opposite another
    About the algo. We trade all symbols. Algo makes a positive and negative basket per side. And one trades opposite another

    And because we create it from a lot if symbols we have stability. Look at the photo where we only trade 5 stocks, it just won't work.
     
    #12     Oct 4, 2022
  3. Please. Who is your broker? lol. I just ask because IB has been playing games lately with marketable limit orders. I've shown them log files and the answer is simply, "we don't know. still researching it."
     
    #13     Oct 4, 2022
  4. Keezilla

    Keezilla

    https://lime.co/ or direct exchange.
     
    #14     Oct 4, 2022
  5. It looks like they charge $1000/month to use their API. You still pay commissions, exchange fees, regulatory fees and other minor/usual brokerage fees on top of that. That extra $1000/month just to enter orders through the API sounds punitive. Makes me feel like I'm being bullied. What on earth is that about?
     
    #15     Oct 4, 2022
  6. TrAndy2022

    TrAndy2022

    When it is on the tick level it is on the smallest level possible. So did you backtest your strategy with historical data on level 2 and orderbook data too ? Did you calculate how much P&L per day or month can be made based on actual or average volume data for your stocks ? And last, when you go to higher timeframes a lot of those restriction and low latency and those greater amount of limit orders could be much less restrictive and in addition you could earn more because you have a higher time frame thus more room to generate profit with increased average trade net profit. Did you factor in that too ? Did you make a kind of P&L curve based on time intervals then ? And why do you not go to higher intraday time intervals and make money that way with less sophisticated setups based on the required technology because it is much easier to make money here for the first. And then you got in enough money you could finance yourself for the tick based timesframes ? Did you also factor in those thoughts ? Let me know your answers.
     
    #16     Oct 4, 2022
  7. We collected tick data from Nasdaq Basic Feed (dxfeed). Average PnL showed 2-3% daily. We are using seconds as a basic timeframe and rebalance the positions each seconds after our estimation. It's not possible to use 1 min or 5 min timeframe - the algo will loose because we need to trade as much as possible and collect profit from the estimation to avoid possible losses.
     
    #17     Oct 5, 2022
    TrAndy2022 likes this.
  8. hanneswas

    hanneswas

    Thanks for sharing! Quite inspiring.
    What max spread is manageable for your approach?
     
    #18     Oct 5, 2022
  9. TrAndy2022

    TrAndy2022

    Maybe you should buy some very good quality tick data here and make more backtests to be sure for the process how you are doing it. Then you also have more data and can make more (forward) tests to see how it works in different time periods. Then I would like to see some results again, if you do not mind.
     
    #19     Oct 5, 2022
  10. This is not right way. It's better use self collected data for avoid problems with the model when it'll go live.
    PS. As I said base period for algo 1 second. It's possible use 1/2 second also.
     
    #20     Oct 6, 2022