System development and trading

Discussion in 'Journals' started by RockMachine, Jul 10, 2024.

  1. The 2.3 Mod Sharpe was an average of 10,15,20 month periods.

    mod sharpe.png

    I expanded this out to 40 months and got an avg. 2.18 Mod Sharpe.

    mod sharpe2.png

    Most of the strategies are negatively skewed due to mean reversion in the short term - medium term. Screenshot above "QQQ" is a trend system that balances out neg skew. This was what was tested. In realtime, the trend based strategy "QQQ" is hedged with a put (not tested). Knowing this, I wouldn't expect ANY of these numbers to be precise. My main goal here is to keep correlation to a minimum and to aim for consistency via Mod Sharpe.
     
    Last edited: Jul 11, 2024
    #11     Jul 11, 2024
  2. I suspected that there's negative skew involved somewhere in there. You don't get these numbers with positive skew strategies. I don't know the details of your strategies, but I would be extremely cautious of risk warehousing and risk of blow up.
     
    #12     Jul 11, 2024
  3. You must be quite a heavy hitter!
     
    #13     Jul 11, 2024
  4. traider

    traider

    You do get them with really unique alphas but most people do not find these
     
    #14     Jul 11, 2024
  5. Today is a good wake-up call as we see several outside reversal bars in market leaders QQQ,SPY, NVDA, TSLA etc.

    I mentioned earlier that in my longer term strategy QQQ, that I wanted to buy an insurance policy for my long position via Puts. It made sense in my head and on paper, that if I was correct on my initial purchase, I could subsidize the hedge via the unrealized PNL of the 1st leg. Realizing that most option buyers lose and that EVERYONE loves to sell premium, I had to take the other side.

    Here is my current risk profile (untested) with a long biased trend position(tested). This is the the 1st roll of an already realized spread. The max loss here per contract is less than the initial subsidized realized PNL. So no stops required, it cannot lose and I would hope for a major event in either direction. Options can be the most conservative investment/trading tool in your Quiver.

    risk.png
     
    Last edited: Jul 11, 2024
    #15     Jul 11, 2024
  6. maler

    maler

    RockMachine, may I ask what fill assumptions you use in your backtest for the option positions in your strategies: do you cross the spread (lift ask for buys, hit bid for sells) on each leg when you get in and out?

    That would be quite conservative and it may be possible to get better fills in live trading.
     
    #16     Jul 11, 2024
  7. All strategies were backtested on underlying shares NOT options. All entries and exits were based on the 2 most accurate data points, the open and close of the day. Real time results are ONLY options and I always bid or offer mid point, then i cancel if not filled > then midpoint with less contracts (try this) and get a fill 30% of the time (fookin algo's). Generally the contracts i'm stalking have small spreads and are liquid.
     
    #17     Jul 11, 2024
  8. maler

    maler

    Thanks, appreciate clarifying this.
     
    #18     Jul 11, 2024
    RockMachine likes this.
  9. How curve-fit do you think your strategies are? Are all your signals purely price-based?
     
    #19     Jul 11, 2024
  10. poopy

    poopy


    Are you using marked PNL to buy the insurance or are you buying puts at the outset (synthetic call)?
     
    #20     Jul 11, 2024