I am engaged in algorithmic trading, employing specialized models that utilize various parameters to signal trading opportunities. Once I get a signal, I could execute a trade and everything is determine in advance, i.e. I know the entry point, the exit point and I have a stop loss if the price goes against me.

While the existing models are functional and able to generate these signals, I'm facing a challenge in optimizing the parameters to maximize the expected value of my trading strategy. I'm aware that this optimization is likely an iterative process, but I currently lack a systematic method to tweak the parameters effectively.

One example of this challenge is the identification of a "gap-up short pattern." A threshold such as a 70% gap in pre-market trading affects the expected value, and I suspect that there may be an optimal percentage that can enhance my results further. However, I'm uncertain how to explore different thresholds methodically to find this optimal value.

I'm keen on exploring optimization techniques that are tailored to my situation, taking into consideration my data constraints and the nature of the trading models. My goal is to find a robust yet practical approach to iteratively fine-tune these critical parameters, thereby maximizing the expected value without risking overfitting or adding undue complexity.

I have not yet embarked on any specific optimization method, and I'm actively seeking guidance on the best course of action to address this specific challenge.

EDIT I have thought using a reinforcement learning approach, but I am not sure at all it would be possible. I have about 2000 to 3000 signals per models so the data seems a bit limited.

  • 1
    $\begingroup$ Some comments: 1) This topic may be more suited to the "Quantitative Finance" site. 2) If you want to tap into general expertise, you either have to avoid narrow technical terms such as "gap-up short pattern" or you need to explain them thoroughly. 3) If you are looking for techniques which are "tailored to my situation taking into consideration my data constraints and the nature of the trading models" then you need to explain those in detail. 4) As a final remark: Maximizing expectation in algorithmic trading does not seem to be a good approach, as it ignores risk. $\endgroup$
    – g g
    Aug 10, 2023 at 13:00
  • $\begingroup$ @gg I do not ignore the risk. I know in advance my entry point, my exit point and the stop loss point. if the movement trend goes against me, then the stop loss just kick me out and protect me from losing more money. I have thought of using Baysian Optimization approach. What do you think about that? $\endgroup$
    – David
    Aug 10, 2023 at 14:24


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.