Hey guys, I am new to algo trading. I know that many top hedge funds have complex quantitative systems. I wonder how do you guys take advantage of machine learning. Have any of the ML tools actually help you make better returns? Do you think machine learning guided trading is over-rated? I am thinking about developing some like that for myself. I'd really appreciate your input! Thank you!
If you have a successful manual trading system, quant background and programming skillset required go for it.
The question is faulty. Contemporary AI works best when it works in conjunction with humans. For example, you can train AI to detect when you're taking trades you really shouldn't.
Most supervised machine learning is really just automation. For me, I use machine learning to automate strategy construction, backtesting, filtering, combining, etc.
Many of the quant hedge funds have routing capabilities that are subpar nanoseconds. Also, AI and Algo could be similar, but they are different as well in execution and development so that you can cant use the term interchangeably. You need to realize what specific strategies you are going to deploy, their trade frequencies and execution.
I have an idea of the conditions and exit criteria that could make up a profitable strategy. My machine learning code will try different combinations on different slices of data, all the while filtering results based on my preferences. At this point, I have a bunch of strategy candidates. The next routine combines strategies based on user defined criteria and backtests, filters the different portfolio combinations with the output being a collection of strategy portfolio candidates. Manually, this process would take me months. On a fast machine, it can take a few hours.
I doubt it as I am performing zero parameter optimization and strategies must pass filters in and out of sample data. Plus, the profits in my live trading account point to strategy robustness.