I'd like to start this journal to test a discuss ideas mainly pertaining to the algorithmic realm. I been trading for 10+ years so i have a general idea and a lot of biases for the good or the bad but I want to go back to the "beginner" mindset stage with a theoretically empty canvas so i thought this would be a good time to share the process. I like algos, they are limited in analytical capabilities compared to the human mind but neither do they possess fear, greed, inconsistencies, lack of energy to focus, you know and this is also a good study in my opinion for any human to enhance their understanding of systematic trading, which can be done by hand as well.
The coding will be mainly in python, which is a language suitable for this task perfectly. We may call this low frequency trading and I don't plan to aim for anything sub-second or to focus to gain the edge in the race of speed but to focus on risk management, sizing, positioning, entry timing etc. generally what a human trader would aim to do.
At this point the test are also somewhat market agnostic, I do have a preferred market (ES futures) but right now i need a place where i can run multiple sim accounts simultaneously and fx brokers are better at that (to my knowledge) so it really doesn't matter at this point if the algo's trade on a etf, cfd or futures because i want to start with observing general patterns and actions and their results. The broker for the sim accounts i have in mind is darwinex or oanda. Both can run multiple sim accounts. Darwinex has a great system of looking through historical performance since it is also a trader - investor hub but I'd be requiring to ran a "middle man" to access it, either MetaTrader 4 or 5 on the other hand Oanda has REST api, which i can access directly. I will test both. I haven't seen a great sim environment in crypto but later down the line when we get to systems that are worth to run live, I will add crypto as well. I'm mostly familiar with Centralized crypto and will probably stick to that, rather than DeFI but as i said, open mind. The best solution would be to trade through Interactive Brokers if i could create infinite amount of sim accounts, which might be possible in the future.
To test ideas i will use Jupyter notebooks, which are interactive notebooks that can run code. For heavier backtesting tasks I plan to use either QuantConnect, custom frameworks. I'm also hearing good info about VectorBT and its speed so one day I'd like to learn it but at this point speed isn't the main concern in backtesting, although it will be important down the line.
Most importantly I'd like this to be a fun research project.