Personal opinion here as I'm terrible at programming, but I went about it the opposite end from ideas I'd read about and wanting to confirm whether they were as accurate as they claimed. If you read through some popular trading strategies, it should give you a general idea, but most people use either technical indicators (ie trend/info about the prices) or fundamental info (information about the company) to choose stocks. Then they have some mechanisms to decide entries and exits. Sometimes, what is "common sense" like using stop losses can be detrimental with certain strategies, so it's really up to you.

I feel like the statistical concepts and finance textbooks go hand-in-hand; it's a powerful combination, especially in building trading algorithms. I recommend first thinking about a strategy you'd like to test and write out some pseudo-code to map out what steps you're algorithm needs to execute to achieve that strategy. That way, you don't have to learn all the documentation all at once. Instead you can learn how to do certain things in Quantopian as you progress through the steps that your algorithm will take. Took me about 6 months to write a working algorithm, with little to no experience with Python.

it took me about 3 months to build things that worked with real momey but another 5 years to get to the comfortable level of the algo game (I trade opm with my algo's). it really depends on what your goal is.....learning never stops

My goal is to get comfortable with creating a simple algorithm, or anything that would work based on the given financial terms. My goal is not to have a "perfect" algorithm, but I want to use my knowledge of conditional statements, loops, and things I have learned in python as leverage to get a grip on building algorithms. My gap in understanding comes with what exactly I must check for, or what goes into an algorithm. Essentially even a workflow to algorithms, or data or numbers I should focus on, and I feel like that knowledge would come from learning trading strategies but I feel like I shouldn't have to resort to reading finance textbooks for this. In terms of workflow, for example, I became comfortable with a typical data science/ML workflow, (import data, clean data, preprocessing, hyperparameter tuning, model selection, model fitting/training, model testing, and evaluation, loading model for deployment). Something like this I had no idea of for building trading algorithms.

I started working on trading algos about a month ,or a month and a half ago. For creating my first algorithm it took me a couple days. However it was pretty terrible. So take that as you will.