It normally involves a lot of digging through new datasets and speaking to expert networks .e.g., not insider trading, but sometimes ideas are sparked by following some of the latest research on: ArXiv, GitHub, SSRN, RePec, and yes, sometimes even reddit, twitter and LinkedIn (all using scrapping bots).
Nowadays some techniques are found using open code repositories as well as through coding competitions, you can see more of the daily stream here https://www.ml-quant.com/
Leave a Reply