LogFAQs > #962275434

It's not about "learning speed" it's about raw size of the data set to learn from. An AI can't make a pattern to base buying and selling logic if there's not enough data to use and test on. We don't have more data now than we did then because the environment is presumably different-- we have less. We can't go back to 2019 and use its data to build a training set for a different retail-driven market, back then we had years and years of back data to tap into.

Yeah you can take a stab at it but you're trading billions of dollars here you don't want to be incorrect. It's probably safer to use the old proven machine learning algorithms for now and just assume most squeezes won't take off like they did and figure out how to game the system gradually with a lot and lot of test runs.