Recently came across blog post on reinforcement learning on stock data to trade. However in reinforcement setting agent should be able to play with simulated environment to explore and learn. However there is no way to simulate the price all it does is takes historical fragment of price and replays it.In my mind it is equivalent to historical data labeled and resampled with redundancy after shuffling. Am I missing something here?
Blog that I was reading was: https://towardsdatascience.com/deep-reinforcement-learning-for-automated-stock-trading-f1dad0126a02