Generating data for future testing

I’ve been tinkering with an LSTM algo for stock trading the past week, based on the StockBot LSTM project done at stanford university a few years back.

I can get it training and testing against historical stock data fine, but when I try to get it to predict future values its accuracy nosedives.

If anyone else has done LSTM’s for stock predictions, how do you do your future look stuff?
(disclaimer, I’m not all too versed with this kind of thing)