So, if I parsed the training data correctly, one's output algorithm is completely agnostic to any actual market conditions. It's merely learned on the anonymize

So, if I parsed the training data correctly, one's output algorithm is completely agnostic to any actual market conditions. It's merely learned on the anonymized feature set of 130 variables. Making it qualitatively no different that any other abstract ML forecasting problem. There's no considerations of market microstructure, news driven events, leverage, etc?