Human-only analysis versus computational capability in trading

Determine whether there exists any form of trading analysis that a human agent can perform which is not computationally reproducible or achievable by algorithmic trading systems within the context of the broader debate on algorithmic versus human trading.

Background

The paper situates its contribution within the ongoing debate about whether algorithmic trading systems can surpass human traders across different market conditions. After reviewing advantages and limitations of algorithmic approaches and outlining the concept of a cognitive ATS, the authors articulate a central unresolved question: whether any human-performed analysis is fundamentally non-computable.

Their study builds a unified LSTM-based forecasting framework using both fundamental macroeconomic variables from the US and Euro Area and a broad set of technical indicators. By feeding the model with the same information a human trader might consider and validating through machine learning metrics and trading simulations, the work aims to empirically probe the boundary between human and computational analysis. The explicit question seeks a principled determination of whether any inherently non-computable human analysis exists in trading.