by Fabian Krause, Jan-Peter Calliess
Continue on SelfMadeQuant
Suggested foundations for trading research and honest backtests and financial machine learning. Specialist prerequisites vary by paper.
Use only information available at the simulated decision time.
Preserve time order and avoid future information in training.
Establish an interpretable baseline before a flexible model.
Reconcile signals, executions, positions and P&L.
Include the costs that can erase a paper strategy.
Account for dependence and overlapping observations.
Choose a reproducible strategy, specify the trading rule, and test what remains after costs and estimation uncertainty.