You are asked to build a model that predicts stock prices from historical market data. The goal is to produce forecasts that could support downstream investment research, while avoiding common mistakes like leakage and unrealistic backtests.
How would you approach building a predictive model for stock prices?