Financial data's own quirks
Financial time series follow the same general forecasting principles as other time series data, but they have quirks that make them distinctly harder to work with. One is volatility clustering, the tendency for large price swings to cluster together in time, so a calm market tends to stay calm and a turbulent one tends to stay turbulent, at least for a while. Models that assume constant, steady volatility throughout will systematically misjudge risk during these clustered periods.
Another quirk is regime change, where the underlying statistical relationships in the data shift, sometimes abruptly, due to a shift in economic conditions, policy, or market structure. A model trained entirely on data from one regime can perform poorly once the market shifts into a different one, which is a persistent challenge for any forecasting approach applied to financial data, however sophisticated.