Economic vs Statistical Forecast Evaluation
Background: Quantile-based trading strategies (QBTS) are frequently used in energy market literature to evaluate the economic utility of probabilistic price forecasts by placing limit orders based on specific quantiles.
Question / Future Work: It remains unclear whether and under what conditions the economic performance of trading strategies, such as QBTS or those based on stochastic programming, can serve as a valid substitute for (strictly) proper scoring rules in model selection and forecast evaluation. The potential for these application-based metrics to exhibit low discriminatory power or to incentivize non-optimal forecasting behaviors necessitates further investigation into the consistency between statistical accuracy and decision quality.
Why It Matters: Understanding this relationship is fundamental for the field of electricity price forecasting, as it determines whether researchers should rely on statistical scores or application-specific outcomes when selecting and validating forecasting models.
Evidence: From a practical perspective, this implies that economic backtest performance can be decision-relevant, but it is generally unsuitable as a stand-alone strictly proper scoring rule for model selection over full predictive distributions.
Metadata & Links
- created_at
- 2026-04-24T05:51:07Z