Energy-Arena: Dynamic Benchmarking Platform Addresses Comparability Crisis in Time-Series Forecasting
arXiv·low signal
Kleinebrahm, Berrisch, and Eiser address the persistent comparability gap in energy forecasting research where reported accuracy gains are incomparable due to study-specific datasets, time periods, and scoring setups. Energy-Arena introduces dynamic benchmarking with standardized evaluation that updates over time rather than using fixed historical windows. The methodology — evolving benchmarks that prevent overfitting to stale test sets — generalizes to any time-series forecasting domain.