1[2 {"id": "C1", "text": "Over 60 origins (2021-09 to 2026-08, h=1), the ridge forecast has RMSE 0.561 vs 0.883 for naive, a 36.5% RMSE improvement.", "results_keys": ["rmse_ridge", "rmse_naive", "rel_improvement_vs_naive", "n_origins"], "status": "supported", "caveat": "Current-vintage data; publication lags ignored."},3 {"id": "C2", "text": "The ridge forecast does not beat the expanding mean: RMSE 0.561 vs 0.544 (relative improvement -3.1%), and MAE is also higher.", "results_keys": ["rmse_ridge", "rmse_mean", "rel_improvement_vs_mean", "mae_ridge", "mae_mean", "rel_improvement_mae_vs_mean"], "status": "supported", "caveat": "Difference is small and not statistically significant (C4)."},4 {"id": "C3", "text": "Ridge's advantage over the naive forecast is statistically significant (DM-HLN p = 0.0014).", "results_keys": ["dm_stat_ridge_vs_naive", "dm_pvalue_ridge_vs_naive"], "status": "supported", "caveat": "Only 60 errors; DM assumptions (stationary loss differential) assumed."},5 {"id": "C4", "text": "Ridge and expanding-mean forecast accuracy cannot be distinguished statistically (DM-HLN p = 0.59).", "results_keys": ["dm_stat_ridge_vs_mean", "dm_pvalue_ridge_vs_mean"], "status": "supported", "caveat": "Low power with 60 evaluation points."},6 {"id": "C5", "text": "All standardisation statistics and the target mean are computed from each origin's training window only; forecasts are invariant to corruption of post-origin data.", "results_keys": ["normalisation_window", "n_analysed"], "status": "supported", "caveat": "Verified by leakage test recorded in out/state/checks.md."}7]8