1[2 {"id":"C1","text":"Over the 60 origins 2021-09 to 2026-08, the ridge forecast has lower RMSE than the naive forecast (rel. improvement about 36%) and lower MAE.","results_keys":["rmse_ridge","rmse_naive","mae_ridge","mae_naive","rel_improvement_vs_naive"],"status":"supported","caveat":"Current-vintage data, publication lags ignored; 60 forecasts."},3 {"id":"C2","text":"The ridge forecast does not beat the expanding historical mean: its RMSE is about 3% higher (rel. improvement about -0.031) and its MAE is also higher.","results_keys":["rmse_ridge","rmse_mean","mae_ridge","mae_mean","rel_improvement_vs_mean"],"status":"supported","caveat":"Small evaluation sample; single fixed lambda."},4 {"id":"C3","text":"The ridge-vs-naive difference in squared error is statistically significant by a Diebold-Mariano test (HLN-corrected, p about 0.001), while the ridge-vs-mean difference is not (p about 0.59).","results_keys":["dm_stat_naive_vs_ridge","dm_pvalue_naive_vs_ridge","dm_stat_mean_vs_ridge","dm_pvalue_mean_vs_ridge"],"status":"supported","caveat":"DM test asymptotics with n=60; ridge and mean share the same intercept (nested-like comparison), for which DM can be conservative."},5 {"id":"C4","text":"All standardisation statistics and the target mean were computed from each origin's training window only (normalisation_window = expanding); an independent re-implementation with future data corrupted reproduced every forecast.","results_keys":["normalisation_window","n_analysed"],"status":"supported","caveat":"Verification recorded in out/state/checks.md."}6]7