Out-of-sample recovery ====================== The fixed-factor recovery test reused the original annual factors for 2028-2031. This test keeps the first-year risk-ranked deferral fixed and resamples future cost, productivity, and capacity factors independently for each future path. Future years are re-solved with the rolling persistence forecast and a two-year lookahead. This is a stochastic recovery test, not a stochastic dynamic program. Method ------ - 20 rolling episodes. - First-year risk-ranked deferral: 20% target, 21.3% actual mean deferral. - Five independent future-factor paths per episode, 100 paths total. - 2028-2031 re-solved for each path. - Future factors sampled from the same lognormal assumptions as the rolling policy. Results ------- .. list-table:: :header-rows: 1 * - Metric - P10 - P50 - P90 - Mean * - Final risk gap - -0.0078 - -0.0010 - 0.0072 - -0.0008 * - Recovery - 0.0340 - 0.0421 - 0.0491 - 0.0421 * - Final gap within 0.005 - — - 87.0% of paths - — - — A negative final gap means the reoptimized path ends below the full policy's final risk. The median gap is slightly negative; 87% of paths are within 0.005 of the full policy. Cost bands: - reoptimized: 284,835-286,788 CAD, mean 285,868; - no reoptimization: 227,290-272,072 CAD, mean 251,312; - full policy: 273,213-326,472 CAD, mean 299,052. The reoptimized path spends more than no reoptimization to recover risk, but less than the full policy on average. Limits ------ - First-year deferral is fixed. - Future factors are independent draws; no serial correlation or calibrated forecast process. - Future years use a persistence forecast and two-year lookahead, not a stochastic dynamic program. - First-year risk ranking only. - No recurrence, weather, outage, safety, or asset-owner priority model.