{
  "created_utc": "2026-10-15T04:13:07.157747+00:00",
  "rule": "config/bias_correction/recommendation_rule.json",
  "rule_contents": {
    "description": "Rule by which the platform suggests a bias-correction method for each cycle, target and product (scripts/bias_recommend.py, docs/47). Fixed before the independent validation seasons are scored. A suggestion is advice for a decision; it never changes the published forecast.",
    "benchmarks": [
      "config/bias_correction/jan_fmam.json",
      "config/bias_correction/feb_fmam.json",
      "config/bias_correction/may_jjas.json",
      "config/bias_correction/sep_ondj.json"
    ],
    "products": {
      "probabilities": {
        "score": "rps",
        "selection": "SEL-P",
        "label": "Published tercile probabilities",
        "unit": "RPS"
      },
      "amounts": {
        "score": "crps",
        "selection": "SEL-A",
        "label": "Rainfall amounts",
        "unit": "CRPS (mm)"
      }
    },
    "historical": "A method is a candidate only if the confirmation (CHIRPS, outer leave-one-year-out over the training years, full pipeline) finds it better than the operational method with Holm-adjusted p < 0.05 across every declared method and target, and not worse over the test years. Variants outside the declared family are not candidates.",
    "choice": "Among candidates, the method the selection-aware procedure chooses from all training years (training-only information); if that method is not a candidate, the candidate with the best training-period score. Other candidates are listed as alternatives: a confirmation establishes improvement against the operational method, not a ranking among candidates.",
    "validation": {
      "minimum_seasons": 3,
      "region": "ethiopia",
      "rule": "Independent validation seasons are forecast years scored after the benchmark (bias_validation.py). A candidate is 'recommended' once at least minimum_seasons have been scored for it and the operational method, and its mean score over them is not worse; with fewer seasons it stays a 'candidate'; with enough seasons and a worse mean it is 'not supported by validation'."
    },
    "statuses": {
      "keep": "Keep the operational method: no method is confirmed better on the historical evidence.",
      "candidate": "Confirmed better on the historical evidence; independent validation seasons are still being collected.",
      "recommended": "Confirmed better on the historical evidence and not worse over the independent validation seasons.",
      "not_supported": "Confirmed better on the historical evidence but worse over the independent validation seasons."
    }
  },
  "cycles": [
    {
      "benchmark": "config/bias_correction/jan_fmam.json",
      "name": "jan_fmam",
      "reference": "mvs",
      "reference_label": "Mean-Variance Scaling",
      "targets": {
        "Feb": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Feb on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Feb on the historical evidence."
          }
        },
        "Mar": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Mar on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Mar on the historical evidence."
          }
        },
        "Apr": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Apr on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Apr on the historical evidence."
          }
        },
        "May": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for May on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for May on the historical evidence."
          }
        },
        "FMAM": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for FMAM on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for FMAM on the historical evidence."
          }
        }
      },
      "confirmation_commit": "9f397927f5d4ddf2be3a1fa52595f34fc8fcb56a"
    },
    {
      "benchmark": "config/bias_correction/feb_fmam.json",
      "name": "feb_fmam",
      "reference": "mvs",
      "reference_label": "Mean-Variance Scaling",
      "targets": {
        "Feb": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "candidate",
            "method": "sqrt_mvs",
            "method_label": "Square-root-space Mean-Variance Scaling",
            "chosen_by": "selection-aware procedure (training years only)",
            "alternatives": [],
            "historical": {
              "training_difference": -0.03669166891513275,
              "holm_p": 0.0264,
              "years_better": 18,
              "years": 24,
              "test_difference": -0.006634972406173217,
              "test_years_better": 6,
              "test_years": 9
            },
            "validation": {
              "seasons": [
                {
                  "year": 2026,
                  "method": 0.3988891000097336,
                  "reference": 0.40489301924311805
                }
              ],
              "mean_difference": -0.006003919233384447,
              "required": 3
            },
            "why": "Square-root-space Mean-Variance Scaling is confirmed better than Mean-Variance Scaling for Feb (RPS difference -0.03669 over the training years, Holm p 0.0264; -0.006635 over the test years). Independent validation: 1 of 3 season(s), mean difference -0.006004."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Feb on the historical evidence."
          }
        },
        "Mar": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Mar on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Mar on the historical evidence."
          }
        },
        "Apr": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Apr on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Apr on the historical evidence."
          }
        },
        "May": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for May on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for May on the historical evidence."
          }
        },
        "FMAM": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for FMAM on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for FMAM on the historical evidence."
          }
        }
      },
      "confirmation_commit": "9f397927f5d4ddf2be3a1fa52595f34fc8fcb56a"
    },
    {
      "benchmark": "config/bias_correction/may_jjas.json",
      "name": "may_jjas",
      "reference": "mvs",
      "reference_label": "Mean-Variance Scaling",
      "targets": {
        "Jun": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Jun on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Jun on the historical evidence."
          }
        },
        "Jul": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Jul on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "candidate",
            "method": "vs",
            "method_label": "Variance Scaling (unbounded)",
            "chosen_by": "best training score among candidates",
            "alternatives": [],
            "historical": {
              "training_difference": -0.1385532224164859,
              "holm_p": 0.0,
              "years_better": 23,
              "years": 24,
              "test_difference": -0.1399016692539045,
              "test_years_better": 9,
              "test_years": 9
            },
            "validation": {
              "seasons": [
                {
                  "year": 2026,
                  "method": 25.570328752762315,
                  "reference": 25.987114563855812
                }
              ],
              "mean_difference": -0.4167858110934972,
              "required": 3
            },
            "why": "Variance Scaling (unbounded) is confirmed better than Mean-Variance Scaling for Jul (CRPS (mm) difference -0.1386 over the training years, Holm p 0; -0.1399 over the test years). Independent validation: 1 of 3 season(s), mean difference -0.4168."
          }
        },
        "Aug": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Aug on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "candidate",
            "method": "vs",
            "method_label": "Variance Scaling (unbounded)",
            "chosen_by": "selection-aware procedure (training years only)",
            "alternatives": [
              {
                "method": "sqrt_mvs",
                "label": "Square-root-space Mean-Variance Scaling",
                "holm_p": 0.0
              },
              {
                "method": "pt",
                "label": "Power Transformation",
                "holm_p": 0.00168
              },
              {
                "method": "gqm",
                "label": "Gamma Quantile Mapping",
                "holm_p": 0.00028500000000000004
              }
            ],
            "historical": {
              "training_difference": -0.12898291857397526,
              "holm_p": 0.0,
              "years_better": 24,
              "years": 24,
              "test_difference": -0.08486901377524386,
              "test_years_better": 8,
              "test_years": 9
            },
            "validation": {
              "seasons": [
                {
                  "year": 2026,
                  "method": 15.270217703931799,
                  "reference": 15.456123493019991
                }
              ],
              "mean_difference": -0.1859057890881921,
              "required": 3
            },
            "why": "Variance Scaling (unbounded) is confirmed better than Mean-Variance Scaling for Aug (CRPS (mm) difference -0.129 over the training years, Holm p 0; -0.08487 over the test years). Independent validation: 1 of 3 season(s), mean difference -0.1859."
          }
        },
        "Sep": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Sep on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Sep on the historical evidence."
          }
        },
        "JJAS": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for JJAS on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for JJAS on the historical evidence."
          }
        }
      },
      "confirmation_commit": "9f397927f5d4ddf2be3a1fa52595f34fc8fcb56a"
    },
    {
      "benchmark": "config/bias_correction/sep_ondj.json",
      "name": "sep_ondj",
      "reference": "mvs",
      "reference_label": "Mean-Variance Scaling",
      "targets": {
        "Oct": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Oct on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Oct on the historical evidence."
          }
        },
        "Nov": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Nov on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Nov on the historical evidence."
          }
        },
        "Dec": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Dec on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Dec on the historical evidence."
          }
        },
        "Jan": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for Jan on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "candidate",
            "method": "sqrt_mvs",
            "method_label": "Square-root-space Mean-Variance Scaling",
            "chosen_by": "selection-aware procedure (training years only)",
            "alternatives": [
              {
                "method": "gqm",
                "label": "Gamma Quantile Mapping",
                "holm_p": 0.0084
              }
            ],
            "historical": {
              "training_difference": -0.4786897475521504,
              "holm_p": 0.020945000000000002,
              "years_better": 18,
              "years": 24,
              "test_difference": -0.6902295325749415,
              "test_years_better": 6,
              "test_years": 8
            },
            "validation": {
              "seasons": [],
              "mean_difference": null,
              "required": 3
            },
            "why": "Square-root-space Mean-Variance Scaling is confirmed better than Mean-Variance Scaling for Jan (CRPS (mm) difference -0.4787 over the training years, Holm p 0.0209; -0.6902 over the test years). Independent validation: 0 of 3 season(s)."
          }
        },
        "ONDJ": {
          "probabilities": {
            "product": "Published tercile probabilities",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for ONDJ on the historical evidence."
          },
          "amounts": {
            "product": "Rainfall amounts",
            "status": "keep",
            "method": "mvs",
            "method_label": "Mean-Variance Scaling",
            "alternatives": [],
            "why": "No method is confirmed better than Mean-Variance Scaling for ONDJ on the historical evidence."
          }
        }
      },
      "confirmation_commit": "bed63b906c77cfb3450514d7034ec5f68d067bc0"
    }
  ]
}