run/summary.json

{
  "calibration_n_claim1": 128,
  "claim1": {
    "minimum_exact_expected_margin": 0.00011968992248062022,
    "minimum_expected_margin": 0.0005542894056847038,
    "minimum_margin_ci95_low": -0.0015807128982507795,
    "nonexchangeable_negative_margin_count": 70,
    "theory_identity_max_abs_error": 0.002303772609819138,
    "verdict": "SUPPORTED"
  },
  "claim2": {
    "empirical_loglog_slope": -0.49719807619023837,
    "minimum_eq7_coverage": 0.995,
    "understated_bound_negative_margin_count": 7,
    "verdict": "SUPPORTED"
  },
  "claim3": {
    "all_four_settings_monotone": true,
    "destructive_reversal_changes_direction": true,
    "verdict": "SUPPORTED"
  },
  "claim4": {
    "scope": "Controlled domain analogues; not the paper's exact absent artifact.",
    "settings_executed": [
      "linear_program",
      "newsvendor",
      "portfolio",
      "shortest_path"
    ],
    "verdict": "PARTIAL"
  },
  "claim5": {
    "scope": "The controlled comparison demonstrates the validity/accuracy trade-off but is not a reconstruction of Figures 3\u20134.",
    "verdict": "NOT_INDEPENDENTLY_REPRODUCED"
  },
  "destructive_controls": {
    "nonexchangeable_sampling": "breaks expected validity in reported rows",
    "reversed_policy": "reverses the intended miscoverage direction",
    "understated_loss_bound": "produces negative expected margins"
  },
  "elapsed_seconds": 2.5065145499993378,
  "population": 100000,
  "repetitions": 1200,
  "seed": 2026
}