{
  "generated_at_utc": "2026-07-16T00:00:00.000Z",
  "kill_criterion_detail": "mean_diff=-0.044859, se=0.030437, threshold=2.0×SE=0.060873, tripped=False",
  "kill_criterion_tripped": false,
  "meta": {
    "checkpoint_crosscheck": {
      "checkpoint_evaluated_at_utc": "2026-07-07T22:58:53Z",
      "checkpoint_source": "postgres:match_outcomes",
      "matches_r16_checkpoint": true,
      "mean_log_loss_m0": 0.944197,
      "mean_log_loss_mstar": 0.899338
    },
    "constants_sha_note": "The report embeds the live pre_reg_constants.yaml sha. The committed evaluation/constants.sha is the Phase 8 seal-time hash and no longer matches the YAML after post-seal additive sections were appended (commit 39899d24). expected_constants_sha is intentionally not enforced here; the situation is documented in the deviation note rather than silently reconciled.",
    "gaps_note": "This report scores the 72 pre-registered group-stage forecasts from the frozen champion batch, re-scored against the realized group results with the same per-match regulation-outcome construction the R16 kill-criterion checkpoint uses. It populates M0 and the champion M2 (= M_STAR) accuracy with 95 percent seeded bootstrap confidence intervals and the HLN-corrected Diebold-Mariano comparison of M2 and M_STAR against M0. Every other cell is null, and null means null: M1 and M3 have no committed per-match forecasts, so their accuracy is null; there are no committed market lines, so the de-vigged market row, the DM-vs-market column, and the Nyberg market-efficiency test are all null; every trading and CLV column is null because no bet ledger was committed; and the knockout stages are structurally absent because the frozen batch carries no per-match knockout forecasts (knockout slots hold different teams on every Monte Carlo run), so by_stage carries only the single group row (n = 72). No shadow model, market estimate, or knockout score was invented to fill a column.",
    "n_scored": 72,
    "null_cells": [
      "M1.accuracy",
      "M3.accuracy",
      "MARKET_DEVIGGED.accuracy",
      "*.dm_vs_market",
      "*.nyberg",
      "*.trading",
      "knockout_stages"
    ],
    "populated_cells": [
      "M0.accuracy",
      "M2.accuracy",
      "M_STAR.accuracy",
      "M2.dm_vs_m0",
      "M_STAR.dm_vs_m0"
    ],
    "report_kind": "r16_ablation_templated",
    "source_batch_id": "batch_20260512_013228Z",
    "timing_disclosure": "The kill-criterion checkpoint was published within the 72-hour window on 2026-07-07; this full templated report reuses that same frozen re-scoring, adds no new forecasts, and was compiled after that window, dated accordingly."
  },
  "models": {
    "M0": {
      "accuracy": {
        "brier_ci": [
          0.4511,
          0.6591
        ],
        "brier_mean": 0.5543,
        "log_loss_ci": [
          0.7957,
          1.0968
        ],
        "log_loss_mean": 0.9442,
        "n_forecasts": 72,
        "rps_ci": [
          0.1406,
          0.2024
        ],
        "rps_mean": 0.1703,
        "rps_median": 0.1501
      },
      "by_stage": {
        "group": {
          "brier_mean": 0.5543,
          "log_loss_mean": 0.9442,
          "n": 72,
          "rps_mean": 0.1703
        }
      },
      "dm_vs_m0": null,
      "dm_vs_market": null,
      "nyberg": null,
      "trading": null
    },
    "M1": {
      "accuracy": null,
      "by_stage": {},
      "dm_vs_m0": null,
      "dm_vs_market": null,
      "nyberg": null,
      "trading": null
    },
    "M2": {
      "accuracy": {
        "brier_ci": [
          0.4271,
          0.6319
        ],
        "brier_mean": 0.5274,
        "log_loss_ci": [
          0.7567,
          1.0475
        ],
        "log_loss_mean": 0.8993,
        "n_forecasts": 72,
        "rps_ci": [
          0.1313,
          0.1902
        ],
        "rps_mean": 0.1596,
        "rps_median": 0.1494
      },
      "by_stage": {
        "group": {
          "brier_mean": 0.5274,
          "log_loss_mean": 0.8993,
          "n": 72,
          "rps_mean": 0.1596
        }
      },
      "dm_vs_m0": {
        "hln_applied": true,
        "pvalue": 0.2719,
        "reject": false,
        "stat": -1.1073
      },
      "dm_vs_market": null,
      "nyberg": null,
      "trading": null
    },
    "M3": {
      "accuracy": null,
      "by_stage": {},
      "dm_vs_m0": null,
      "dm_vs_market": null,
      "nyberg": null,
      "trading": null
    },
    "MARKET_DEVIGGED": {
      "accuracy": null,
      "by_stage": {},
      "dm_vs_m0": null,
      "dm_vs_market": null,
      "nyberg": null,
      "trading": null
    },
    "M_STAR": {
      "accuracy": {
        "brier_ci": [
          0.4271,
          0.6319
        ],
        "brier_mean": 0.5274,
        "log_loss_ci": [
          0.7567,
          1.0475
        ],
        "log_loss_mean": 0.8993,
        "n_forecasts": 72,
        "rps_ci": [
          0.1313,
          0.1902
        ],
        "rps_mean": 0.1596,
        "rps_median": 0.1494
      },
      "by_stage": {
        "group": {
          "brier_mean": 0.5274,
          "log_loss_mean": 0.8993,
          "n": 72,
          "rps_mean": 0.1596
        }
      },
      "dm_vs_m0": {
        "hln_applied": true,
        "pvalue": 0.2719,
        "reject": false,
        "stat": -1.1073
      },
      "dm_vs_market": null,
      "nyberg": null,
      "trading": null
    }
  },
  "pre_reg_constants_sha": "691993cce8194898e7eef164526a7e81cec94edff2823aa37f3882fedc09c227",
  "schema_version": "7.0",
  "tournament": "FIFA2026"
}