{
  "key": "fresh4-2",
  "idea": "Skew-aware certainty-equivalent price for the existing uncertainty-table lookup, on top of the thr=0.9 lattice base. Shift only the quote lookup using the posterior third central moment; keep the belief and call rule unchanged. Frozen alpha=-0.025.",
  "why": "Mean and variance discard posterior asymmetry created by the ambiguous arb point mass and retail interval. A third-moment correction could account for unequal economic costs without a learned classifier, averaging quotes across hypotheses, or changing the posterior. Both shift directions were tested. The hypothesis did not produce a measurable gain over the threshold base.",
  "settings_tried": [
    {
      "setting": "alpha=0.1",
      "n": 256,
      "mean_diff": -0.2945,
      "se": 0.0316
    },
    {
      "setting": "alpha=0.25",
      "n": 256,
      "mean_diff": -1.1402,
      "se": 0.0542
    },
    {
      "setting": "alpha=0.5",
      "n": 256,
      "mean_diff": -2.3361,
      "se": 0.0738
    },
    {
      "setting": "alpha=-0.1",
      "n": 256,
      "mean_diff": -0.102,
      "se": 0.0251
    },
    {
      "setting": "alpha=-0.25",
      "n": 256,
      "mean_diff": -0.5514,
      "se": 0.0349
    },
    {
      "setting": "alpha=-0.5",
      "n": 256,
      "mean_diff": -1.2721,
      "se": 0.045
    },
    {
      "setting": "alpha=-0.025",
      "n": 256,
      "mean_diff": -0.0245,
      "se": 0.0166
    },
    {
      "setting": "alpha=0.025",
      "n": 256,
      "mean_diff": -0.0446,
      "se": 0.0175
    }
  ],
  "confirm_paired": {
    "n": 256,
    "mean_diff": 0.005,
    "se": 0.016
  },
  "holdout_vs_lattice_qm": {
    "n": 2048,
    "mean_diff": 0.3175,
    "se": 0.0111,
    "worst5": [
      [
        10942,
        -2.506
      ],
      [
        11487,
        -1.606
      ],
      [
        11645,
        -1.302
      ],
      [
        10885,
        -1.249
      ],
      [
        10955,
        -1.207
      ]
    ],
    "sum": 650.1810833243703,
    "sumsq": 722.3843436286687
  },
  "holdout_vs_thr09": {
    "n": 2048,
    "mean_diff": -0.0054,
    "se": 0.0056,
    "worst5": [
      [
        10387,
        -1.743
      ],
      [
        10822,
        -0.901
      ],
      [
        10942,
        -0.898
      ],
      [
        10087,
        -0.788
      ],
      [
        11009,
        -0.773
      ]
    ],
    "sum": -11.101568447496675,
    "sumsq": 132.5958152076539
  },
  "misread_rate": {
    "base": 0.28883400504733875,
    "mine": 0.28898325491730265
  },
  "price_error_bps": {
    "base": 13.277152756808208,
    "mine": 13.237383092084476
  },
  "gas_estimate": "Unmeasured; one additional pass over at most 121 lattice cells per callback, plus scalar division and clipping. No Solidity implementation or gas bound is claimed.",
  "code_zb64": "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",
  "code_sha256": "a6f15134764be0e11502ae89613033d1752dc8db7b143d42ef5a72c9dcbf2ea4",
  "notes": "Negative result: retain thr=0.9 without this correction. All eight nonzero settings use 256 tuning seeds 60000..60255 against the threshold base; alpha=-0.025 was the best tested nonzero setting. Confirmation used 60256..60511 once; no retuning followed. Each holdout arm covered exactly 10000..12047 in four 512-seed runs through run.py Blind. No seed was filled. Scored callbacks used a numba translation; the delivered plain Python policy matched every callback fee and state diagnostic on 20 independent verification runs 60512..60531, and alpha=0 matched the supplied threshold policy on eight tuning runs. Misread rates and RMS log-price errors aggregate all 2048 holdout runs; base means thr=0.9. Reported price error uses the posterior mean. Paired mean, se and worst5 follow pair.py rounding; sum and sumsq use unrounded per-seed differences. code_sha256 hashes the uncompressed tar. No engine evaluation or measured gas."
}