{
  "any_gap_detectable_today": false,
  "claim_boundary": "Reads trading databases and walk-forward curves read-only. Registers no hypothesis identity and opens no return data. 0 trials.",
  "go_live_date": "2026-08-07",
  "schema": "canli.alphac-execution-gap-power.v1",
  "sleeves": [
    {
      "calendar_days_required_to_detect_1pct": 2446.6434985490123,
      "common_days": 5,
      "daily_tracking_sd": 0.0008318286844294201,
      "days_required_to_detect": {
        "0.5%": 6752.133186960724,
        "1.0%": 1688.033296740181,
        "2.0%": 422.00832418504524,
        "5.0%": 67.52133186960724
      },
      "measurable": false,
      "observed_annualised_gap": -0.11812065834455232,
      "observed_over_stderr": -1.1269935263895376,
      "return_observations": 4,
      "sleeve": "AlphaMax",
      "standard_error_of_that_gap": 0.10481041423810693,
      "walkforward": "equity_live_fwd",
      "window": [
        "2026-08-07",
        "2026-08-14"
      ]
    },
    {
      "calendar_days_required_to_detect_1pct": 3356.0783653198646,
      "common_days": 9,
      "daily_tracking_sd": 0.0009742369397852844,
      "days_required_to_detect": {
        "0.5%": 9261.94932852135,
        "1.0%": 2315.4873321303376,
        "2.0%": 578.8718330325844,
        "5.0%": 92.6194932852135
      },
      "measurable": false,
      "observed_annualised_gap": -0.028409575853535318,
      "observed_over_stderr": -0.32729894849721064,
      "return_observations": 8,
      "sleeve": "AlphaTrend",
      "standard_error_of_that_gap": 0.08680008287218019,
      "walkforward": "mf_live_fwd",
      "window": [
        "2026-08-07",
        "2026-08-21"
      ]
    }
  ],
  "the_answer": "THE RECORD IS TOO SHORT. Overlap between the live marks and a walk-forward that extends into the live period exists for two sleeves and is 5 and 9 days \u2014 four and eight return observations. No execution gap is distinguishable from noise at that length, and the observed divergences below are OBSERVATIONS, not estimates.",
  "the_quantity_that_answers_it_soonest": "The TRACKING DIFFERENCE d = r_live - r_backtest, not either Sharpe. The two run the same signal on the same days, so their returns are strongly correlated and the difference series is far less noisy than either level. The record needed to detect an execution gap is therefore much shorter than the record needed to establish the edge itself \u2014 which is the useful thing to know.",
  "verdict": "NOT MEASURABLE YET \u2014 publish the power, not a number",
  "what_would_change_this": "Time, and nothing else. The days_required figures below are the honest answer to 'when can we ask'. They also depend on the tracking sd being estimated from four and eight observations, which is itself a very noisy estimate \u2014 treat them as an order of magnitude, not a date.",
  "when_to_look_again": {
    "a_1pct_annual_gap": {
      "AlphaMax": "6.7 years",
      "AlphaTrend": "9.2 years"
    },
    "a_5pct_annual_gap": {
      "AlphaMax": "3 months",
      "AlphaTrend": "4 months"
    },
    "note": "The useful output is not a number today but a date. Converted from return observations at 252 trading days a year, and rounded outward because the tracking sd behind them is itself estimated from four and eight observations.",
    "the_practical_reading": "A LARGE execution gap \u2014 5% of equity a year \u2014 becomes visible in roughly three to four months, so that is when this is worth re-running. A 1% gap needs the better part of a decade and is not answerable by waiting. If a gap that small matters, it has to be attacked by measuring fills directly against their decision prices, not by comparing equity curves."
  },
  "why_only_two_sleeves": "A comparison needs a backtest that covers the live days. The BLESSED research curves (crypto_carry_wk, k30_dn_63) are frozen by disclosure protocol and end 2026-06-01, before go-live, so they cannot answer this question at any record length. Only the live-forward walk-forwards, regenerated by the daily ticks, overlap at all."
}