Research & Methodology

{
  "price_source": "Yahoo Finance BTC-USD daily Close (auto-adjusted)",
  "phase_source": "PyEphem next_full_moon / next_new_moon, reduced to calendar dates",
  "pivot_detector": "scipy.signal.find_peaks",
  "min_pivot_spacing_days": 30,
  "prominence_pct_of_median_close": 15.0,
  "max_lag_days": 14,
  "signed_lag": "pivot_date - moon_date (positive = pivot AFTER the moon)",
  "full_moon_matches": "nearest significant swing HIGH within +/-14 days",
  "new_moon_matches": "nearest significant swing LOW within +/-14 days",
  "published_benchmarks": {
    "recent_two_year_full_moon": {
      "mean_days": 4.4,
      "median_days": 4,
      "sd_days": 4.2,
      "n": 5,
      "window_from": "2024-09-21"
    },
    "2017_2026_full_moon": {
      "pct_high_after_full_moon": 71.0,
      "mean_days": 3.2,
      "median_days": 6.0,
      "n": 31
    }
  },
  "caveats": [
    "Phase-shift and null controls weaken any claim that the astronomical Full Moon uniquely causes the high.",
    "The New-Moon / local-low correspondence is the stronger empirical lead.",
    "Nodal strength is not a reliable timing predictor."
  ]
}

{
  "new_moon_utc": "2026-09-11T03:27:00",
  "full_moon_utc": "2026-09-26T16:49:00",
  "primary_test": "Does a strict local low form in the window T0:T+3 (11-14 September 2026 inclusive)?",
  "strict_local_low_definition": "A strict 7-day pivot on the daily LOW: a day t qualifies when LOW[t] < LOW[t-1], LOW[t-2], LOW[t-3] AND LOW[t] < LOW[t+1], LOW[t+2], LOW[t+3]. The test is on INTRADAY LOWS, not closing prices.",
  "price_basis": "Daily OHLC. The LOW column only. Closing prices are NOT used.",
  "equality_handling": "Strictly less-than on both sides, as written. A low that merely ties a neighbouring low does not qualify. Applied identically on both sides, so a flat double bottom yields no pivot rather than two.",
  "decidability": "A day cannot be judged until 3 subsequent daily bars exist. Until then the result is UNDETERMINED - never 'failed'.",
  "invalidation": "A qualifying pivot is NOT invalidated by a lower low occurring later, inside or outside the window. The test asks whether a strict local low FORMED in T0:T+3, not whether it was the cycle's lowest price. The deepest low of the cycle is a separate, separately reported measure.",
  "new_moon_strength_recorded_in_advance": 79,
  "strength_caveat": "The 79/100 figure was recorded in advance as context only. It is NOT evidence that a low must occur, and it does not enter the Pattern Fit Score. The strength formula implemented in btcmoon.lunar.phases scores this same New Moon 69.2/100; the two numbers are kept separate rather than reconciled, because tuning a formula to reproduce a historical figure would breach the research constitution.",
  "if_low_forms": "Measure maximum upside at exactly 7, 14 and 21 days from the pivot date.",
  "secondary_tests": [
    "Waxing return from the New-Moon low to the Full Moon (26 Sep).",
    "Cycle-high alignment relative to the Full Moon.",
    "Relationship to a website-defined major high (see website-methodology-v1)."
  ],
  "distinct_concepts": {
    "qualifying_strict_local_low": "The formal pre-registered test. LOW-based 7-day pivot in T0:T+3.",
    "lowest_price_in_a_wider_window": "Descriptive only. Not the test.",
    "absolute_cycle_low": "The deepest low between consecutive New Moons. Reported separately.",
    "informal_nm_plus_0_4": "A private, never-pre-registered intraday timing idea. Requires hourly data. Must never stand in for, or be conflated with, the formal test.",
    "legacy_website_new_moon_pivot": "CLOSE-based scipy.signal.find_peaks major pivot, spacing 30, prominence 15% of median close, matched within +/-14 days. A different measure entirely - see website-methodology-v1."
  },
  "final_review": "After 30 September 2026.",
  "amendment_policy": "None. Lag windows, pivot definitions, scoring weights and the strength formula are fixed for the duration of this test."
}