The informal NM+0.4 timing idea - assessed separately, and not a formal test

This record exists to keep the informal idea separate from the pre-registered
protocol. They are different claims, tested against different data, and one must
never be allowed to stand in for the other.

What the informal idea is. A private, working short-horizon notion that the
local reaction low tends to arrive around NM+0.4 days - roughly ten hours
after the exact New Moon. It was never pre-registered, has no written pass/fail
criteria, and is not part of the paper.

Why it cannot be tested on daily data. NM+0.4 days after the September New
Moon (11 Sep 03:26:55 UTC) is 11 Sep ~13:03 UTC. A daily bar cannot resolve a
ten-hour offset. Anyone claiming to have confirmed or refuted NM+0.4 from daily
candles has not tested it.

What hourly data shows. Assessed on hourly BTC-USD bars, purely as an
observation:

• The lowest hourly low within +/-12 hours of the NM+0.4 target printed at
  11 Sep 12:00 UTC, i.e. NM+0.36 days, at $76,088.52.
• That is roughly an hour earlier than the NM+0.4 estimate - which matches
  what the private note recorded at the time.
• For August (New Moon 12 Aug 17:36:39 UTC), the nearest hourly low to the
  NM+0.4 target came at NM+0.06 days, while the deepest hourly low of that
  reaction was at NM+1.85 days.

What this does and does not mean.

• It does not validate NM+0.4. Two observations, chosen after the fact, with
  no pre-registered window, no null control and no correction for the fact that
  some hourly low always exists near any chosen timestamp. This is an
  anecdote, and it is recorded as one.
• It does not feed the formal September result. That test stands on its own
  pre-registered LOW-based rule.
• It is interesting enough to deserve a proper pre-registration: fix a
  window, fix a definition of "reaction low", fix the null control, and then
  test it forward. Until that happens it remains a private hypothesis.

Note on data. Hourly and daily feeds disagree slightly on the exact low
(hourly $76,088.52 vs daily $76,162.91 for 11 September). That is normal
aggregation variance between vendor series. The formal test uses the daily OHLC
series and is unaffected; it is flagged here so the discrepancy is on the record
rather than discovered later.