Residuals beat visual rhythm.
Month-day labels use a non-leap reference calendar only to measure intervals. They do not assign a source year.
LOAD THE STATED MAXIMA TO BEGIN
SCAN RESULTBEST SCAN: --
CANDIDATE A10.5-DAY RMSE: --
CANDIDATE B14-DAY RMSE: --
INTERVALS: --
14-day residuals
What the dates support
They can compare in-sample spacing once analyzed.
What remains untested
No raw polar-motion signal, uncertainty, out-of-sample event, solar current, earthquake catalog, lunar torque, charge layer, ECDO, LLSVP, or deep-Earth mechanism is analyzed here.
SPACING IS NOT CAUSATIONRESIDUALS BEFORE MECHANISMSTHREE EVENTS REQUIRE OUT-OF-SAMPLE TESTINGSPACING IS NOT CAUSATIONRESIDUALS BEFORE MECHANISMSTHREE EVENTS REQUIRE OUT-OF-SAMPLE TESTING
A correction becomes useful when the method travels.
Normalize
Convert ordered dates into elapsed-day offsets without letting an arbitrary source year alter the gaps.
Predict
For each candidate period, predict event n at n times the period from the first observed maximum.
Measure
Residual RMSE summarizes how far the observed timing sits from each candidate cadence.
Challenge
Add future events and uncertainty. A three-point in-sample fit has no independent forecast test.
Worked fit
At 10.5 days, predicted offsets are 0, 10.5, 21 against observed 0, 11, 22. Residuals are 0, 0.5, 1 day.
Counterexample
At 14 days, predicted offsets are 0, 14, 28. Residuals are 0, -3, -6 days, creating far more timing error.
Transfer
Keep the candidate periods frozen, add later maxima with timing uncertainty, and judge whether residual error remains small out of sample.