Combine seasonal, election, and yield rules as one procedure

Write three rules as one schedule

The archive workflow turns monthly equity-index closes into invest-or-cash rules. Seasonal-trading writes calendar months and election-cycle years as testable entry, cash, and abstention rules over the system holding period. System-optimization then combines those rules in steps so the whole procedure can be tested, with every extra cell treated as a fitting risk.

The calendar layer used windows around November to January, March to April, and July, with September treated as the weak window from 1950 to 1995. The four-year election calendar was a second regime: remain invested in preelection and election years and move to cash in postelection and midterm years.

Lag the yield-trend-filter

A third rule compares the 30-year yield with its six-month average and treats a gap above that average of more than 0.1% as the cash or exit state. The yield-trend-filter is that lagged gap between the long Treasury yield and its recent average, used to stay invested, reduce size, or stand aside.

The implementable filter uses the prior month's yield minus the prior month's six-month average, because the current month's close is not known before month-end. A fitted line from last month's yield gap to this month's equity return is only a coarse inverse-correlation check. The intended input is the trend of the yield, not its level.

Combine through a few regime cells

A many-exception rule set is treated as more likely to be fitted to one sample, and harder to follow under stress, than a few exception-free rules. A regime-cell is one bucket formed by crossing a calendar window with an election-cycle pair so exposure can be assigned without a thicket of exceptions.

Crossing a half-year seasonal split with lumped election pairs yields four cells that can be added, multiplied, or weighted. Expanding to 12 monthly slots or 24 month-by-cycle slots raises fitting risk. A coarse-permutation keeps a small starting grid, such as two half-year windows crossed with two lumped election pairs plus a yield state, used to limit fitting. Restricting the design that way leaves four starting permutations and is offered as a check against a fully iterated month-by-year table.

Cap stacked exposure

Stacking seasonal and yield exposure scores without a cap can take leverage above 2:1. Leverage-control is a hard bound on exposure, set from account risk and regime strength before entry and kept while the position is open. A 200% ceiling is applied so the loss or exposure decision is bounded before the combined signal is used.

Check stability for compounding-illusion

Ending equity can look superior only because one or two early outperforming years later compound. Compounding-illusion is that pattern: ending equity that looks superior only because one or two early years kept multiplying, not because later years still beat a fully invested path. Stability is checked with return variability, worst 12-month loss, decade windows, and small weight perturbations.