{
  "stage": "Before evaluating the complete component grid in portfolios, after the standalone component screen",
  "reason": "Standalone winners need not be portfolio winners because of diversification and quarterly rebalancing. Apply the already frozen993 rules to one stock component at a time for all three declared baseline allocations; do not change indicators or parameter thresholds.",
  "universe": "For each of three bases,993 Nasdaq-only and993 dividend-only rules, with their common fixed benchmark deduplicated:5955 unique specifications. Previously computed joint shortlisted combinations remain a separate exploratory sample.",
  "training_selection": "For each base and stock component, maximize2016-10-21..2020-12-31 market CAGR subject to market maximum drawdown<=min(15%,the fixed baseline training maximum drawdown), average operation days<=8/year, and full-calendar-year maximum<=12. If no candidate qualifies, choose the fixed baseline. Signals and parameters then remain fixed in the2021+chronological evaluation. Selection uses market data only, available at the year-end weekend.",
  "exploratory_frontier": "Rank full-period CAGR subject to both NAV and market maximum drawdown<=12/15/20%, and the same operation limits. These rankings are explicitly in-sample exploration, not validation or a future risk cap.",
  "reporting": "Retain all metrics and source specification IDs; display training-selected rules, plain fixed controls and selected full-period frontier examples, with post2020 results and sensitivity checks. Parameter-grid sizes are not counts of independent tests."
}