{
  "created_before_results": true,
  "date": "2026-10-24",
  "objective": "Test simple weekly Nasdaq100 and dividend-low-vol timing overlays. Preserve/raise compound return while reducing drawdown and limiting actual operation days. No leverage, derivatives, shorts or additional securities.",
  "data_freeze": "2016-10-21 through 2026-10-21, same cached adjusted ETF total-return prices and NAVs as prior research; 2014+ history warms signals. 512890 before 2019-01-18 is official H20269 total-return index minus assumed annual 0.6% drag. No original source is changed.",
  "primary_signal": "Completed Friday-ending weekly total-return market close. Split/dividend adjusted. At a weekend, examine the latest completed week; execute at the closing price on the first subsequent common Chinese trading day. Never use the signal week's close as the execution price. In weeks without observed quotes, do not create confirmation observations. A NAV-signal sensitivity uses an additional whole-week lag for publication availability.",
  "execution": "At most one operation day per week. Keep current holdings if target state does not change. At the first weekly operation opportunity of a new calendar quarter, rebalance to the existing state target. On a state change, rebalance the whole portfolio. Between operations holdings drift. Sell-side allocations go to 511010 five-year government bond ETF. A state has a minimum 0 or 4 complete calendar weeks before switching again; confirmation requires 1 or 2 observed weekly closes. Initial purchase and every buy/sell notional cost 0.1%, solved self-consistently after fees.",
  "families": {
    "sma": "10/20/30/40/52 week simple MA, symmetric buffer 0/1/2%; outside lower band cut, outside upper band restore",
    "ema": "EMA fast/slow 4/13,8/26,13/40,20/52, symmetric buffer 0/1%",
    "momentum": "13/26/52 week stock return versus government bond return; symmetric excess-return threshold 0/2%",
    "channel": "Exit below prior 8/13/20/26-week closing low; re-enter above prior 4/8/13-week closing high, entry lookback<=exit lookback; excludes current week in comparison",
    "pullback": "Rolling peak since latest entry, capped at 26/52 weeks; exit on 6/10/15% drawdown; restore after 3/6% bounce from low since exit and positive latest week",
    "hybrid": "20/40-week SMA and 6/10% drawdown below past26-week peak needed together to exit; restore above SMA, or bounce3/6% from low since exit AND close above4-week MA",
    "staged": "13/40 or20/52 week MA; if above both then100%, below both then floor, between then midpoint; profiles100/75/50 or100/50/0, buffers0/1%"
  },
  "common_grid": "Binary-state defensive sleeve exposure0/25/50/75% of its normal stock allocation. Confirmation1/2 observed weeks, minimum hold0/4 calendar weeks. Finite coarse grid published in full; no repeated rescue-tuning after results. A buy-and-hold control is included.",
  "lazy_limits": "Primary candidate filter: average operation days<=8/year and maximum operation days in any full calendar year<=12, excluding initial purchase. Also disclose state-change count and operations per calendar year. These are realized historical counts, not hard future limits.",
  "chronology": "Training2016-10-21..2020-12-31; lock choices then evaluate2020-12-31..2026-10-21. Choose the highest training CAGR among lazy rules with training daily drawdown no worse than the hold control, using only market-price metrics; include hold as an eligible choice. Rank tie by smaller drawdown and fewer trades. Alternative training choice accepts at most2pp annual return sacrifice for >=15% relative drawdown reduction. These years and instruments were already examined, so this is chronological retrospective validation, not an untouched holdout.",
  "portfolio_tests": "Fixed40/20/10/30,50/20/10/20,50/15/15/20 in Nasdaq/divlow/gold/gov5 order. Compare no overlay, Nasdaq only, dividend only, both. Independently selected training rules and prespecified 20w/40w buffered-MA reference rules are tested. Also retain clearly labelled full-sample exploratory winners; do not promote them as validated.",
  "robustness": "Market and NAV valuations use the same market signals and operation calendar. Show allperiod, training, post2020, 2021..2023 and2024..2026, actualETF return period. For shortlisted rules: one-side costs0.3/0.5%, an extra week delay, lagged NAV signals, neighboring parameters, matched average-exposure static controls, cut/re-entry episodes and whipsaw opportunity cost.",
  "whipsaw": "For completed defensive episodes report stock and bond total returns from executed cut close through executed restoration close. Stock minus bond positive means omitted stock outperformed parking in bonds; report duration, raw count, <=8-week whipsaw count, and whether re-entry price exceeds cut. This diagnostic is distinct from actual portfolio P&L and fees.",
  "limits": "Historical closing-price execution, fractional shares, theoretical ex-date dividend reinvestment; not an exact intraday fill or live implementability guarantee. Portfolio targets are not hard daily weight caps. No assurance of future15%/20% maximum drawdown or persistent return enhancement. Larger grids raise selection bias; show failures as well as winners."
}