{
  "date": "2026-10-24",
  "status": "Written before this broader study's results; earlier studies have already examined the same historical sample, so no period is an untouched holdout.",
  "purpose": "Compare distinct economically motivated, low-frequency enhancement families, not keep tuning the user's example. Permit modest tactical increases funded from bonds, and stock rotation, in addition to defensive overlays. Assess return/drawdown trade-offs, not only strict dominance.",
  "data": "Reuse the frozen 2016-10-21..2026-10-21 RMB dividend-reinvested ETF market-price and NAV data. Pre-2019-01-18 dividend-low-vol uses official H20269 total return less assumed 0.6% annual fee. Warm-up starts in 2014. No original archived files change. Daily valuation dates are unioned; all execution dates must be actual common mainland ETF trading days.",
  "monthly_execution": "Signals use the last completed calendar month's actual observations. First common trading day's CLOSE of the next month executes the decision. No signal-day fills. Check at most monthly. Trade when target states change, or at the first monthly opportunity of a new quarter. Quarterly unchanged-state rebalance is the standard benchmark. Extra 7 calendar-day delay and NAV signals with 7-day publication allowance are sensitivities. Fractional shares; initial and buy/sell notional cost 0.1%, with 0.3% and 0.5% stress tests.",
  "asset_order": ["nasdaq_etf", "sp500_etf", "divlow", "gold", "gov5"],
  "bases": {"user40": [0.4,0,0.2,0.1,0.3], "growth50": [0.5,0,0.2,0.1,0.2]},
  "families": {
    "ma": "Completed month-end total-return price versus 6/10/12-month SMA, exposure 0/100 or 50/100 percent of normal allocation. No fitting.",
    "momentum": "6/12-month total return versus actual five-year-bond return, exposure 0/100 or 50/100. Bond return is not a risk-free rate; this is an adaptation.",
    "ensemble": "Equal voting across MA 3/6/10/12 or momentum 1/3/12 and 3/6/12 months. Average binary signals, floors 0 or 50 percent. Explicit ensemble of MA10, momentum12, and volatility16 with 63 observations, also with 50 percent permanent core.",
    "volatility": "Target 12/16/20 percent annual standalone asset volatility divided by prior 63/126 actual daily-return standard deviation times sqrt252; cap 1, floor .25, round exposure to nearest .25. Inverse-variance adaptation uses past 252-observation volatility squared divided by 21/63/126-day volatility squared, same caps. No full-sample scale fitting or leverage.",
    "conditional_vol": "If trailing 21/63-day volatility exceeds the 80th percentile of the PREVIOUS 36 completed monthly observations (minimum12), retain half exposure; otherwise full. Downside-only adaptation of conditional-volatility literature, not a paper replication.",
    "trend_vol": "MA10 floor50 or MA3/6/10/12 voting floor50, multiplied by capped .16/63-day realized volatility, rounded to .25 with floor.25.",
    "tilt": "MA10 trend: .5/1.25 normal exposure. MA3/6/10/12 vote: .5+.75*vote, rounded .25. Additions funded from bonds; no leverage.",
    "contrarian": "Price/MA10 deviation beyond +/-10 or20 percent: factors .75/1.25, otherwise1. Weekly Wilder RSI14 observed at month end: >70=>.75,<30=>1.25, otherwise1. Trend-filtered RSI: belowMA10=>.5; aboveMA10 and RSI<40=>1.25; aboveMA10 and RSI>70=>.75; otherwise1. These are exploratory conventional technical heuristics, with weaker evidence for these exact ETF rules.",
    "rotation": "Within original stock budget, rotate across Nasdaq100, S&P500 and dividend-low-vol using 12-month or mean3/6/12-month returns. Allocate to top1/top2, only if each beats five-year bonds; otherwise that slice goes to bonds. Permanent stock core 0 or50 percent. This is asset selection, not pure Nasdaq/dividend timing, and is not Antonacci GEM replication.",
    "rebalance": "Quarterly fixed baseline plus monthly, annual, and monthly-inspected absolute drift bands5/10 percentage points; a rule without return forecasts. Band variants rebalance only on band breach."
  },
  "limits": "Portfolio targets: each stock <=60%, combined stocks <=80%, bonds >=10%, gold stays10%. Proportional scaling if combined cap binds. Between trades weights can drift. Standalone sleeves cap stock at100%; tilt rules are clipped, explicitly labelled as such. Every rule uses identical parameters on Nasdaq, dividend-low-vol and S&P500. Portfolio overlays applied to Nasdaq only, dividend only and both. No hard future maximum drawdown guarantee.",
  "selection": "Report ALL rules and failures. Primary evaluation uses NAV, with market-price robustness. Anchor rules selected by published prevalence, not full-sample rank. A historical choice trained through2020 selects highest minimum(NAV,market) CAGR among portfolios with worst(NAV,market) drawdown<=15%, <=12 operations per complete year; if none qualify choose lowest worst drawdown. Annual expanding walk-forward repeats selection using past data only, starting2021. Same examined historical sample; retrospective validation only. Separate labelled full-sample candidates show risk/return frontier.",
  "validation": "Chronological periods full/train/post2020/2021-2023/2024-onward and actual-ETF2019+. Stress 2020 shock,2022 decline. For shortlisted methods: costs, 7-day execution delay, delayed NAV signals, reset2019, adjacent parameter choices. Compare static portfolios matched to average actual exposure as diagnostic hindsight controls, not tradable ex-ante rules. Include paired 6-month moving-block bootstrap of monthly log-return differences to show uncertainty; confidence intervals are not multiple-testing corrected and are not proof of alpha.",
  "excluded": "No macro/valuation/breadth rules without reliably timestamped point-in-time data; no fitting future macro regimes, machine learning search, leverage, short selling, derivatives, tax assumptions, premium arbitrage with unpublished NAV, or claim that costs stay below real QDII premium effects. No intraday fills, actual dividend payment-delay cashflow, minimum commission or whole-lot constraints."
}