THE SIGNAL日本語
Menu

ETF MODEL / HOW IT WORKS

How the model sets direction and target weights

Current model v4.0.1. The 20 published instruments and calculation settings below are generated from the implementation and configuration.

  1. 1. Regular session closes

    Collection starts 30 minutes after the calendar-defined close. U.S. and Japanese markets update independently.

  2. 2. Validate daily data

    Check the target session, OHLC, splits, missing observations and reference prices. A valid 220-trading-day history is required.

  3. 3. Determine direction

    Apply the selected instrument-specific timing rule below.

  4. 4. Calculate instrument weights

    Use the actual ETF's volatility and its instrument-level cap.

  5. 5. Apply account constraints

    Scale down sequentially for target-group totals, invested capital, leverage-weighted exposure and covariance-based risk.

  6. 6. Publish / next opening

    Publish validated results. Assumed execution is at the next market trading day's opening.

Instrument-specific models

Changed November 5, 2026. Selection prioritizes net completed-trade win rate in the first 60% of observed history, with at least 10 trades. Ties use CAGR. The later period is retrospective, not untouched validation. Higher win rate does not imply higher returns. Insufficient evidence retains the previous rule. SPXL uses the 200-day trend-following rule by explicit operator selection.

SPXLSPY: Reference close relative to SMA200: above => LONG; otherwise CASH. Decision: LONG.
SPXSSPY: Reference close relative to SMA200: below and 20-session return is negative => LONG; otherwise CASH. Decision: CASH.
TQQQQQQ: Contrarian: Reference close relative to SMA200: below => LONG; otherwise CASH. Decision: CASH.
SQQQQQQ: Reference close relative to SMA200: below and 20-session return is negative => LONG; otherwise CASH. Decision: CASH.
TECLXLK: Contrarian: Reference close relative to SMA200: below => LONG; otherwise CASH. Decision: CASH.
SOXLSOXX: Contrarian: Reference close relative to SMA200: below => LONG; otherwise CASH. Decision: LONG.
FASXLF: Reference close relative to SMA200: above and 20-session return is positive => LONG; otherwise CASH. Decision: LONG.
TECSXLK: Reference close relative to SMA200: below and 20-session return is negative => LONG; otherwise CASH. Decision: CASH.
SOXSSOXX: Reference close relative to SMA200: below and 20-session return is negative => LONG; otherwise CASH. Decision: LONG.
FAZXLF: Reference close relative to SMA200: below => LONG; otherwise CASH. Decision: CASH.
NVDLNVDA: Reference close relative to SMA200: above and 20-session return is positive => LONG; otherwise CASH. Decision: CASH.
MUUMU: Reference close relative to SMA200: above => LONG; otherwise CASH. Decision: LONG.
TSLLTSLA: Reference close relative to SMA100: above => LONG; otherwise CASH. Decision: LONG.
SPCHSPCX: Reference close relative to SMA200: above => LONG; otherwise CASH. Decision: UNAVAILABLE.
NVDDNVDA: Reference close relative to SMA200: below => LONG; otherwise CASH. Decision: LONG.
MUDMU: Reference close relative to SMA200: below and 20-session return is negative => LONG; otherwise CASH. Decision: CASH.
TSLSTSLA: Reference close relative to SMA100: below => LONG; otherwise CASH. Decision: CASH.
SSPCSPCX: Reference close relative to SMA200: below => LONG; otherwise CASH. Decision: UNAVAILABLE.
15701321.T: Contrarian: Reference close relative to SMA200: below => LONG; otherwise CASH. Decision: LONG.
13571321.T: Contrarian: Reference close relative to SMA200: above => LONG; otherwise CASH. Decision: CASH.

Published instruments and reference prices

SymbolTarget / daily leverageReference priceInstrument cap
SPXLS&P500 · Bull 3×SPY30%
SPXSS&P500 · Bear 3×SPY30%
TQQQNasdaq-100 · Bull 3×QQQ30%
SQQQNasdaq-100 · Bear 3×QQQ30%
TECLInformation technology · Bull 3×XLK30%
SOXLSemiconductors · Bull 3×SOXX30%
FASFinancials · Bull 3×XLF30%
TECSInformation technology · Bear 3×XLK30%
SOXSSemiconductors · Bear 3×SOXX30%
FAZFinancials · Bear 3×XLF30%
NVDLNVIDIA · Bull 2×NVDA5%
MUUMicron · Bull 2×MU5%
TSLLTesla · Bull 2×TSLA5%
SPCHSpaceX · Bull 2×SPCX5%
NVDDNVIDIA · Bear 1×NVDA5%
MUDMicron · Bear 1×MU5%
TSLSTesla · Bear 1×TSLA5%
SSPCSpaceX · Bear 2×SPCX5%
1570Nikkei 225 · Bull 2×1321.T30%
1357Nikkei 225 · Bear 2×1321.T30%

Fixed selections: S&P 500 uses SPXL; Nasdaq-100 uses TQQQ. Product direction and the model's LONG/CASH state are separate.

Target-weight calculation

When LONG: min(instrument cap, 100%, 25% / max(actual ETF annualized volatility, 5%))
When CASH: 0%

Actual ETF daily returns include splits and distributions over the latest 60 trading days, with at least 40 observations. Annualized volatility is sample standard deviation × √252. Product leverage is not multiplied into ETF volatility again.

Instrument capsStandard: 30% / Single-stock: 5%
Aggregate target capsS&P 500: 30% / Nasdaq-100: 30% / All single-stock products: 10% / Overlapping technology exposure: 50%
Scale within market accountsReduce weights only as needed to satisfy constraints. Residual funds remain cash; weights are not renormalized to 100%.
AccountInvestment capLeverage-weighted exposure capAnnual volatility target
US (USD)80%150%25%
JP (JPY)80%100%25%

Weights use each market account's total assets as the denominator. USD and JPY have independent budgets; their weights are not added together. Risk targets do not guarantee a maximum loss.

Earnings-event guard: Disabled, reflecting the current configuration.

Update and execution timeline

Target market closeThe reference date is the market's local session date: ET for the U.S. and JST for Japan.
Close + 30 minutesStart collection and validation. If unavailable, retry every 5 minutes until 120 minutes after close. Time alone does not establish final data.
Next trading day's openingMaintain the next-opening execution assumption. Results past their order deadline are not activated or extended.

Exchange calendars determine holidays, shortened sessions and daylight saving time. On retrieval failure, retain the last valid result with its actual reference date. Never relabel old daily prices as today's data.

Relationship to validation

Published models use instrument-specific conditions shared by daily signals and backtests. Historical performance does not establish future profitability.

Backtest results and assumptions → / Latest signals and prices →

Configuration / implementation fingerprint: 85e3ae8ea53b3032. This explanation is regenerated from the actual settings on every production build.