Strategy Card

Time-Series Momentum w/ Trend Filter + Volatility/Drawdown Crash Protection (US Stocks) — strategy card

Spec ID: spec-tsm-trendfilter-crashprotected-us-stocks-1772186900 · Generated: 2026-05-18 10:18 UTC

One-line description

Long-only weekly time-series momentum strategy on the 50 largest US stocks, selecting names with positive multi-horizon momentum, a 200-day trend filter, and sufficient ADX trend strength. Portfolio risk is moderated with inverse-vol weighting, volatility targeting, ATR trailing stops, SPY regime filters, and drawdown de-leveraging.

Why this trade exists

Time-series momentum and trend-following attempt to capture the empirical tendency for assets with positive recent returns and established uptrends to continue outperforming over intermediate horizons. The strategy combines stock-level continuation signals with a broad-market risk-on filter, reflecting the view that single-name momentum is more reliable when the market proxy is also above its long-term trend.

Crash protection is included because equity momentum can suffer during volatility spikes, sharp reversals, and correlated selloffs. Volatility targeting, inverse-volatility sizing, ATR stops, and portfolio drawdown de-leveraging are intended to reduce exposure when realized risk or losses rise, while still allowing participation during persistent equity trends.

Algorithm

[code omitted from public view]

Parameters

Param	Value	Notes
Universe	Top 50 US stocks by capitalization	Static symbol list supplied in spec; includes AAPL, MSFT, NVDA, AMZN, etc.
Side	Long only	No short positions.
Bar size	1 day	Daily OHLCV inputs.
Backtest window	2020-01-01 to 2025-10-08	Includes COVID crash, 2022 bear market, and 2023-2025 rally period.
Rebalance frequency	Weekly	Uses last trading day of each week.
Entry signal	Majority vote, at least 3 of 4 votes	Votes: close > 200D SMA; 21D return > 0; 63D return > 0; 126D return > 0.
Trend strength filter	ADX(14) >= 18	Filters weaker or choppier trends.
Ranking	126D return, descending	Selects strongest eligible names.
Max positions	15	Concentrated large-cap momentum basket.
Weighting	Inverse volatility	Uses 20D realized volatility.
Per-name cap	10%	Applied during portfolio construction.
Volatility target	12% annualized	Always applied, based on 20D realized vol.
Vol scale bounds	0.25 to 1.5	Limits de-risking and re-risking from vol targeting.
Max leverage	4.0	Portfolio-level leverage ceiling.
ATR stop	3.0 * ATR(14) trailing from highest close since entry	Exit on close breach, executed next open.
Market risk-on filter	SPY close > SPY 200D SMA	Required for new risk-on entries.
Volatility regime filter	SPY RV20 / RV252 <= 1.5	De-risks when short-term market vol is elevated.
Drawdown control	63D drawdown trigger at 12%	Risk scale multiplied by 0.5 for at least 10 trading days.
Drawdown reset	SPY above 200D SMA for 5 consecutive bars	Restores normal risk scale after condition is met.
Execution	Signal at close, market order at next open	Partial fills allowed.
Costs	Not specified / null	Backtest metrics should be interpreted as before explicit commissions, borrow, tax, and slippage assumptions unless the engine applied defaults outside the spec.

Look-ahead audit

#	Concern	Status
1	Indicator timing	Signals are defined using close data and trades occur at the next open, reducing same-bar look-ahead risk.
2	Rolling windows	SMA, returns, ADX, ATR, and realized volatility use historical rolling windows with full-window minimums.
3	Universe construction	Universe is described as top 50 by capitalization; if selected using future information or survivorship-biased constituents, results may be overstated.
4	Corporate actions	Requires split- and dividend-adjusted prices for return and trend calculations; adjustment methodology should be verified.
5	Execution assumptions	Next-open market fills may be optimistic around gaps, halts, high volatility, or large rebalance trades.
6	Risk-control state	Drawdown triggers, trailing stops, and highest-close-since-entry must be updated only with information available at or before each decision time.

Caveats / known limitations

• Backtest covers a relatively short 2020-2025 sample and may be sensitive to the strong performance of mega-cap technology stocks.
• Static top-cap universe can introduce survivorship and selection bias if constituents were not point-in-time.
• Costs are unspecified, while the strategy generated 3,977 trades; turnover, bid-ask spread, slippage, and taxes could materially reduce realized returns.
• Volatility targeting and drawdown de-leveraging can reduce tail exposure but may also de-risk after losses and miss sharp rebounds.
• ATR stops executed at next open do not guarantee stop prices and can gap through during market stress.
• Per-name cap, max positions, and large-cap concentration may still leave meaningful factor, sector, and mega-cap exposure.
• Parameter set is fairly complex, increasing the risk of overfitting without walk-forward or out-of-sample validation.

Results

The backtest produced a positive total return of 65.52% over 2020-01-01 to 2025-10-08, with moderate risk-adjusted performance: Sharpe 0.73, Sortino 0.87, and Calmar 0.36. The realized volatility was 15.63% and maximum drawdown reached -25.54%, indicating that the crash-protection layer reduced but did not eliminate substantial equity drawdown risk. Win rate was 64.93% with profit factor 1.35 across 3,977 trades, suggesting a broadly profitable but turnover-heavy implementation whose live viability depends heavily on realistic execution costs and point-in-time universe handling.

Backtest metrics snapshot

Metric	Value
Total Return	65.52%
Sharpe	0.73
Sortino	0.87
Calmar	0.36
Max Drawdown	-25.54%
Volatility	15.63%
Win Rate	64.93%
Profit Factor	1.35
Total Trades	3977
Symbols	50 (AAPL, ABBV, ABT, ACN, ADBE, AMD, AMZN, AVGO, AXP, BAC, +40 more)