Strategy Card

Correlation-Breakdown Momentum Crash Defender (Winners Crowding Trigger + Defensive Rotation) — strategy card

Spec ID: spec-corrbreakdown-momcrash-us-stocks-1771868534 · Generated: 2026-06-23 07:05 UTC

Cluster: Momentum · Sub Cluster: Momentum Crash Defensive Rotation

One-line description

Buys recent large-cap US equity momentum winners in normal regimes, but rotates into low-volatility, quality, and cash exposure when crowding/correlation-breakdown crash signals appear. Signals are computed at the close and traded at the next open with weekly rebalancing.

Why this trade exists

Cross-sectional momentum can earn a premium from underreaction, slow information diffusion, and institutional benchmarking, but the trade is prone to sharp reversals when winners become crowded and their correlations/betas rise together. This strategy attempts to preserve the upside of long-only momentum while reducing exposure during likely momentum-crash regimes.

The defensive trigger is structural rather than purely price-based: it watches whether the winner basket becomes unusually internally correlated, whether winner correlations exceed loser correlations, whether winner beta shifts higher, whether the momentum factor suffers a material drawdown, or whether the broad market trend turns risk-off. In those states, the portfolio rotates toward stocks with lower realized volatility and quality proxies, plus a cash sleeve.

Algorithm

[code omitted from public view]

Parameters

Param	Value	Notes
Universe	Top 200 US stocks by capitalization	Daily bars; provided symbol set includes large-cap US equities and some listed non-US/ADR-like names.
Backtest window	2018-01-01 to 2023-01-01	One contiguous historical test period.
Rebalance frequency	Weekly	Signals at close; trades at next open.
Momentum lookback	252 trading days	Cross-sectional rank signal.
Skip period	20 trading days	Avoids short-term reversal / look-ahead around most recent returns.
Winners / losers fraction	20% / 20%	Winners are held in risk-on; losers are used for diagnostics.
Correlation window	60 trading days	Used for winner/loser internal correlation estimates.
Correlation z-score lookback	252 trading days	Used to normalize crowding/correlation signals.
Winner internal corr z threshold	1.5	Crash signal when winner crowding is unusually high.
Corr spread z threshold	1.0	Crash signal when winners are more correlated than losers unusually.
Beta shift lookback	60 trading days	Measures change in winner basket market sensitivity.
Beta shift z threshold	1.0	Crash signal when winner beta rises unusually.
Momentum drawdown trigger	-12%	Crash signal from recent momentum-factor drawdown.
Market trend filter	50-day MA vs 200-day MA	Risk-off when fast MA is below slow MA on largest ETF proxy.
Crash regime rule	Enter if >=2 signals; exit if <1 signal	2-week smoothing applied to signals.
Risk-on allocation	100% long momentum winners	Equal weight with caps.
Risk-off allocation	50% low-vol, 30% quality, 20% cash	Defensive rotation sleeve.
Position sizing	Equal weight with 10% cap	Max 40 positions.
Risk limits	Max leverage 4.0; sector cap 35%; stop loss 12%	Portfolio and position-level controls.
Costs	Not specified	Backtest appears to use no explicit transaction-cost/slippage assumption.

Look-ahead audit

#	Concern	Status
1	Signal/execution timing	Signals use close(t) information and execute at next open, reducing same-bar look-ahead risk.
2	Momentum formation	Uses 252-day lookback with 20-day skip; avoids using the most recent month in ranking.
3	Rolling z-scores	Correlation and beta z-scores should be computed only from historical windows ending at or before signal close.
4	Universe construction	Top-200-by-cap universe may embed survivorship or point-in-time membership risk unless constituents/caps are point-in-time.
5	Fundamental quality proxy	Quality metrics over 8 quarters require reporting-lag handling; must not use restated or future filings.
6	ETF market proxy	Largest ETF proxy selection should be fixed or point-in-time to avoid future capitalization knowledge.

Caveats / known limitations

• No explicit trading costs or slippage were supplied despite high turnover: 10,249 trades over the test.
• Universe may have survivorship and point-in-time selection bias if today’s top-cap symbols were applied historically.
• Quality proxy details are abstract; results depend heavily on data availability, lagging, and restatement handling.
• Defensive rotation can lag abrupt crashes if correlation/beta diagnostics trigger after losses begin.
• Risk-off rules are parameter-rich and may be overfit to the 2018-2022 regime.
• Long-only momentum remains exposed to broad equity beta; the strategy is not market neutral.
• Sector caps and position limits may materially alter realized exposure versus the theoretical signal basket.

Results

From 2018-01-01 to 2023-01-01, the backtest produced a 53.05% total return with 13.56% volatility and a Sharpe ratio of 0.78. Drawdown control was moderate rather than exceptional, with a -21.74% maximum drawdown and Calmar of 0.41. The 76.37% win rate and 1.73 profit factor are encouraging, but the large trade count and absence of explicit costs mean implementation frictions could materially reduce realized performance.

Backtest metrics snapshot

Metric	Value
Total Return	53.05%
Sharpe	0.78
Sortino	0.92
Calmar	0.41
Max Drawdown	-21.74%
Volatility	13.56%
Win Rate	76.37%
Profit Factor	1.73
Total Trades	10249
Symbols	200 (AAPL, ABBV, ABNB, ABT, ACN, ACON, ADBE, ADI, ADM, ADP, +190 more)