Method
Rule / pseudocode
_Cross-sectional momentum: recent relative winners tend to keep outperforming over 1-12 month horizons. Documented across equities, sectors and countries; attributed to under-reaction and delayed information diffusion._
Out-of-sample equity curve
Out-of-sample performance
Hypothesis
Holding the top-2 of nine US sector ETFs ranked by trailing 3-month return, rebalanced monthly, earns positive after-cost OOS risk-adjusted returns versus equal-weight.
Results
Out-of-sample Sharpe was 0.79, versus 0.26 in-sample. A bootstrap test returned p = 0.005 (statistically significant). Walk-forward Sharpe stayed positive in 5 of 5 folds. Worst out-of-sample drawdown was -33.1%.
Analysis
In-sample (2007 to late-2019) the top-2 sector-momentum portfolio managed only a 0.26 Sharpe on 4.6% annualised return, and — because it concentrates capital in just two sectors — suffered a brutal -51% drawdown through the 2008 crisis. Out-of-sample (late-2019 to mid-2025) it looks far healthier: 18.2% annualised, a 0.79 Sharpe and a -33% drawdown. That is a very large IS-to-OOS improvement, and improvement of that shape is a warning sign, not a reassurance: it means the apparent edge is concentrated in the recent, strongly trending regime rather than being stable across history.
Crucially, the strategy is always fully invested in equities, so most of its return is simply equity beta. The honest benchmark is an equal-weight basket of all nine sectors, which posted a 0.63 OOS Sharpe over the same window. The momentum tilt lifts that to 0.79 — a real but modest improvement — while raising volatility and concentration risk. The gain over naive diversification is not large enough to be confident it is momentum alpha rather than luck in sector selection.
The block-bootstrap on 67 non-overlapping monthly returns gives p = 0.005, but this tests against a zero-return null, not against the equity risk premium the strategy is loaded on. A long-only equity strategy will clear a zero-edge test almost by construction, so the low p-value does not establish a momentum edge here.
Walk-forward folds were 0.17, 0.93, 1.95, 1.48, 0.04 — nominally 100% positive but with the first and last folds essentially flat, again pointing to a middle-of-sample phenomenon rather than a persistent effect. The IS/OOS Sharpe gap (0.26 vs 0.79) is exactly the sample-window dependence CLAUDE.md flags as grounds for caution.
Verdict: INCONCLUSIVE. Despite clearing the mechanical gates (54 position entries, OOS Sharpe 0.79, p < 0.10, all folds positive), the edge is confounded with equity beta, barely beats an equal-weight benchmark, and is heavily sample-window dependent with a poor in-sample record. It is not promoted; a cleaner test would measure the long/short momentum spread (top sectors minus bottom sectors) to strip out beta before drawing any conclusion.
Provenance
Synthesized from 1ui4ptb. Same thread flags equity strategies with strong cross-sectional dispersion this year. The cleanest credible test of that is classic cross-sectional momentum: rank sector ETFs by trailing return and hold the leaders, monthly rebalance — a well-studied premium distinct from the time-series breakout above.
Evidence
Walk-forward Sharpe by fold
---
_Generated by labs-algo-trading. Automated research — not financial advice. Backtests overfit; treat verdicts as hypotheses._
Evidence boundary
This record preserves the research boundary: Method was not retained in the source record. No result, threshold, or source has been inferred to fill that gap.