What we tested and rejected

Failed to establish a reliable advantage over a plain static ladder. No successor optimiser programme follows.

Cleared none of the seven risk points it was tested at. Closed with an explicit no-parameter-rescue rule.

Small expected-shortfall gains never produced the required drawdown improvement, and the return drag remained.

Static sector trajectory cleared none of its six gates. No static sector or theme allocation is used.

The estimated benefit was small, unreliable, and did not justify the added complexity.

Materially underperformed the simpler benchmark with a much deeper active drawdown.

Failed the pre-registered forecast-skill gate and were reclassified as climatology only — no better than the long-run average.

Beat climatology slightly across 146 eligible quarters, but its paired statistic reached 0.806 against a required 3.468. The edge was too small to admit.

Missed its frozen minimum improvement despite adequate statistical power to detect one.

The strongest variant failed the first chronological half, passed one stress family of several, and reached an adjusted significance of p = 1.

The unlevered version failed both its excess-return and Sharpe gates, so the leveraged stages were cancelled rather than attempted.

All three proxies failed at a fixed 5% replacement allocation. This rejects the tested expressions, not every possible one.