zenigame-fx-alpha-sieve

Out-of-sample validation for Stage C survivors is automated by running the Alpha Sieve script, which applies a holdout embargo and a 90-day OOS window to re-test candidates and generate robustness reports.

Running an Alpha Sieve across multiple instruments is handled by invoking the automated sieve execution with a configurable run-id or run-number, processing prior Stage C candidates and surfacing pass rates and top performers.

A holdout embargo in trade analysis isolates a segment of data from the optimization process, and the 90-day OOS window then tests the Stage C survivors against this unseen data to assess strategy robustness.

Yes, out-of-sample backtest validation requires the scripts/alpha_factory/run_alpha_sieve.py script to execute the sieve and produce result reports in the reports/alpha-sieve directory.

Alpha Sieve reports for risk analysis include key metrics such as status, candidate counts, pass rates, Sharpe ratios, PnL, and top performing candidates to summarize the OOS validation results.

Re-testing prior trading strategy candidates under extended OOS conditions is necessary when you need to confirm the robustness of Stage C survivors across different runs and instruments before deployment.