SRBMPR validation engine implementation
A modular, production-ready quantitative architecture built to enforce information leakage control, structural purging, and combinatorial purged cross-validation (CPCV) under strict institutional standards.
Modular architecture and structural containers
The core package separates behavioral event detection from structural validation. Within srbmpr_container.py, immutable data structures record event boundaries (t0 to t1) alongside version-locked methodology tags. The engine does not reconstruct or alter historical market structures; it treats them as verified structural records to maintain total auditability.
Strict purging logic and leak prevention
Implemented inside srbmpr_purge.py, the purge protocol acts as an unyielding firewall against temporal overlap. Any candidate training event whose information span intersects a designated test interval is systematically excised before training splits are formed. Combined with configurable post-event embargo windows, look-ahead bias is eliminated at the foundation.
Combinatorial purged cross-validation runner
The ValidationRunner module coordinates combinatorial path generation across grouped market intervals. By evaluating multiple out-of-sample paths simultaneously, the engine tests the stability of psychological market structures without overfitting. Institutional researchers and data teams can inspect reproducible run identifiers, group partitions, and diagnostic verification reports.
Create Your Own Website With Webador