No system predicts markets every time — anyone claiming otherwise is selling you something. The real question is whether an I-Ching signal adds information on top of ordinary momentum and trend models. The answer from our 30-day back-test on liquid A-shares: the year-pillar signal hit 67% on its own; momentum alone hit 36%; a fused multi-agent score (I-Ching cycles + quant data) hit 53% with far fewer, higher-conviction signals.
The year pillar (annual heavenly stem and earthly branch) is a slow, structural signal. It does not flip with daily noise; it describes the energy regime of the whole year. In trending markets, regime signals beat noise-chasing. That is why our fused engine gives the year pillar the largest vote weight and damps weak low-confidence signals automatically.
Single signals are fragile. Our engine runs six independent agents — three I-Ching cycle engines (year, month, solar term) and three quantitative engines (XGBoost direction, amplitude, momentum) — then fuses them by confidence-weighted voting. Strong signals require agreement across at least four agents. That structure is what separates a research tool from a lucky streak.
---