E-20. Long/medium/short daily trend × 5m independent entry, 6 families (2026-07-15)
- Preregistered D-026. 2 macro types × 6 entry types × hold 30/60/120m = 36 runs.
- Every strategy is preserved as a separate branch in
macro_trend_families.py. Indicators are not all stacked into one signal; only the direction context is shared. A cap of 5 per day and a 1-hour cooldown apply to all. - Activity: 1.30~3.21/day on the best of each family, meeting the user's target range.
- Best before costs: slow 20d/60d/EMA90 + EMA20 pullback reclaim + 120m, 1,766 signals, 2.16/day, +0.02542%/signal, 5 of 6 windows positive, gross positive at 3 horizons.
- After market-order costs -0.17458%, at 1.5x costs -0.27458%. Train pass 0; val/test not opened.
- Conclusion: entry frequency and period stability improved, but 2-hour directional moves are smaller than costs.
Source and migration evidence
Source: backtest/docs/experiment-log.md#E-20
Original SHA-256: 069c4aaaf53a20930ca8d94ba6365b4ce2b6bc560171d662a4c179ea55a6857c
Raw data and account information are not part of the public package. The original results do not replace statistical or scientific revalidation. The English text is a translation; numbers, tables and charts are carried over unchanged.