Fractal-zigzag entry plug-in on 3-minute Parquet

[lgbarn]

Parent

#1

What to build

The first concrete RLStrategy entry side: detect_entries driven by the fractal-zigzag pivot detectors (the trigger-scan winners) with live_edge newest-bar confirms, over the six-symbol 3-minute Parquet data (NQ, ES, RTY, YM, GC, SI; 2021-04 through 2026-06). Each candidate entry carries direction, entry price, and a 1x ATR initial stop. All observation features strictly causal (bar i uses only bars <= i).

Acceptance criteria

• Entries are truncation-invariant: running detection on a truncated series yields identical entries up to the truncation point (prior art: the fractal detector lookahead proofs)
• Every candidate entry has a 1x ATR initial stop attached at detection time
• All six symbols load from Parquet and produce plausible entry counts over the full 5-year span (reported in the PR)
• Feature computation is causal — verified by the same truncation test applied to the observation vectors

Blocked by

• Restore the generic RL pipeline from git #2

[lgbarn]

PR opened: #12 — all four acceptance criteria covered by TDD seams; CI green; per-symbol entry counts reported in the PR body.