ATR state defined trade management.

Two regimes, two management structures.

The switch classifies the environment and the classification selects the doctrine wholesale:

• Normal / out-of-session → BE at 1:1.4, 50% partial at 1R, static runner to 2R.
• Trend + in-session → BE at 1:1.6, 50% partial at 1R, ATR trailing on the remainder (typically ×1.5 volatility-adjusted) — with the occasional no-partial slot from quota.
• The switch feeds branch selection before entry — it never re-labels a trade after it starts moving.

Tuning changes how often Trend fires — nothing else.

The parameters adjust sensitivity, not doctrine:

• atrLen (default 28) — longer is smoother and slower to switch; shorter is snappier with more whipsaws. 20–28 suits intraday regimes.
• maLen (default 10) — the threshold's averaging window. Raise to 14–20 to reduce flicker; drop to 7–9 to react sooner.
• k (default 1.20) — the sensitivity knob. 1.10–1.15 yields more Trend days; 1.30–1.40 admits only the strongest expansions.

The rules were extracted from trades, then frozen.

The original ATR-state doctrine wasn't designed on a whiteboard — it was distilled from journaled trades whose management outcomes were graded against their volatility states, then written down and versioned. Provenance matters: rules extracted from evidence can be re-audited against new evidence; rules from intuition can only be argued about.