cyb50-quant/research/dragon/v2/openspec/changes/followthrough-mid-exit-shadow-probe/proposal.md

Why

The lean profit-loop review showed one clear next question: the only subtype worth pursuing is mid_zone_very_weak_b1, and its current probe path likely fails because the reentry is clipped too early by the default glued take-profit logic.

We do not need a broad sell-system rewrite. We need one narrow, shadow-only exit experiment that asks:

• if a repaired mid-zone followthrough reentry is still structurally healthy,
• should it be allowed to survive the first mild kdj_sell instead of being cut by knife_take_profit_2_glued?

What Changes

• Add a narrow helper for followthrough-reentry-specific kdj_only exit hold.
• Add config gates for this exit treatment, all default-off.
• Add a new shadow branch:
• alpha_first_glued_followthrough_mid_exit_probe
• Add a focused review script comparing:
• RC1 base
• original mid followthrough probe
• new mid exit probe
• Add a regression test pinning the new branch fixture.
• Expose the new branch in daily pipeline outputs only.

Impact

• Affected code: config, branch config, strategy sell path, daily pipeline, new review script, new test, new OpenSpec pack.
• Runtime impact: negligible.
• Governance impact: none to RC1 default path.