# Next Decision

This was a DryRun sample expansion only. No real PostHedge exit was implemented.

## Candidate Assessment
- RF_120_NO_IMPROVE_1000: strongest balanced candidate. Stable across all required periods and both deposits, simple enough, no BasketClose/RecoveryClose kill in the sampled results.
- MH_240: also stable and simpler than RF. It may be easier to implement, but pure time-based exit has higher risk of cutting a Basket that would later improve.
- ATV_TICK30_R2_ADVERSE: stable as a risk signal and can be used as an auxiliary condition. It should not be the first standalone implementation because tick-count behavior may be broker/data dependent.
- ATV_M1_R2_ADVERSE: stable but slower than 30-second tick signal and slightly lower total estimated improvement.

## Next Candidates Max 3
1. PostHedgeRecoveryFailureExit Compare: RF_120_NO_IMPROVE_1000 only.
2. PostHedgeMaxHoldingExit Compare: MH_240 only, if RF compare is rejected or too complex.
3. PostHedgeRecoveryFailure + TickVolume Composite Compare: only after RF standalone is tested, to reduce false exits.

## Decision
P0_A2_BASE_D4 remains the baseline. The sample expansion is strong enough to justify one real-exit comparison, but not enough to adopt a P1 default. The next implementation should test RF_120_NO_IMPROVE_1000 alone against P0.