# Dragon Deep Oversold Audit

- audited trades: `14`
- all buy dates aligned with workbook: `True`
- all sell dates aligned with workbook: `True`
- winners: `5` / `14`
- fast failures (holding <= 6d and negative return): `7`

## Entry Subtype Summary
- `positive_b1_rebound`: trades `2`, win_rate `0.00%`, avg_return `-2.98%`, avg_mfe `2.72%`, avg_mae `-5.95%`, fast_failures `1`
- `shallow_false_start`: trades `3`, win_rate `33.33%`, avg_return `-2.63%`, avg_mfe `2.20%`, avg_mae `-3.52%`, fast_failures `2`
- `deep_capitulation`: trades `1`, win_rate `0.00%`, avg_return `-2.05%`, avg_mfe `1.91%`, avg_mae `-2.08%`, fast_failures `1`
- `mixed_oversold`: trades `3`, win_rate `33.33%`, avg_return `-1.80%`, avg_mfe `2.82%`, avg_mae `-5.57%`, fast_failures `2`
- `classic_oversold`: trades `5`, win_rate `60.00%`, avg_return `0.04%`, avg_mfe `3.54%`, avg_mae `-2.28%`, fast_failures `1`

## Candidate Pressure Points
- `2022-03-11 -> 2022-03-15`: `positive_b1_rebound` / `immediate_failure` | buy a1 `-0.0526` b1 `0.0297` c1 `15.11` | return `-5.78%`
- `2022-02-09 -> 2022-02-11`: `mixed_oversold` / `immediate_failure` | buy a1 `-0.0550` b1 `-0.0510` c1 `10.04` | return `-5.42%`
- `2024-01-23 -> 2024-01-29`: `shallow_false_start` / `rebound_then_fail` | buy a1 `-0.0345` b1 `-0.0227` c1 `12.24` | return `-4.72%`
- `2024-01-18 -> 2024-01-22`: `shallow_false_start` / `immediate_failure` | buy a1 `-0.0318` b1 `-0.0234` c1 `12.90` | return `-3.30%`
- `2018-09-18 -> 2018-10-08`: `classic_oversold` / `rebound_then_fail` | buy a1 `-0.0439` b1 `-0.0667` c1 `9.73` | return `-3.25%`
- `2019-06-11 -> 2019-06-17`: `mixed_oversold` / `immediate_failure` | buy a1 `-0.0321` b1 `-0.0467` c1 `10.41` | return `-3.03%`
- `2022-09-26 -> 2022-09-30`: `deep_capitulation` / `flat_noise` | buy a1 `-0.0650` b1 `-0.0866` c1 `10.95` | return `-2.05%`
- `2024-08-15 -> 2024-08-20`: `classic_oversold` / `flat_noise` | buy a1 `-0.0283` b1 `-0.0843` c1 `11.49` | return `-0.92%`

## Quant Judgment
- This rule family cannot be bluntly removed because every current deep-oversold trade is workbook-aligned.
- The weakest local pattern is not the deepest capitulation bucket; it is the shallow or positive-B1 rebound subset.
- Any redesign should therefore prefer subtype gating or delayed confirmation for shallow rebounds rather than tighter global oversold thresholds.