cyb50-quant/research/dragon/v2/dragon_deep_oversold_confirmation_experiments.md

Dragon Deep Oversold Confirmation Experiments

• Goal: test an alpha-first branch for weak deep-oversold subtypes by replacing immediate entry with QL confirmation inside a short waiting window.
• Scope: only positive_b1_rebound and shallow_false_start; default baseline behavior remains unchanged.

Summary

• baseline: trades 107, avg_return 2.52%, profit_factor 3.31, real BUY 106, real SELL 105, deep weak trades 5, confirmed deep trades 0
• confirm_weak_with_ql_1bar: trades 104, avg_return 2.67%, profit_factor 3.56, real BUY 101, real SELL 102, deep weak trades 2, confirmed deep trades 2
• confirm_weak_with_ql_2bar: trades 104, avg_return 2.67%, profit_factor 3.56, real BUY 101, real SELL 102, deep weak trades 2, confirmed deep trades 2
• confirm_weak_with_ql_3bar: trades 103, avg_return 2.73%, profit_factor 3.66, real BUY 101, real SELL 101, deep weak trades 1, confirmed deep trades 1

Delta Vs Baseline

• confirm_weak_with_ql_1bar: delta_avg_return 0.15%, delta_profit_factor 0.25, delta_deep_weak_avg_return -0.15%, real BUY 101, real SELL 102
• confirm_weak_with_ql_2bar: delta_avg_return 0.15%, delta_profit_factor 0.25, delta_deep_weak_avg_return -0.15%, real BUY 101, real SELL 102
• confirm_weak_with_ql_3bar: delta_avg_return 0.21%, delta_profit_factor 0.35, delta_deep_weak_avg_return 0.11%, real BUY 101, real SELL 101

Quant Judgment

• Best alpha-first branch in this pack: confirm_weak_with_ql_3bar with avg_return 2.73% and profit_factor 3.66.
• This pack is expected to lose workbook date alignment because entries are delayed by confirmation; treat it as a research branch, not a baseline patch.
• If the best branch improves weak-subtype trade quality materially, the next step should be to audit shifted dates trade-by-trade rather than porting it directly into the baseline.