---
run_id: 21
date: "2026-03-11 13:36:58 UTC"
question: "validate 0DTE strangle profit target"
total_hypotheses: 23
confirmed: 0
rejected: 0
pending: 23
rounds: 8
---

# 研究迴圈報告 — Run #21

> 研究主題：validate 0DTE strangle profit target
> 產生時間：2026-03-11 13:36:58 UTC

## 執行摘要

本次研究迴圈共進行 **8** 輪，產生 **23** 個假說。
最高信心假說（confidence=1.0000）：
> 0-DTE ATM_straddle strangle with 50% PT, $1 slippage

## 各輪摘要

| 輪次 | Agent 數 | 假說數 | 持續時間 | Token 用量 |
|------|----------|--------|----------|------------|
| R1 | 4 | 5 | — | — |
| R2 | 3 | 6 | — | — |
| R3 | 3 | 12 | — | — |
| R4 | 3 | 0 | — | — |
| R5 | 4 | 0 | — | — |
| R6 | 3 | 0 | — | — |
| R7 | 2 | 0 | — | — |
| R8 | 2 | 0 | — | — |

## 輪次評分（Keep/Discard Gate）

| 輪次 | 綜合分數 | Sharpe | Drawdown | 樣本量 | 決策 | 改善% |
|------|----------|--------|----------|--------|------|-------|
| R1 | 0.2750 | 0.0000 | 0.5000 | 0.0000 | KEEP | — |
| R2 | 0.2656 | 0.0000 | 0.5000 | 0.0000 | KEEP | -3.42% |
| R3 | 0.2221 | 0.0000 | 0.5000 | 0.0000 | DISCARD | -19.24% |
| R4 | 0.2188 | 0.0000 | 0.5000 | 0.0000 | DISCARD | -20.44% |
| R5 | 0.2228 | 0.0000 | 0.5000 | 0.0000 | DISCARD | -18.98% |
| R6 | 0.2188 | 0.0000 | 0.5000 | 0.0000 | DISCARD | -20.44% |
| R7 | 0.1250 | 0.0000 | 0.5000 | 0.0000 | DISCARD | -54.55% |
| R8 | 0.1893 | 0.0000 | 0.5000 | 0.0000 | DISCARD | -31.16% |

## 假說清單

### Round 1

- ⏳ **[validate 0DTE strangle profit target]** (confidence=0.9200)
  H1: 25% profit target maximizes risk-adjusted returns for 0DTE short strangles
- ⏳ **[validate 0DTE strangle profit target]** (confidence=0.9000)
  H2: Combining 25% profit target with 1.5x credit stop-loss yields optimal Sharpe for 0DTE strangles
- ⏳ **[validate 0DTE strangle profit target]** (confidence=0.8000)
  H3: Profit targets should NOT be dynamically scaled by implied volatility for 0DTE strangles
- ⏳ **[validate 0DTE strangle profit target]** (confidence=0.8800)
  H4: 0DTE strangle profit targets are hit predominantly in the first 90 minutes, enabling morning-only trading
- ⏳ **[validate 0DTE strangle profit target]** (confidence=0.9300)
  H5: Higher profit targets increase average P&L but degrade risk-adjusted returns due to tail risk exposure

### Round 2

- ⏳ **[entry_timing]** (confidence=0.0000)
  0DTE short strangle entry timing sweep: test 7 entry times (09:31 through 13:00) with 1% OTM strikes and 50% profit target
- ⏳ **[entry_timing]** (confidence=0.9100)
  H1: 09:31 market-open entry dominates all later entry times for 0DTE short strangles on risk-adjusted returns
- ⏳ **[entry_timing]** (confidence=0.8900)
  H2: A minimum credit filter ($3.50) dramatically improves 09:31 entry Sharpe from 0.22 to 0.34 by eliminating unprofitable low-vol days
- ⏳ **[entry_timing]** (confidence=0.8800)
  H3: Delaying entry to 10:00 to let opening volatility settle does NOT improve returns — it destroys the premium advantage
- ⏳ **[entry_timing]** (confidence=0.9300)
  H4: The 25% PT is typically hit within 31 minutes of 09:31 entry, making this a morning-only rapid-fire strategy
- ⏳ **[entry_timing]** (confidence=0.8700)
  H5: The optimal combined strategy is 09:31 entry + credit>=$3.50 filter + 25% PT + 1.5x SL, achieving Sharpe 0.34 with 93.5% win rate

### Round 3

- ⏳ **[None]** (confidence=1.0000)
  0-DTE ATM_straddle strangle with 50% PT, $1 slippage
- ⏳ **[None]** (confidence=1.0000)
  0-DTE OTM_0.5pct strangle with 50% PT, $1 slippage
- ⏳ **[None]** (confidence=1.0000)
  0-DTE OTM_1.0pct strangle with 50% PT, $1 slippage
- ⏳ **[None]** (confidence=1.0000)
  0-DTE OTM_1.5pct strangle with 50% PT, $1 slippage
- ⏳ **[None]** (confidence=1.0000)
  0-DTE OTM_2.0pct strangle with 50% PT, $1 slippage
- ⏳ **[None]** (confidence=0.0000)
  0-DTE OTM_3.0pct strangle with 50% PT, $1 slippage
- ⏳ **[None]** (confidence=1.0000)
  0-DTE Asym_P2_C05 strangle with 50% PT, $1 slippage
- ⏳ **[strike_selection]** (confidence=0.0000)
  H1: Fixed-dollar OTM strike widths ($20/$30/$50) outperform percentage-based (0.5%/1%/2%) for 0DTE strangles because SPX intraday moves are more absolute-dollar-bounded than percentage-bounded
- ⏳ **[strike_selection]** (confidence=0.0000)
  H2: Asymmetric strike placement (put 1.5% OTM, call 0.7% OTM) outperforms symmetric 1% OTM because SPX has persistent negative skew — downside vol is overpriced relative to upside
- ⏳ **[strike_selection]** (confidence=0.0000)
  H3: Wide strangle (2% OTM) with NO profit target outperforms narrow strangle (0.5% OTM) WITH 50% profit target on risk-adjusted basis because the narrow strangle's tail risk from breaches exceeds the benefit of early exits
- ⏳ **[strike_selection]** (confidence=0.0000)
  H4: Strike width should adapt to VIX level — use 0.7% OTM when VIX<20, 1.5% when VIX 20-30, 2.5% when VIX>30 — outperforming any fixed-width strangle
- ⏳ **[strike_selection]** (confidence=0.0000)
  H5: Iron condor (buying wings $20 further OTM) at same short strikes as naked strangle reduces max loss enough to improve risk-adjusted returns despite lower net credit

## 負面結果（已排除的方向）

- **R3** [expiration_choice] 0-DTE ATM_straddle strangle with 50% PT, $1 slippage; 0-DTE OTM_0.5pct strangle with 50% PT, $1 slippage; 0-DTE OTM_1.0pct strangle with 50% PT, $1 sl...
  失敗原因：Composite score 0.2221 below baseline
  分數：0.2221，交易數：0
- **R4** [position_sizing] Round 4 research on position_sizing
  失敗原因：Composite score 0.2188 below baseline
  分數：0.2188，交易數：0
- **R5** [exit_rules] Round 5 research on exit_rules
  失敗原因：Composite score 0.2228 below baseline
  分數：0.2228，交易數：0
- **R6** [regime_filter] Round 6 research on regime_filter
  失敗原因：Composite score 0.2188 below baseline
  分數：0.2188，交易數：0
- **R7** [volatility_surface] Round 7 research on volatility_surface
  失敗原因：Composite score 0.1250 below baseline
  分數：0.1250，交易數：0
- **R8** [correlation] Round 8 research on correlation
  失敗原因：Composite score 0.1893 below baseline
  分數：0.1893，交易數：0

## 錯誤日誌摘要

- **backtest_crash**: 1 次

最近錯誤：

- [backtest_crash] Script entry_timing_strangle_0dte.py timed out after 120s

## 方法評分

| Agent | 方法 | 成功率 | 提出數 | 確認數 | 樣本數 |
|-------|------|--------|--------|--------|--------|
| researcher | validate 0DTE strangle profit target | 0.00% | 5 | 0 | 5 |
| optimizer | entry_timing | 0.00% | 1 | 0 | 1 |
| researcher | entry_timing | 0.00% | 5 | 0 | 5 |
| optimizer | general | 0.00% | 7 | 0 | 7 |
| researcher | strike_selection | 0.00% | 5 | 0 | 5 |

## 建議與後續行動

- 23 個假說仍待驗證，可增加輪次或手動檢視。
- 最常見錯誤類型為 **backtest_crash**（1 次），建議優先修復。