03 / EVIDENCE & ARCHIVED VARIANTS
ETF is judged by what it does to the whole capital curve. The strongest evidence is portfolio-level risk efficiency.
ETF 要按它对完整资金曲线的影响来评价。最重要的证据，是组合层风险效率。
| Portfolio | CAGR | MDD | Calmar | Correlation to CB | Interpretation |
|---|---|---|---|---|---|
| CB only | 18.60% | -18.76% | 0.99 | 1.00 | Main return engine, but deeper drawdown. |
| 70% CB + 30% Treasury fallback | 15.35% | -10.92% | 1.41 | 0.26 | Earlier defensive version; already improves the path. |
| 70% CB + 30% ETF Stabilizer | 15.49% | -10.77% | 1.44 | 0.25 | Current sealed candidate. |
| 70% CB + 30% 50/50 Gold fallback | 15.70% | -10.53% | 1.49 | 0.23 | Observation; better metrics but more Gold-regime dependence. |
The result should not be read as “ETF is better than convertible bonds.” A cleaner reading is: convertible bonds provide the main return structure, while ETF absorbs part of the portfolio risk budget with lower correlation and a more defensive asset mix.
这个结果不应该被读成“ETF 比可转债好”。更准确的读法是：可转债提供主要收益结构，ETF 用较低相关性和更防守的资产组合承接一部分风险预算。
| CB / ETF | CAGR | MDD | Calmar | Reading | 中文说明 |
|---|---|---|---|---|---|
| 80 / 20 | 16.55% | -13.14% | 1.26 | Return engine heavier. | 收益引擎更重。 |
| 70 / 30 | 15.49% | -10.77% | 1.44 | Balanced compromise. | 平衡折中。 |
| 60 / 40 | 14.41% | -9.18% | 1.57 | More defensive. | 更防守。 |
| 50 / 50 | 13.31% | -7.56% | 1.76 | Smoothest path, but more alpha dilution. | 更平滑，但收益引擎被稀释更多。 |
The point is not that 70/30 is mathematically optimal. It is a readable compromise: convertible bonds remain the return engine, while ETF exposure is large enough to materially reduce drawdown.
这里不把 70/30 写成数学意义上的最优。它是一个容易解释的折中：可转债仍然承担收益引擎，ETF 的权重又足够大，可以明显压低回撤。
| Item | Boundary | 中文说明 |
|---|---|---|
| Convertible-bond data | Sealed net return series, 2017-02-14 to 2026-04-30. | 可转债使用封板后的净收益序列。 |
| ETF data | Domestic ETF close-price series, active from 2015-02-02 to 2026-05-15. | ETF 使用境内 ETF close 价格序列。 |
| Common window | Portfolio comparison uses 2017-02-14 to 2026-04-30. | 组合比较只使用两条收益序列的重叠区间。 |
| Cross-border ETF | S&P 500 exposure uses 513500.SH with a 5% close/NAV premium filter. |
标普500 使用境内跨境 ETF，并加入 5% 折溢价过滤。 |
This matters because ETF research can easily become messy when price source, NAV, premium and proxy choices are mixed together. Here, the portfolio result is judged only after the data boundary is made explicit.