PHASE 2 / RESEARCH JOURNEY
This page is the research path, not a parameter page. It shows how a traditional dual-low premise evolved into a credit-aware, equity-aware and execution-aware convertible-bond rotation framework.
这一页讲研究路径，不以参数为叙事中心。它要说明：一个传统“双低”前提，如何逐步演化成带有信用边界、正股弹性和执行纪律的可转债轮动框架。
这条研究不是从某个内部参数名开始的。更基础的问题是：在个人资金、公开数据和低频执行约束下，可转债市场是否存在可以被清楚解释、反复检验、并且有现实可执行性的横截面机会？
The research did not start from an internal parameter label. The starting question was simpler: under personal-capital size, public data and low-frequency execution, does the convertible-bond market contain a cross-sectional opportunity that can be explained, tested repeatedly and implemented realistically?
Figure 1 / 图 1. Research path map
这张图只保留研究顺序：先判断市场适配，再验证双低 baseline，再加入正股动量，再定义样本边界，最后处理组合构建和执行噪声。没有通过机制、平台或前向解释的模块，不进入主线。
The chart keeps the research order simple: market fit, dual-low baseline, stock-momentum score, sample boundaries, then portfolio construction and execution noise. Modules that fail mechanism, platform or forward explanation do not enter the main line.
可转债研究不能直接从因子开始。这个市场有强赎、退市期、自然到期、停牌、转股价、转股溢价率、正股复权、发行规模和成交状态。任何一个字段处理错，都会让回测表现混入数据口径收益。
Convertible-bond research cannot start directly from factors. The market contains strong-call events, delisting periods, natural maturity, suspensions, conversion prices, conversion premiums, adjusted stock prices, issue size and tradability status. If one of these fields is mishandled, backtest performance can include data-convention returns.
| Research check | Why it mattered | 中文说明 |
|---|---|---|
| Tradability and delisting status | Do not rank bonds outside normal trading conditions. | 先排除无法正常交易或退市阶段的标的。 |
| Strong-call PIT status | Use call-risk information only after it becomes observable. | 强赎只能用时点可得信息，不能偷看未来。 |
| Natural maturity, not noisy delist date | Avoid excluding strong performers simply because they later redeemed. | 持有期过滤应看自然到期，避免误伤飞升后赎回的样本。 |
| Stock adjustment and momentum field | The score later depends on underlying-stock behavior. | 正股字段和复权错位，会直接污染动量信号。 |
| Realistic execution lag | Separate signal observation from implementable execution. | 信号观察和可执行成交之间需要现实延迟。 |
这一步的意义很朴素：先保证研究对象是真的可交易、可观察、可复现。没有这一层，后面的 alpha 讨论很容易建立在错误样本上。
The point is simple: make sure the research object is tradable, observable and reproducible. Without this layer, later alpha discussion can easily be built on the wrong sample.