PHASE 2 / CORE STRATEGY & EVIDENCE
This page explains the frozen candidate itself. The research journey explains how the strategy arrived here; this page explains what the candidate does, what evidence supports it, and what the evidence still cannot prove.
这一页解释冻结候选策略本身。研究路径页负责讲它怎么走到这里；这一页负责讲它到底做什么、证据支持到哪里、还有什么不能证明。
A weekly, long-only convertible-bond rotation strategy that ranks a cleaned universe by bond-side cheapness and equity-side momentum, then uses a retention band to reduce noisy turnover.
这是一条周频、只做多的可转债轮动策略：先清洗可交易债池，再按债性便宜和正股动量排序，最后用持仓保留区间减少无意义换手。
它不是预测单只转债的完整定价，也不是静态持有最便宜标的。它更像一个低频横截面实验：在信用和制度风险已经被基本处理的样本池里，持续提高组合对便宜、低溢价和正股近期强势候选的暴露。
It does not fully price individual convertible bonds, and it is not a buy-and-hold basket of the cheapest bonds. It is a low-frequency cross-sectional experiment: inside a cleaned universe, the framework repeatedly allocates toward candidates with better cheapness, lower premium and stronger recent underlying-stock behavior.
| Layer | What the rule does | 中文说明 |
|---|---|---|
| Universe boundary | Keep only bonds that are tradable and not dominated by known institutional risks. | 先保留可交易、制度风险不过度显性的债。 |
| Bond-side score | Prefer lower price and lower conversion premium. | 偏向价格较低、转股溢价率较低的候选。 |
| Equity-side score | Add underlying-stock 20-day momentum. | 加入正股 20 日动量，避免只暴露在便宜但缺少弹性的候选上。 |
| Portfolio construction | Select top candidates weekly, but retain existing holdings if rank remains acceptable. | 每周形成排名靠前的候选组合；已有持仓若未出现明显恶化，则允许继续保留。 |
| Execution assumption | Use weekly rebalance, T+2 entry and 15 bps one-way cost as the main research convention. | 用周频调仓、T+2 入场和单边 15bp 成本作为主要研究口径。 |
Implementation label. Internally, the frozen candidate is top12_keep37: maintain a 12-bond target basket, while allowing an existing holding to remain if it stays within the top 37.
实现标签。 内部版本名是 top12_keep37：目标组合宽度为 12 只；已有持仓如果仍在前 37 名以内，就允许继续保留。
Research interpretation. The rule slightly broadens portfolio breadth and treats small weekly rank changes as potential noise rather than immediate evidence of economic deterioration.
研究解释。 这条规则略微提高组合宽度，并把单周排名的小幅波动视为潜在噪声，而不是立即触发组合调整的经济恶化证据。
图示说明。 这张图先定义四条对照线，再比较 NAV 和回撤：最终候选、旧候选、无 keep 缓冲版本，以及清洗后可转债池的等权 beta 锚。