13F-RS-v1.0 · versioned research framework
13F-RS 衡量一份公开申报对解释管理人持仓意图的研究可用度。它不是未来收益预测、管理人能力排名或跟单建议。
13F-RS measures how usable a public filing is for interpreting manager intent. It is not a return forecast, manager-skill ranking or copy-trading recommendation.
先验收数据，再解释持仓Validate first, interpret second
13F-RS-v1.0
| 维度Dimension | 满分Max | 判断规则Decision rule |
|---|---|---|
| 申报完整性Filing integrityfiling_integrity | 15 | 申报行数、汇总值、修订语义与单位能否对上。X0202 也不机械假设美元：用非期权普通股的 value / shares 隐含价格做二次校验。Whether rows, summary value, amendment semantics and units reconcile. Even X0202 is checked with value-per-share implied prices on non-option common holdings. |
| 实体映射Entity mappingentity_mapping | 10 | 品牌、法律申报人、CIK、Other Manager 与母公司关系是否解析完整，并避免把 13F-NT 误称为保密申请。Whether brand, legal filer, CIK, Other Manager and parent relationships are resolved, without mislabeling 13F-NT as confidential treatment. |
| 策略适配度Strategy fitstrategy_fit | 15 | 13F 与策略的匹配程度。集中多头通常更可读；平台、宏观、量化和多策略的可见切片更窄。How well 13F matches the strategy. Concentrated long portfolios are usually more readable; platform, macro, quant and multi-strategy slices are narrower. |
| 集中度信息密度Concentration signalconcentration | 10 | 前十集中度和有效持仓数衡量信息密度；集中不等于高水平，也不等于低风险。Top-ten concentration and effective position count measure information density; concentration is neither skill nor low risk. |
| 高确信度可识别性Conviction identifiabilityconviction_proxy | 15 | 当前用单仓 ≥2% 作为有限代理；没有基准主动权重时，不把绝对大仓位直接解释成主动确信度。Currently uses positions at or above 2% as a limited proxy; without benchmark active weights, absolute size is not treated as active conviction. |
| 延续性 / 换手Persistence / turnoverpersistence | 10 | 用股数延续、前十大延续和同价漂移组合估算季度快照换手；不是已实现交易流水。Uses share persistence, top-ten persistence and a same-price drift portfolio to estimate snapshot turnover; it is not realized transaction history. |
| 申报时滞Filing lagfiling_lag | 10 | 以 EDGAR 实际受理日计算可获得时滞，而不是假设季末即可知道；10 日内满分，45 日降至零。Uses the actual EDGAR acceptance date rather than pretending quarter-end availability; full freshness inside 10 days, decaying to zero at 45 days. |
| 不可见风险Opacity riskopacity | 10 | 期权、修订、保密遗漏与策略特有的账外敞口越多，越难恢复净方向。期权股数是标的股数，不是合约数。Options, amendments, confidential omissions and strategy-specific invisible exposure make net direction harder to recover. Option shares are underlying shares, not contract counts. |
| 独立共识Independent consensusindependent_consensus | 5 | 按独立品牌母体与策略类型去重；同向变化提高研究优先级，也可能意味着拥挤风险。Deduplicates by independent brand parent and style; shared change raises research priority but can also signal crowding. |
RS = Σ(weightᵢ × ratioᵢ)
覆盖度与 A–D 数据置信等级单独展示，不藏进总分。关键字段缺失时，页面不做精确排名。所有公式变化都提升版本号；如无法重跑历史，则保留旧分数版本。
Coverage and A–D data confidence are displayed separately rather than hidden inside the total. Missing critical fields prevent precise ranking. Formula changes increment the version; old score versions remain if history cannot be rerun.
Living method · 2026-08-16
Vanguard 与 Pershing Square 说明：NT 是 Notice，不等于保密处理。必须逐个报告期独立重建 Other Manager 图，并从相邻两期已知品牌节点的并集出发；本季关系不能直接倒套上季。
Vanguard and Pershing Square show that NT is a notice, not confidential treatment. Rebuild the Other Manager graph independently for every report date, starting from the union of brand nodes known across adjacent quarters; never project the current graph backward.
部分申报仍表现为旧式“千美元”口径。保留原值与倍数，并用普通股 value / shares 的中位隐含价格检测；异常会降低完整性置信度。
Some filings still behave like legacy thousands. Preserve raw value and multiplier, then test median common-stock value / shares; anomalies reduce integrity confidence.
RESTATEMENT 替换原表，NEW HOLDINGS 追加遗漏行。原始申报与修订链应能重建时点视图；当前页采用截至 as_of 的有效结果，回测只能使用当时已公开的链条。
RESTATEMENT replaces the original table; NEW HOLDINGS appends omitted rows. Source filings and the amendment chain should support point-in-time reconstruction; this page uses the effective result as of as_of, while backtests may use only the chain then public.
同一平台旗下多个申报实体不算多票；还要看策略类型是否独立，避免把同一数据源误当广泛共识。
Multiple filers under one platform are not multiple votes; style independence also matters so one source is not mistaken for broad consensus.
回测和判断只能使用 EDGAR accepted_at 之后的信息。季末日不是投资者可见日，45 天滞后会显著改变可交易性。
Backtests and decisions may use information only after EDGAR accepted_at. Quarter-end is not the public-availability date; the 45-day lag changes tradability.
13F 没有执行价、到期日、权利金或配套对冲。Put/Call 名义价值不能直接解释成看空/看多，也不能与股票总值机械相加。
13F omits strike, expiry, premium and paired hedges. Put/call notional value is not a direct bearish/bullish read and should not be mechanically added to stock exposure.
Research foundations
谁要申报、报告证券、季度时点与 45 天期限的制度原文。Primary instructions for filers, reportable securities, quarter-end snapshots and the 45-day deadline.
字段、版本、修订和电子申报结构的技术合同。Technical contract for fields, versions, amendments and electronic filing structure.
SEC 工作人员对申报主体、可报告证券、期权、合并申报与保密处理等操作问题的说明。SEC staff guidance on filers, reportable securities, options, combination reports and confidential treatment.
SEC 结构化汇总数据适合批量研究；单笔核验仍回到 EDGAR accession 与原始 XML。SEC structured datasets support bulk research; filing-level verification still returns to the EDGAR accession and source XML.
审计指出历史上存在准确性、完整性与监督缺口，支持逐份做行数、金额、证券和修订验收。The audit documented historical accuracy, completeness and oversight gaps, supporting filing-by-filing row, value, security and amendment checks.
研究高确信度持仓与组合其余部分的差异；提示应关注主动权重，而非仅看绝对仓位。Studies high-conviction holdings versus the rest of the portfolio; motivates active weights rather than raw size alone.
用持仓披露检验对冲基金选股和组合特征，同时展示只看公开多头的局限。Uses holdings disclosure to study hedge-fund stock selection and portfolio traits, while illustrating limits of the visible long book.
研究保密持仓与延迟披露，说明公开 13F 可能系统性漏掉最敏感的想法。Studies confidential holdings and delayed disclosure, showing that public 13F can systematically omit sensitive ideas.
讨论强制披露的成本、行为反应与复制风险。Examines costs, behavioral responses and imitation risk from mandatory portfolio disclosure.
在大型银行控股公司样本中发现证券匹配、市场估值和修订报告可能存在显著差异，支持把 13F 当作需验收的数据而非无误真值。Finds material security-matching, valuation and amendment discrepancies in a sample of large bank holding companies, supporting validation rather than blind trust.
系统研究 restatement 及其信息含量，说明修订不能静默覆盖历史；研究与回测应区分当时可见申报和后来修订后的有效结果。Studies restatements and their information content, showing why research and backtests should distinguish the filing visible at the time from the later effective result.
用实际 EDGAR 申报日期研究时滞选择，发现前置交易和投票动机与延迟有关；支持用 accepted_at 而不是统一假设 45 天。Uses actual EDGAR filing dates to study delay choice and links delays to front-running and voting motives; supports accepted_at rather than a blanket 45-day assumption.
把伯克希尔长期持仓与质量、低风险和杠杆因子联系起来；不是简单复刻当季 13F。Links Berkshire's long record to quality, low-risk and leverage factors; it is not a recipe for copying one quarter's 13F.
展示机构共同持有结构可用于研究产业与治理问题，超越“谁买了哪只股票”。Shows that common institutional ownership supports industry and governance research beyond who bought which stock.