Free Stock Recommendations API
Signals You Can Verify免费股票推荐 API
可验证的推荐信号
Evaluate recommendation endpoints by their evidence, timestamps, methodology, and limits—not by a buy or sell label alone.
评估股票推荐接口，应关注证据、时间戳、方法和使用限制，
而不能只看买入或卖出标签。
TL;DR核心摘要
An endpoint may rank symbols, but it should not place trades or replace suitability checks.
Useful results expose factors, observation time, methodology, and universe—not only a label.
Check request caps, data delay, coverage, and commercial-use terms before integration.
Backtest with point-in-time inputs and monitor drift before using a signal.
接口可以对股票代码排序，但不应直接代替交易和适当性判断。
有用的结果会说明因子、观察时间、方法和股票池，而不是只返回标签。
接入前要确认请求次数、数据延迟、覆盖范围和商业使用条件。
使用当时可得的数据回测，并监控信号漂移后再用于研究。
What a free stock recommendations API should return免费股票推荐 API 应该返回什么
A recommendation API converts market observations into ranked research output, often combining price behavior, fundamentals, estimates, news, or technical indicators. Without an input window, scoring rule, and effective time, a symbol list cannot be reproduced or evaluated.
股票推荐 API 把市场观察结果转化为有序的研究输出，常见输入包括价格走势、基本面、盈利预测、新闻或技术指标。若缺少输入时间窗、评分规则和生效时间，股票列表就无法复现或评价。
State whether the screen covers common stocks, ETFs, one exchange, or another universe, including delisting rules.
Give each candidate an as-of time and reason. Separate source observation time from recommendation generation time.
Document the score scale, holding horizon, invalidation condition, and refresh time.
For broker research, preserve firm, analyst when licensed, action, previous rating, new rating, announcement time, source ID, and whether the event is initiated, reiterated, upgraded, or downgraded.
State the lookback window, included firms, rating-to-score mapping, duplicate policy, stale-report expiry, and weighting. A consensus without its denominator is not comparable.
说明筛选对象是普通股、ETF、某个交易所还是其他股票池，并交代退市规则。
给每个候选标的提供数据截至时间和理由，并区分数据观察时间与推荐生成时间。
说明评分范围、持有周期、失效条件和更新时点。
对于券商研报，应在许可允许范围内保留机构、分析师、操作类型、原评级、新评级、发布时间、来源 ID，并区分首次覆盖、重申、上调和下调。
说明回看窗口、纳入机构、评级到分数的映射、去重规则、陈旧报告失效时间和权重。缺少分母的共识无法横向比较。
If target prices are included, store the currency, horizon, publication time, previous target, new target, and the stock price observed at that time. Never aggregate targets across currencies or mix a newly issued rating with an old target that happened to remain in the source.
如果响应包含目标价，还要保存币种、目标期限、发布时间、旧目标价、新目标价，以及当时观察到的股价。不能跨币种直接聚合，也不能把新发布的评级与来源中残留的旧目标价拼在一起。
Response fields that make a signal testable让推荐信号可检验的响应字段
| Field字段 | Meaning含义 | Why it matters主要用途 | Check核查重点 |
|---|---|---|---|
| symbol | Stable exchange-qualified identifier.带交易所信息的稳定证券标识。 | Joining prices and fundamentals.关联价格和基本面数据。 | Ticker reuse across venues.不同市场可能复用同一代码。 |
| as_of | UTC time when inputs were valid.输入数据有效时对应的 UTC 时间。 | Point-in-time testing.按当时可得信息回测或复核。 | Not the same as delivery time.不等同于接口返回时间。 |
| score / rank | Documented ordering value.有明确定义的排序数值。 | Comparing candidates consistently.以统一尺度比较候选标的。 | Requires scale and direction.必须说明范围和数值方向。 |
| reasons | Factors that contributed to output.促成推荐结果的具体因子。 | Review and debugging.人工复核与问题排查。 | Avoid generated explanations without evidence.避免没有证据支持的生成式解释。 |
How to evaluate a free recommendation endpoint如何评估免费股票推荐接口
Confirm the universe, schedule, sources, corporate-action handling, and whether observations are delayed or real time.
Use only information available at each as-of time. Include fees, slippage, delistings, and empty periods.
Track score changes and timestamp progress. Stale data should return an explicit status, not silently repeat a recommendation.
Record quotas, attribution, commercial rights, retention limits, support, and upgrade pricing.
The same broker action may arrive through multiple wires or summaries. Deduplicate by source event ID and a stable event fingerprint before counting analysts.
A higher consensus can result from bearish analysts dropping out of the window. Track analyst count, additions, removals, and rating changes alongside the average.
确认股票池、更新周期、数据源、公司行动处理方式，以及数据属于延迟行情还是真实时间行情。
每条推荐只能使用其数据截至时间之前公开的信息，并计入费用、滑点、退市股票和空窗期。
记录评分变化和时间戳推进情况。数据过期时应返回明确状态，而不是静默重复旧推荐。
记录请求配额、署名要求、商业使用权、数据保留限制、支持方式和升级价格。
同一券商评级可能通过多个资讯源或摘要重复到达。计入分析师数量前，应按来源事件 ID 和稳定事件指纹去重。
共识变高也可能只是偏空分析师退出统计窗口。除了平均值，还要同时跟踪分析师数量、新增、移除和评级变动。
Use QVeris to find and inspect market data capabilities用 QVeris 查找并检查市场数据能力
There is no verified QVeris result dedicated to every recommendation methodology. Use the QVeris documentation to understand capability calls and the Playground to inspect candidate market or fundamental inputs. Keep the recommendation logic, evidence, and timestamps in your own auditable layer.
目前没有一个经过核实的 QVeris 结果，可以覆盖所有股票推荐方法。可以先通过 QVeris 文档了解能力调用，再到 Playground 检查候选行情或基本面输入。推荐逻辑、证据和时间戳应保留在你自己的可审计层中。
- Search by required data capability, not by a promised outcome.
- Inspect authentication, schema, examples, and terms before use.
- Keep the logic testable and retain every source timestamp.
- 按所需数据能力搜索，而不是按承诺的投资结果搜索。
- 使用前检查鉴权方式、响应结构、示例和服务条款。
- 保持推荐逻辑可检验，并保存每条信号对应的数据源时间戳。
FAQ常见问题
No. Suitability, risk tolerance, portfolio context, and execution remain outside its response.
Only after checking its scale and calibration on unseen periods.
Not always. Compare observation, generation, and delivery times separately.
Keep the request, response, timestamps, method version, universe, and source identifiers.
Keep the original label, then map it through a versioned provider-specific table. Do not assume every “outperform” or five-point scale has the same meaning.
Yes. Reports can age out, coverage can start or stop, duplicates can be corrected, and the eligible analyst set can change.
不构成。适当性、风险承受能力、投资组合背景和交易决定都不在响应范围内。
只有了解评分范围，并在未参与建模的时间段验证校准效果后才有参考价值。
不一定。应分别比较数据观察时间、推荐生成时间和接口返回时间。
保存请求、响应、时间戳、方法版本、股票池和数据源标识。
先保留原始标签，再通过带版本的供应商专属映射表转换。不能默认所有“跑赢大盘”或五级评分含义相同。
会。旧报告可能退出窗口，覆盖关系可能开始或结束，重复事件可能被修正，符合条件的分析师集合也会变化。
References and next steps参考资料与下一步
Financial Modeling Prep documentation
Alpha Vantage documentation
QVeris documentation
QVeris Playground
Financial Modeling Prep 文档
Alpha Vantage 文档
QVeris 文档
QVeris Playground