作者： 什么是FINRA卖空头寸与回补天数
了解FINRA如何每半月报告卖空头寸，通过回补天数与流通股比例衡量做空程度，助您掌握市场风险。
卖空头寸是指股票已卖出但尚未回购的总股数。这是未平仓卖空头寸的累计余额，由 FINRA 每半月报告一次。该指标是衡量交易员看空股票程度的标准衡量标准，通过两个比率来体现：回补天数和流通股卖空比例。本文所有数据均源自该监管数据集，点击即可查询各项数据背后的详情。
什么是股票的卖空头寸 (Short Interest)？
卖空交易颠覆了常规的交易顺序：卖空者先借入股票，按当前价格卖出，目标是以更低的价格买回。买回股票的行为称为“平仓” (covering)，即将借来的股票归还给出借方。在平仓之前，该头寸保持开启状态；每个经纪商持有的未平仓卖空头寸总和，即为该股票的卖空头寸。
卖空头寸是一个“水平”指标——即某一时间点的未偿余额，类似于贷款余额。当新卖空单开启时，该数值上升；当卖空单平仓时，该数值下降。它反映的是存量而非单日交易情况。这与每日卖空“成交量” (short volume) 不同，后者是标记为卖空的交易流，且大多属于常规的 做市商 业务；两者常被混淆，我们在 卖空头寸与卖空成交量 中对两者进行了测量。为了使原始股数在不同股票间具有可比性，我们使用了两种标准化指标：卖空股数占流通股比例 (short percent of float) 和 回补天数 (days to cover)，下文将针对实际代码进行测量。
任何定义都必须包含一项风险提示：当股价上涨时，卖空会亏损。由于股价没有上限，未平仓卖空的损失是无限的——这与多头头寸不同，多头头寸的最大损失仅限于投入的本金。
做空头寸是如何计量与报告的？
做空头寸属于监管披露内容，而非交易所实时行情。根据 FINRA Rule 4560，每家会员经纪商必须在指定的结算日报告其账面上的未平仓做空头寸，每两周一次：一次在月中，一次在月末。该规则涵盖所有上市及场外交易的权益证券。其统计的是头寸而非交易量——即公司在结算日当天，其客户及自有交易部门仍需偿还的借入股票数量，无论期间发生了何种交易。FINRA 会将这些申报文件汇总为每个结算日一份文件：每种证券占一行，包含做空股数、日均成交量，以及基于两者计算出的回补天数。FINRA 文件指南介绍了其来源机制；做空最严重的排行榜则展示了最新的数据：
每个数字背后的完整 SQL
WITH (SELECT max(settlement_date) FROM global_markets.stocks_short_interest) AS latest
SELECT concat(monthName(latest), ' ', toString(toDayOfMonth(latest)), ', ', toString(toYear(latest))) AS latest_settlement_date,
count() AS tickers_reported,
multiIf(count() < 1000, toString(count()),
concat(toString(intDiv(count(), 1000)), ',', lpad(toString(count() % 1000), 3, '0'))) AS tickers_reported_fmt,
round(sum(si) / 1e9, 1) AS total_shares_short_b
FROM
(
SELECT ticker, max(short_interest) AS si
FROM global_markets.stocks_short_interest
WHERE settlement_date = latest
GROUP BY ticker
)截至 June 30, 2026 结算日，该文件涵盖了 22,207 个代码，总计包含 56.8 十亿 股未平仓做空头寸——这是对整个美国股市做空情况的每两周一次的快照。
做空头寸报告日期
“每月两次”包含两条规则：月中结算日为每月15日，若15日遇周末或节假日，则提前至前一个工作日；月末结算日为当月最后一个工作日。报告文件通常在结算后数天至数周内公布。以下是过去五个月内每次结算的日期、星期，以及数据公布前的延迟天数：
每个数字背后的完整 SQL
SELECT concat(monthName(d), ' ', toString(toDayOfMonth(d)), ', ', toString(toYear(d))) AS settlement_date,
weekday,
securities_on_file,
multiIf(securities_on_file < 1000, toString(securities_on_file),
concat(toString(intDiv(securities_on_file, 1000)), ',', lpad(toString(securities_on_file % 1000), 3, '0'))) AS securities_on_file_fmt,
publication_lag_days
FROM
(
SELECT settlement_date AS d,
formatDateTime(settlement_date, '%W') AS weekday,
uniqExact(ticker) AS securities_on_file,
dateDiff('day', settlement_date, toDate(min(_ingest_time))) AS publication_lag_days
FROM global_markets.stocks_short_interest
WHERE settlement_date >= today() - INTERVAL 5 MONTH
GROUP BY settlement_date
)
ORDER BY d过去五个月的 9 结算日期，星期列显示了工作日调整情况（例如，若15日为周末，则结算日期显示为13日）。最新的 June 30, 2026 结算于 Tuesday，涵盖 22,207 证券，数据公布耗时 11 天；前一次耗时 16 天。
结论：您今天看到的任何做空头寸数据，反映的都是最近一次结算时的仓位，即一到数周前的数据（我们测量了每次报告的延迟时间）。“当前”做空头寸永远不是当前的。
实例解析：特斯拉的卖空头寸
结合股票代码来理解定义会更直观。以下是特斯拉过去两年的每次结算数据，该文件包含了卖空头寸与同期日均成交量的对比：
每个数字背后的完整 SQL
SELECT concat(monthName(d), ' ', toString(toDayOfMonth(d)), ', ', toString(toYear(d))) AS settlement_date,
short_interest_m_shares,
avg_daily_volume_m_shares
FROM
(
SELECT settlement_date AS d,
round(max(short_interest) / 1e6, 1) AS short_interest_m_shares,
round(max(avg_daily_volume) / 1e6, 1) AS avg_daily_volume_m_shares
FROM global_markets.stocks_short_interest
WHERE ticker = 'TSLA'
AND settlement_date >= today() - INTERVAL 2 YEAR
GROUP BY settlement_date
)
ORDER BY d截至 June 30, 2026 结算日，TSLA 的卖空股数为 79.1 百万 股，而日均成交量为 46 百万股；两年前该余额为 87 百万股。数据点统计：共有 47 个结算日 —— 每月两次，期间无其他数据。
做空股本占比与回补天数
回补天数（Days to cover）——即卖空倍数——是将卖空股数除以日均成交量：按股票的正常交易速度，需要多少个完整交易日才能回补所有空头头寸？本文件为您计算该指标，需注意一个约定：该数值最小为 1.00，不会显示低于此值的结果。
做空股本占比（Short percent of float）是将卖空股数除以流通股数（即总股本减去内部持股、受限股及其他紧密持有的股份块），用以衡量可交易股份中卖空的比例。需要注意的是，正如 流通股说明 开篇所述：没有任何申报文件会直接报告流通股数。 公司每季度仅报告“发行在外的股份总数”；您看到的每个流通股数值均为供应商计算的结果。下表改用经审计的分母进行计算——即最新季度申报中的基本发行股份总数——这使得每个百分比都代表一个保守的底线，因为流通股数绝不会大于发行股份总数。
每个数字背后的完整 SQL
WITH (SELECT max(settlement_date) FROM global_markets.stocks_short_interest) AS latest,
shares AS
(
SELECT tk AS ticker,
argMax(basic_shares_outstanding, (filing_date, period_end)) AS shares_out
FROM global_markets.stocks_income_statements
ARRAY JOIN tickers AS tk
WHERE tk IN ('AAPL', 'KO', 'MSFT', 'NVDA', 'TSLA', 'GME')
AND timeframe = 'quarterly'
AND filing_date >= today() - INTERVAL 1 YEAR
AND basic_shares_outstanding > 0
GROUP BY tk
),
si AS
(
SELECT ticker,
max(short_interest) AS shares_short,
max(days_to_cover) AS dtc
FROM global_markets.stocks_short_interest
WHERE settlement_date = latest
AND ticker IN ('AAPL', 'KO', 'MSFT', 'NVDA', 'TSLA', 'GME')
GROUP BY ticker
)
SELECT si.ticker AS ticker,
round(si.shares_short / 1e6, 1) AS shares_short_m,
round(shares.shares_out / 1e6, 0) AS shares_outstanding_m,
multiIf(shares.shares_out < 1e9, toString(toUInt64(round(shares.shares_out / 1e6, 0))),
concat(toString(intDiv(toUInt64(round(shares.shares_out / 1e6, 0)), 1000)), ',',
lpad(toString(toUInt64(round(shares.shares_out / 1e6, 0)) % 1000), 3, '0'))) AS shares_outstanding_m_fmt,
round(100.0 * si.shares_short / shares.shares_out, 2) AS short_pct_of_shares_out,
round(si.dtc, 2) AS days_to_cover
FROM si
INNER JOIN shares ON si.ticker = shares.ticker
ORDER BY indexOf(['AAPL', 'KO', 'MSFT', 'NVDA', 'TSLA', 'GME'], si.ticker)查看第五行的 Tesla：在 3,225 百万发行股份中，有 79.1 百万股被卖空，占比为 2.45%，回补天数为 1.72 天。这是典型的超大市值特征：单个股份数量看似庞大，但实际处于极高的流动性和巨大的股份基数之上。Nvidia 的数据更具代表性：在六家公司中，其卖空头寸绝对值最大，为 310.1 百万股，但在 1.99 天的回补天数下，仅占发行股份总数的 1.28%。仅看绝对数值并无意义。
GameStop 是个例外：其卖空股份占发行股份总数的 12.48%，是表中其他任何股票的数倍，回补天数为 10.75 天。
何谓高做空比例？
目前并无官方界定。“高”仅是一种惯例。最准确的校准方式是衡量市场整体水平。以下是最新文件中所有流动性股票（日均成交量至少 100 万股）的卖空天数：
每个数字背后的完整 SQL
WITH (SELECT max(settlement_date) FROM global_markets.stocks_short_interest) AS latest
SELECT count() AS liquid_tickers,
multiIf(count() < 1000, toString(count()),
concat(toString(intDiv(count(), 1000)), ',', lpad(toString(count() % 1000), 3, '0'))) AS liquid_tickers_fmt,
round(quantileExact(0.5)(dtc), 1) AS median_days_to_cover,
round(quantileExact(0.9)(dtc), 1) AS p90_days_to_cover,
min(dtc) AS lowest_days_to_cover,
countIf(dtc >= 10) AS names_at_10_plus
FROM
(
SELECT ticker, max(days_to_cover) AS dtc, max(avg_daily_volume) AS adv
FROM global_markets.stocks_short_interest
WHERE settlement_date = latest
GROUP BY ticker
)
WHERE adv >= 1000000在 3,202 只流动性股票中，卖空天数中位数为 2.4 天，第 90 百分位数为 5.9 天，最低值为 1 —— 即基准线。仅有 45 只股票的卖空天数达到 10 天或以上：几天属于常态，两位数比例才代表市场确实拥挤。
接下来是另一种指标：针对近期有季度申报记录的流动性股票，按做空股数占发行总股数比例进行分组，并列出各组的卖空天数中位数。
每个数字背后的完整 SQL
WITH (SELECT max(settlement_date) FROM global_markets.stocks_short_interest) AS latest,
si AS
(
SELECT ticker,
max(short_interest) AS shares_short,
max(avg_daily_volume) AS adv,
max(days_to_cover) AS dtc
FROM global_markets.stocks_short_interest
WHERE settlement_date = latest
GROUP BY ticker
HAVING adv >= 1000000
),
shares AS
(
SELECT tk AS ticker,
argMax(basic_shares_outstanding, (filing_date, period_end)) AS shares_out
FROM global_markets.stocks_income_statements
ARRAY JOIN tickers AS tk
WHERE timeframe = 'quarterly'
AND filing_date >= today() - INTERVAL 9 MONTH
AND basic_shares_outstanding > 0
GROUP BY tk
HAVING shares_out >= 10000000
),
joined AS
(
SELECT si.ticker AS ticker,
100.0 * si.shares_short / shares.shares_out AS short_pct,
si.dtc AS dtc
FROM si
INNER JOIN shares ON si.ticker = shares.ticker
WHERE 100.0 * si.shares_short / shares.shares_out <= 50
)
SELECT multiIf(short_pct < 2, 'Under 2%',
short_pct < 5, '2-5%',
short_pct < 10, '5-10%',
short_pct < 20, '10-20%',
'20%+') AS short_pct_bucket,
count() AS tickers,
round(100.0 * count() / sum(count()) OVER (), 1) AS pct_of_tickers,
round(quantileExact(0.5)(dtc), 2) AS median_days_to_cover
FROM joined
GROUP BY short_pct_bucket
ORDER BY min(short_pct)此处的分布比大多数读者预期的更广。仅有 11.9% 的股票做空比例低于发行总股数的 2%；中间两组（28.6% 和 29%）涵盖了大部分处于 2% 至 10% 之间的股票；8.9% — 169 只股票的比例超过 20%。若以流通股计算，所有数值都会更高。
最后一列显示了回报关系：随着做空比例上升，卖空天数中位数也随之增加，从做空最少组的 1.83 天上升到做空最多组的 5.37 天。两种不同的分母（一个是股本，一个是成交量）得出的排名一致：相对于股本被大量做空的股票，其平仓速度通常也较慢（卖空天数的解释）。
GameStop, January 2021: What a Squeeze Looks Like in the Data
Every discussion of high short interest arrives at the same reference case. GameStop's record across the 10 settlements from November 2020 to March 2021, with the closing price on each of those dates (prices as-traded, before GME's later 4-for-1 split):
每个数字背后的完整 SQL
WITH px AS
(
SELECT toDate(toTimeZone(window_start, 'America/New_York')) AS day,
argMax(toFloat64(close), window_start) AS close_px
FROM global_markets.delayed_stocks_minute_aggs
WHERE ticker = 'GME'
AND window_start >= '2020-11-01 00:00:00'
AND window_start < '2021-04-02 00:00:00'
AND toHour(toTimeZone(window_start, 'America/New_York')) BETWEEN 9 AND 15
GROUP BY day
),
si AS
(
SELECT settlement_date,
round(max(short_interest) / 1e6, 1) AS shares_short_m,
round(max(days_to_cover), 1) AS days_to_cover
FROM global_markets.stocks_short_interest
WHERE ticker = 'GME'
AND settlement_date >= '2020-11-01'
AND settlement_date <= '2021-03-31'
GROUP BY settlement_date
)
SELECT concat(monthName(si.settlement_date), ' ', toString(toDayOfMonth(si.settlement_date)), ', ', toString(toYear(si.settlement_date))) AS settlement_date,
si.shares_short_m AS shares_short_m,
si.days_to_cover AS days_to_cover,
round(px.close_px, 2) AS gme_close
FROM si
INNER JOIN px ON si.settlement_date = px.day
ORDER BY si.settlement_dateThe setup sits in the top rows: at the November 13, 2020 settlement, 67.5 million GME shares were short, days to cover stood at 14, and the stock closed at $11.02. The balance kept building into year-end — 71.2 million shares at the December 31, 2020 settlement, the largest of the 10 prints — with the stock at $18.81.
Then the table turns over. At the January 29, 2021 settlement, with the stock closing at $328.24, short interest printed 21.4 million shares — roughly a third of the year-end balance — and days to cover had collapsed to 1, the floor. Both halves of the ratio moved at once: shorts covered (the numerator fell) and volume exploded (the denominator rose). By the March 31, 2021 settlement the balance sat at 10.7 million.
Two lessons a price chart will not teach. The record is retrospective — the January 15, 2021 snapshot, taken with the stock at $35.49, reached the public only on the usual multi-week lag. And a low days-to-cover reading is not always calm: the ratio hit its floor in the wildest stretch of the episode, describing a torrent of volume rather than a small position. The full squeeze anatomy walks those weeks.
做空头寸常见问题解答
什么是高做空头寸？
目前没有官方界定标准。在最近一次结算中，美国流动性股票的中位数为 2.4 天回补天数，仅有 45 的股票回补天数达到 10 天或更多。另一方面，8.9% 的有申报记录的流动性股票，其做空股数占发行总股数的 20% 以上——交易台通常将此区间视为重度做空。
做空头寸的报告日期是什么时候？
每月两次：每月 15 日（若遇周末或节假日则提前至前一个工作日）以及每月最后一个工作日。根据 FINRA Rule 4560，经纪商会报告截至这些结算日的未平仓头寸，数据会在几天至几周后公布——此处最新的数据结算日为 June 30, 2026，并在 11 天后发布。
什么是流通股做空比例？
做空头寸除以流通股数（即可自由交易的股数）。由于申报文件中不包含流通股数据，因此所有公布的版本均为供应商的估算值。若以经审计的发行总股数（始终不小于流通股数）为基准，Tesla 的最新比例为 2.45%，GameStop 为 12.48%。
什么是回补天数？
做空头寸除以日均成交量：即所有未平仓做空头寸全部回补所需的典型交易天数。Tesla 的最新数据为 1.72 天，该文件将该比例的下限设定为 1.00。
做空头寸与做空成交量是否相同？
不相同。做空头寸统计的是截至结算日的未平仓头寸——这是一个存量指标，每月公布两次。做空成交量统计的是每日卖空的股数——这是一个流量指标，其中大部分属于做市商对冲 (两者皆是，但衡量方式不同)。
以上每个面板均为存储的查询指令——请打开下方的 SQL 语句，替换任意股票代码，并在 Strasmore 终端上重新运行。