Strasmore Research
市場回顧 Matt Connor作者： Matt Connor · 更新於 2026-07-25

2026年7月7日市場回顧

半導體重挫但指數波瀾不驚：晶片股大跌，NVDA逆勢收紅，下跌家數為上漲兩倍，SPY價差創一個月最窄。

2026年7月7日週二，不僅吐回了週一重啟行情中成長股的那一半漲幅，還倒貼更多：QQQ收盤-1.82%，完全抹去週一漲幅，而DIA則為-0.29%。盤面底下，下跌家數與上漲家數之比達二比一：4034檔下跌對上2091檔上漲。當日真正的故事，是半導體板塊出現數倍於大盤指數的跌幅——而該板塊的幾大巨頭並未參與其中。以下每個數字皆取自預存查詢——展開任一區塊即可查看確切SQL。

計分板

每一項變動皆比較 7 月 7 日最後一個正常交易分鐘 K 線，與前一個交易日 7 月 6 日週一。

查詢SPY / QQQ / DIA / IWM — 7月7日 與 7月6日 收盤比較，常規交易時段
每個數據背後的精確 SQL 語法
WITH prior AS (
    SELECT ticker, argMax(close, window_start) AS prior_close
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker IN ('SPY', 'QQQ', 'DIA', 'IWM')
      AND window_start >= '2026-07-06 13:30:00' AND window_start < '2026-07-06 20:00:00'
    GROUP BY ticker
),
sess AS (
    SELECT ticker,
           argMin(open, window_start) AS day_open,
           argMax(close, window_start) AS day_close,
           max(high) AS day_high,
           min(low) AS day_low,
           round(toFloat64(sum(volume)) / 1e6, 1) AS shares_traded_m
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker IN ('SPY', 'QQQ', 'DIA', 'IWM')
      AND window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00'
    GROUP BY ticker
)
SELECT
    s.ticker AS ticker,
    round(toFloat64(p.prior_close), 2) AS prior_close,
    round(toFloat64(s.day_open), 2) AS day_open,
    round(toFloat64(s.day_close), 2) AS day_close,
    round((toFloat64(s.day_open) / toFloat64(p.prior_close) - 1) * 100, 2) AS gap_pct,
    round((toFloat64(s.day_close) / toFloat64(s.day_open) - 1) * 100, 2) AS intraday_pct,
    round((toFloat64(s.day_close) / toFloat64(p.prior_close) - 1) * 100, 2) AS pct_change,
    round(toFloat64(s.day_high), 2) AS day_high,
    round(toFloat64(s.day_low), 2) AS day_low,
    s.shares_traded_m AS shares_traded_m
FROM sess s
JOIN prior p ON s.ticker = p.ticker
ORDER BY s.ticker

四大指數 ETF 當中，QQQ 的 -1.82% 跌幅領先；SPY 收 -0.48%，IWM 收 -0.9%，而 DIA 的 -0.29% 是最溫和的數字。跳空與盤中走勢的分歧顯示，開盤時磁帶的兩端看法並不一致：QQQ 的隔夜跳空開在 -1.17%，而開盤到收盤的盤中走勢為 -0.65% —— 主要跌幅來自跳空 —— 而 DIA 跳空 上漲 0.47%，最終仍收黑，其開盤到收盤的走勢印出 -0.76%。自週四以來，領漲權每次都出現輪動：週四由 DIA 領漲，週一由 QQQ 領漲，週二再度換成 DIA。

當天是否異常？

查詢QQQ 與 SPY：7月7日相對過去一個月交易日的絕對波幅排名（排名1 = 最大絕對變動）
每個數據背後的精確 SQL 語法
SELECT
    round(anyIf(cc_pct, ticker = 'QQQ' AND d = toDate('2026-07-07')), 2) AS qqq_close_over_close_pct,
    arrayCount(x -> x > abs(anyIf(cc_pct, ticker = 'QQQ' AND d = toDate('2026-07-07'))), groupArrayIf(abs(cc_pct), ticker = 'QQQ' AND d != toDate('2026-07-07'))) + 1 AS qqq_abs_move_rank,
    countIf(ticker = 'QQQ') AS qqq_sessions_compared,
    round(anyIf(oc_pct, ticker = 'SPY' AND d = toDate('2026-07-07')), 2) AS spy_open_to_close_pct,
    arrayCount(x -> x > abs(anyIf(oc_pct, ticker = 'SPY' AND d = toDate('2026-07-07'))), groupArrayIf(abs(oc_pct), ticker = 'SPY' AND d != toDate('2026-07-07'))) + 1 AS spy_abs_move_rank,
    countIf(ticker = 'SPY') AS spy_sessions_compared,
    toString(min(d)) AS first_session
FROM (
    SELECT ticker, d,
           (close_px / lagInFrame(close_px) OVER (PARTITION BY ticker ORDER BY d) - 1) * 100 AS cc_pct,
           oc_pct
    FROM (
        SELECT ticker, toDate(toTimeZone(window_start, 'America/New_York')) AS d,
               argMax(toFloat64(close), window_start) AS close_px,
               (argMax(toFloat64(close), window_start) / argMin(toFloat64(open), window_start) - 1) * 100 AS oc_pct
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker IN ('SPY', 'QQQ')
          AND window_start >= toDateTime('2026-06-04 00:00:00')
          AND window_start < toDateTime('2026-07-08 00:00:00')
          AND (toHour(window_start) * 60 + toMinute(window_start)) BETWEEN 810 AND 1199
        GROUP BY ticker, d
    )
)
WHERE isFinite(cc_pct)

就指數層面而言，並不特別。QQQ 的 -1.82% 收盤對收盤波動，按絕對值計算，在過去 21 個交易日中排名第 9；SPY 的開盤至收盤波動 -0.34%，則在過去 21 個交易日中排名第 15——包含 6 月下旬記憶體族群震盪的過去一個月，設下了很高的比較基準。7 月 7 日的特殊之處，在於表層之下的個股分歧。

市場廣度：每有一檔上漲，就有兩檔下跌

查詢7月7日成交額至少100萬美元標的之上漲與下跌家數
每個數據背後的精確 SQL 語法
WITH per_ticker AS (
    SELECT
        ticker,
        toFloat64(argMaxIf(close, window_start, window_start < '2026-07-07 00:00:00')) AS prior_close,
        toFloat64(argMaxIf(close, window_start, window_start >= '2026-07-07 00:00:00')) AS day_close,
        sumIf(toFloat64(close) * toFloat64(volume), window_start >= '2026-07-07 00:00:00') AS day_dollar_volume
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE (window_start >= '2026-07-06 13:30:00' AND window_start < '2026-07-06 20:00:00')
       OR (window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00')
    GROUP BY ticker
)
SELECT
    countIf(day_close > prior_close AND day_dollar_volume >= 1000000) AS advancers,
    countIf(day_close < prior_close AND day_dollar_volume >= 1000000) AS decliners,
    countIf(day_close = prior_close AND day_dollar_volume >= 1000000) AS unchanged,
    countIf(day_dollar_volume >= 1000000) AS liquid_tickers,
    count() AS tickers_traded_both_sessions,
    count() - countIf(day_dollar_volume >= 1000000) AS dropped_by_liquidity_filter,
    round(100.0 * countIf(day_close > prior_close AND day_dollar_volume >= 1000000)
        / countIf(day_dollar_volume >= 1000000), 1) AS advancer_pct
FROM per_ticker
WHERE prior_close > 0 AND day_close > 0

2091 檔上漲，4034 檔下跌，59 檔持平 — 流動性較高的標的中，有 33.8% 收紅，週一的收紅格局遭到逆轉。此篩選條件剔除 11460 檔雙交易時段標的中，成交金額低於 100 萬美元的 5276，在此列入統計。

半導體板塊崩解——唯獨最大咖的兩家沒事

賣壓集中在半導體族群：晶片製造商、供應它們的設備廠、圍繞它們的儲存生態系，以及疊在上面的槓桿型 ETF。下方列出連動性與跌幅；數據本身並未說明原因。

查詢7月7日晶片板塊：相較週一收盤變動、區間時點與美元交易量
每個數據背後的精確 SQL 語法
WITH per_name AS (
    SELECT
        ticker,
        toFloat64(argMaxIf(close, window_start, window_start < '2026-07-07 00:00:00')) AS prior_close,
        toFloat64(argMaxIf(close, window_start, window_start >= '2026-07-07 00:00:00')) AS day_close,
        maxIf(toFloat64(high), window_start >= '2026-07-07 00:00:00') AS day_high,
        minIf(toFloat64(low), window_start >= '2026-07-07 00:00:00') AS day_low,
        argMinIf(window_start, (toFloat64(low), toInt64(toUnixTimestamp(window_start))), window_start >= '2026-07-07 00:00:00') AS low_bar,
        argMaxIf(window_start, (toFloat64(high), -toInt64(toUnixTimestamp(window_start))), window_start >= '2026-07-07 00:00:00') AS high_bar,
        round(sumIf(toFloat64(close) * toFloat64(volume), window_start >= '2026-07-07 00:00:00') / 1e9, 2) AS day_dollar_bn
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker IN ('AMD', 'AVGO', 'INTC', 'KLAC', 'LRCX', 'MRVL', 'MU', 'NVDA', 'SNDK', 'SOXL', 'SOXS', 'STX', 'TER', 'WDC')
      AND ((window_start >= '2026-07-06 13:30:00' AND window_start < '2026-07-06 20:00:00')
        OR (window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00'))
    GROUP BY ticker
)
SELECT
    ticker,
    round(prior_close, 2) AS prior_close,
    round(day_close, 2) AS day_close,
    round((day_close / prior_close - 1) * 100, 2) AS pct_chg,
    formatDateTime(toTimeZone(high_bar, 'America/New_York'), '%H:%i') AS day_high_et,
    formatDateTime(toTimeZone(low_bar, 'America/New_York'), '%H:%i') AS day_low_et,
    toHour(toTimeZone(high_bar, 'America/New_York')) * 60 + toMinute(toTimeZone(high_bar, 'America/New_York')) AS high_minute_et,
    toHour(toTimeZone(low_bar, 'America/New_York')) * 60 + toMinute(toTimeZone(low_bar, 'America/New_York')) AS low_minute_et,
    round((day_high / day_low - 1) * 100, 2) AS range_pct,
    day_dollar_bn
FROM per_name
ORDER BY ticker

泰瑞達（-9.63%）與英特爾（-9.59%）跌掉將近十分之一的市值，邁威爾（-7.41%）、科磊（-7.23%）、科林研發（-6.8%）和超微（-6.41%）緊追在後。那些週一勉強反彈一半的記憶體暨儲存股直接反轉向下：SanDisk -7.1% 高低區間達 10.36%，威騰 -7.81%，希捷 -4.84%，以及美光 -4.61%——美光成交量高達 41.6 億美元，比另外三家儲存股加總還多（美光深度解析提供背景脈絡）。

板塊兩大巨頭置身事外：輝達收在 0.67% 的紅盤，博通收在 -0.83% 近乎平盤——整個板塊一片慘綠，唯獨最大兩檔成分股沒事（輝達六月深度解析）。

時間點完全同步——同一個早盤的真空區：泰瑞達、科林研發和美光分別在美東時間 10:41、10:41 和 10:41 打出盤中低點，SanDisk、邁威爾、科磊、博通和 SOXL 也在前後幾分鐘內見低（面板上載有各自的時間戳記），而 SOXS，那檔三倍反向半導體 ETF，則在 10:42 打出盤中高點——整段走勢上下顛倒。英特爾在其他個股止穩後仍持續下探，低點出現在 15:18。槓桿包裝直接反映跌幅規模：SOXL -15.15% 對比 SOXS +15.87%。

磁帶的另一面

二比一的紅色磁帶，仍有三分之一的綠色。在晶片股下跌時，哪些股票逆勢上漲：

查詢晶片股下跌時的輪動亮點 — 及當日最大單筆成交
每個數據背後的精確 SQL 語法
WITH per_name AS (
    SELECT
        ticker,
        toFloat64(argMaxIf(close, window_start, window_start < '2026-07-07 00:00:00')) AS prior_close_raw,
        toFloat64(argMinIf(open, window_start, window_start >= '2026-07-07 00:00:00')) AS day_open_raw,
        toFloat64(argMaxIf(close, window_start, window_start >= '2026-07-07 00:00:00')) AS day_close_raw,
        round(sumIf(toFloat64(close) * toFloat64(volume), window_start >= '2026-07-07 00:00:00') / 1e9, 2) AS day_dollar_bn
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker IN ('CRNX', 'CVX', 'JNJ', 'LLY', 'META', 'TSLA', 'UNH', 'XOM')
      AND ((window_start >= '2026-07-06 13:30:00' AND window_start < '2026-07-06 20:00:00')
        OR (window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00'))
    GROUP BY ticker
)
SELECT
    ticker,
    round(prior_close_raw, 2) AS prior_close,
    round(day_open_raw, 2) AS day_open,
    round(day_close_raw, 2) AS day_close,
    round((day_open_raw / prior_close_raw - 1) * 100, 2) AS gap_pct,
    round((day_close_raw / day_open_raw - 1) * 100, 2) AS intraday_pct,
    round((day_close_raw / prior_close_raw - 1) * 100, 2) AS pct_chg,
    day_dollar_bn
FROM per_name
ORDER BY ticker

能源與醫療保健撐起了綠色欄位：埃克森美孚 +3.78%、雪佛龍 +3.47%、嬌生 +3.07%、禮來 +2.75%、聯合健康 +2.41%。此面板上的兩檔巨型股走勢分歧：META 上漲 +2.52%，而 TSLA 幾乎回吐了週一的單向漲勢，收在 -4.02%。磁帶上最響亮的單一報價：Crinetics Pharmaceuticals (CRNX) 收盤 +98.79%，成交量 $4.19 億，幾乎翻倍——整個漲幅來自隔夜跳空（開盤價 98.91%；常規交易時段僅推動其 -0.06%），這種盤中重新定價的現象，發生在盤前交易之處。數據並未說明原因。

資金流向何處

查詢成交量領先者雙維度：交易金額前6大、交易股數前4大（一重複代號因實體驗證中暫予排除）
每個數據背後的精確 SQL 語法
SELECT leaderboard, ticker, dollar_volume_bn, shares_m,
    round(100 * if(leaderboard = 'by dollars traded', dollar_volume_bn, shares_m)
        / max(if(leaderboard = 'by dollars traded', dollar_volume_bn, shares_m)) OVER (PARTITION BY leaderboard), 1) AS pct_of_board_leader
FROM (
    SELECT
        'by dollars traded' AS leaderboard,
        ticker,
        round(sum(toFloat64(close) * toFloat64(volume)) / 1e9, 2) AS dollar_volume_bn,
        round(sum(toFloat64(volume)) / 1e6, 1) AS shares_m
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00'
      AND ticker NOT IN ('SPCX')
    GROUP BY ticker
    ORDER BY dollar_volume_bn DESC
    LIMIT 6
    UNION ALL
    SELECT
        'by shares traded' AS leaderboard,
        ticker,
        round(sum(toFloat64(close) * toFloat64(volume)) / 1e9, 2) AS dollar_volume_bn,
        round(sum(toFloat64(volume)) / 1e6, 1) AS shares_m
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00'
      AND ticker NOT IN ('SPCX')
    GROUP BY ticker
    ORDER BY shares_m DESC
    LIMIT 4
)
ORDER BY leaderboard ASC, if(leaderboard = 'by dollars traded', dollar_volume_bn, shares_m) DESC

MU 以 $41.6 億 美元領先成交金額榜——在 讓出榜首 僅一個交易日後，便重新奪回——緊隨其後的是 SPY（$28.68 億）與 QQQ（$26.54 億），另外三檔晶片股——NVDA（$19.95 億）、SNDK（$19.73 億）及 INTC（$13.82 億）——則佔據前六名的其餘席位。單一個股成交集中度正是 相對成交量 所揭示的現象。

在股數榜上，SOXS 成交 566.3 萬股——這檔三倍放空半導體 ETF 已連續第三個交易日位居股數榜首（週一 與 週四 留有前次紀錄），而這次正值其追蹤的板塊實際破位之日。CPOP 則是反向教材：161.3 萬股僅成交 $0.02 億——股數會美化低價股。基礎：7 月 7 日常規交易時段；一檔重複使用代碼的標的因待實體驗證而排除——其收據。

盤中交易輪廓

查詢每30分鐘交易股數區間，常規交易時段（十億股）
每個數據背後的精確 SQL 語法
SELECT
    formatDateTime(toStartOfInterval(toTimeZone(window_start, 'America/New_York'), INTERVAL 30 MINUTE), '%H:%i') AS et_time,
    round(sum(toFloat64(volume)) / 1e9, 2) AS shares_bn,
    round(100 * sum(toFloat64(volume)) / max(sum(toFloat64(volume))) OVER (), 1) AS pct_of_biggest_bucket
FROM global_markets.delayed_stocks_minute_aggs
WHERE window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00'
GROUP BY et_time
ORDER BY et_time

開盤前半小時成交 2 億股，13:30 時出現 0.64 億股的日內低量，尾盤 15:30 則放大至 2.14 億股——典型的微笑曲線，尾盤量能最重。

選擇權成交紀錄

查詢整日選擇權摘要：成交量、買權佔比、0DTE，以及包覆SPY收盤價的兩檔合約
每個數據背後的精確 SQL 語法
WITH
    (
        SELECT (groupArray(und), groupArray(strike), groupArray(typ), groupArray(vol), groupArray(avg_px), groupArray(is_0dte))
        FROM (
            SELECT any(underlying_symbol) AS und, any(toFloat64(strike_price)) AS strike, any(option_type) AS typ,
                   sum(size) AS vol, round(avg(toFloat64(price)), 3) AS avg_px,
                   if(substring(ticker, length(ticker) - 14, 6) = '260707', 1, 0) AS is_0dte
            FROM global_markets.options_trades
            WHERE sip_timestamp >= '2026-07-07 00:00:00' AND sip_timestamp < '2026-07-08 00:00:00'
            GROUP BY ticker
            ORDER BY vol DESC
            LIMIT 2
        )
    ) AS top2,
    (
        SELECT round(toFloat64(argMax(close, window_start)), 2)
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'SPY' AND window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00'
    ) AS spy_regular_close
SELECT
    round(count() / 1e6, 2) AS option_prints_m,
    round(toFloat64(sum(size)) / 1e6, 2) AS contracts_m,
    round(100.0 * sumIf(size, option_type = 'C') / sum(size), 1) AS call_pct_of_volume,
    round(100.0 * sumIf(size, substring(ticker, length(ticker) - 14, 6) = '260707') / sum(size), 1) AS same_day_expiry_pct,
    round(toFloat64(sumIf(size, substring(ticker, length(ticker) - 14, 6) = '260707')) / 1e6, 2) AS same_day_contracts_m,
    round(toFloat64(sumIf(size, underlying_symbol = 'SPY')) / 1e6, 2) AS spy_contracts_m,
    round(toFloat64(sumIf(size, underlying_symbol = 'QQQ')) / 1e6, 2) AS qqq_contracts_m,
    round(toFloat64(sumIf(size, underlying_symbol = 'INTC')) / 1e6, 2) AS intc_contracts_m,
    top2.1[1] AS top_contract_underlying,
    top2.2[1] AS top_contract_strike,
    top2.3[1] AS top_contract_type,
    if(top2.3[1] = 'C', 1, 0) AS top_contract_is_call,
    top2.4[1] AS top_contract_volume,
    top2.5[1] AS top_contract_avg_price,
    round(top2.2[1] - spy_regular_close, 2) AS top_strike_minus_spy_close,
    top2.1[2] AS second_contract_underlying,
    top2.2[2] AS second_contract_strike,
    top2.3[2] AS second_contract_type,
    if(top2.3[2] = 'P', 1, 0) AS second_contract_is_put,
    top2.4[2] AS second_contract_volume,
    top2.5[2] AS second_contract_avg_price,
    round(spy_regular_close - top2.2[2], 2) AS spy_close_minus_second_strike,
    top2.6[1] + top2.6[2] AS both_top_contracts_same_day,
    spy_regular_close AS spy_close
FROM global_markets.options_trades
WHERE sip_timestamp >= '2026-07-07 00:00:00' AND sip_timestamp < '2026-07-08 00:00:00'

選擇權交易量達 61.13 百萬口合約，分佈在 10.6 百萬筆成交中——規模與週一相當——大盤收黑之日，買權仍佔總成交量 55.6%。當日到期合約佔成交紀錄的 31.2%（19.08 百萬口合約）——零日到期合約佔比與尋常週二無異。當日最熱絡的兩檔合約將收盤價夾在中間，均為當日到期的 SPY 履約價：749 美元的買權（873438 口合約，均價 0.647 美元）收盤時價外 1.34 美元，而 747 美元的賣權（785834 口合約，均價 0.88 美元）則反向價外 0.66 美元收場。SPY 收在 747.66，恰好落在當日兩大押注之間，兩者皆到期歸零。

報價紀錄

本頁每個價格背後，都有一條報價串流——全國最佳買賣報價，涵蓋所有掛牌股票，持續不斷地重新報價，是這個資料庫中最難取得的數據集。每個交易時段，無論是否尋常，這裡都會對它進行衡量。

查詢全市場股票NBBO報價流：7月7日 與 7月6日 更新筆數比較
每個數據背後的精確 SQL 語法
SELECT
    round(countIf(toDate(sip_timestamp) = toDate('2026-07-07')) / 1e6, 2) AS jul7_updates_m,
    round(countIf(toDate(sip_timestamp) = toDate('2026-07-06')) / 1e6, 2) AS jul6_updates_m,
    round((countIf(toDate(sip_timestamp) = toDate('2026-07-07')) / countIf(toDate(sip_timestamp) = toDate('2026-07-06')) - 1) * 100, 1) AS day_over_day_pct,
    toHour(toTimeZone(maxIf(sip_timestamp, toDate(sip_timestamp) = toDate('2026-07-07')), 'America/New_York')) * 60
        + toMinute(toTimeZone(maxIf(sip_timestamp, toDate(sip_timestamp) = toDate('2026-07-07')), 'America/New_York')) AS jul7_last_quote_et_minute,
    round(countIf(sip_timestamp >= '2026-07-07 19:00:00' AND sip_timestamp < '2026-07-07 20:00:00') / 1e6, 2) AS jul7_close_hour_updates_m
FROM global_markets.cache_stocks_quotes
WHERE sip_timestamp >= '2026-07-06 00:00:00' AND sip_timestamp < '2026-07-08 00:00:00'

7 月 7 日，股票報價紀錄承載了 492.76 百萬次 NBBO 更新——較週一的 391.96 百萬次多出 25.7%。是重新報價更頻繁，而非報價區間擴大：

查詢空氣缺口中的觸及價差：報價中位數價差與報價速率，美東時間10:36–10:48 對比 午間對照時段（各12分鐘）
每個數據背後的精確 SQL 語法
SELECT
    ticker,
    round(quantileExactIf(0.5)((toFloat64(ask_price) - toFloat64(bid_price)) / ((toFloat64(ask_price) + toFloat64(bid_price)) / 2) * 10000, bid_price > 0 AND ask_price >= bid_price AND sip_timestamp >= '2026-07-07 14:36:00' AND sip_timestamp < '2026-07-07 14:48:00'), 2) AS pocket_spread_bps,
    round(quantileExactIf(0.5)((toFloat64(ask_price) - toFloat64(bid_price)) / ((toFloat64(ask_price) + toFloat64(bid_price)) / 2) * 10000, bid_price > 0 AND ask_price >= bid_price AND sip_timestamp >= '2026-07-07 17:30:00' AND sip_timestamp < '2026-07-07 17:42:00'), 2) AS midday_spread_bps,
    round(quantileExactIf(0.5)((toFloat64(ask_price) - toFloat64(bid_price)) / ((toFloat64(ask_price) + toFloat64(bid_price)) / 2) * 10000, bid_price > 0 AND ask_price >= bid_price AND sip_timestamp >= '2026-07-07 14:36:00' AND sip_timestamp < '2026-07-07 14:48:00')
        - quantileExactIf(0.5)((toFloat64(ask_price) - toFloat64(bid_price)) / ((toFloat64(ask_price) + toFloat64(bid_price)) / 2) * 10000, bid_price > 0 AND ask_price >= bid_price AND sip_timestamp >= '2026-07-07 17:30:00' AND sip_timestamp < '2026-07-07 17:42:00'), 2) AS spread_delta_bps,
    round(countIf(sip_timestamp >= '2026-07-07 14:36:00' AND sip_timestamp < '2026-07-07 14:48:00') / 1e3, 1) AS pocket_updates_k,
    round(countIf(sip_timestamp >= '2026-07-07 17:30:00' AND sip_timestamp < '2026-07-07 17:42:00') / 1e3, 1) AS midday_updates_k,
    round(countIf(sip_timestamp >= '2026-07-07 14:36:00' AND sip_timestamp < '2026-07-07 14:48:00') / countIf(sip_timestamp >= '2026-07-07 17:30:00' AND sip_timestamp < '2026-07-07 17:42:00'), 1) AS quote_rate_ratio,
    round(count() / 1e6, 2) AS session_updates_m,
    countIf(bid_price <= 0 OR ask_price <= 0 OR bid_price > ask_price) AS dropped_invalid_quotes
FROM global_markets.cache_stocks_quotes
WHERE ticker IN ('AMD', 'INTC', 'MU', 'NVDA', 'QQQ', 'SNDK', 'SOXL', 'SOXS', 'SPY', 'WDC')
  AND sip_timestamp >= '2026-07-07 13:30:00' AND sip_timestamp < '2026-07-07 20:00:00'
GROUP BY ticker
ORDER BY ticker

在早盤低點時，受檢視股票的重新報價頻率達到其日中水準的數倍——MU 為 3.4 倍、SanDisk 為 4.1 倍、AMD 為 4.6 倍——而買賣價差幾乎沒有變動：MU 在該時段的中位數價差為 5.02 個基點，對比日中為 4.81；NVDA 為 1.04 對比 1.01；SPY 則完全沒有變動（在 0.27 個基點的基礎上變動了 0 個基點）。真正承受了顯著價差懲罰的兩檔股票是 AMD（+2.6 個基點）和 Intel（+0.95）——Intel 是該時段過後仍持續下跌的唯一一檔——而槓桿型配對在底部實際上收窄了（SOXL -2.59 個基點，SOXS -1.49）。流動性真空看起來像是報價極寬且一片死寂；這次恰恰相反——是一場報價風暴，但價差卻被牢牢釘住。即便如此，交易成本仍因股票而異：SOXS 在底部報價寬達 19.59 個基點，對比 NVDA 的 1.04。

查詢SPY 以更新量加權的平均報價價差：7月7日於過去一個月排名（排名1 = 最窄）
每個數據背後的精確 SQL 語法
SELECT
    round(anyIf(avg_spread_cents, d = toDate('2026-07-07')), 3) AS jul7_avg_spread_cents,
    arrayCount(x -> x < anyIf(avg_spread_cents, d = toDate('2026-07-07')), groupArrayIf(avg_spread_cents, d != toDate('2026-07-07'))) + 1 AS tightness_rank,
    count() AS sessions_compared,
    round(min(avg_spread_cents), 3) AS tightest_session_cents,
    round(max(avg_spread_cents), 3) AS widest_session_cents,
    toString(min(d)) AS first_session,
    sum(dropped_invalid) AS dropped_invalid_quotes
FROM (
    SELECT toDate(toTimeZone(sip_timestamp, 'America/New_York')) AS d,
           avgIf(toFloat64(ask_price - bid_price), bid_price > 0 AND ask_price >= bid_price) * 100 AS avg_spread_cents,
           countIf(NOT (bid_price > 0 AND ask_price >= bid_price)) AS dropped_invalid
    FROM global_markets.cache_stocks_quotes
    WHERE ticker = 'SPY'
      AND sip_timestamp >= toDateTime('2026-06-05 00:00:00')
      AND sip_timestamp < toDateTime('2026-07-08 00:00:00')
      AND (toHour(sip_timestamp) * 60 + toMinute(sip_timestamp)) BETWEEN 810 AND 1199
    GROUP BY d
)

以及本節每個交易時段都會進行的例行檢查：按更新量加權平均報價價差計算，7 月 7 日是 SPY 過去一個月內價差最窄的交易日——1.809 美分，在 21 個交易日中排名第 1，而該月份最寬時曾達到 2.865 美分。一個二比一的下跌磁帶，卻以該月最窄的指數價差定價，這是一個只有報價串流才能釐清的事實。

查詢選擇權NBBO報價流：7月7日 相較 7月6日，以及相較同個交易日的股票報價
每個數據背後的精確 SQL 語法
WITH
    (
        SELECT count()
        FROM global_markets.cache_options_quotes
        WHERE toYYYYMMDD(sip_timestamp) = 20260707
    ) AS jul7_option_rows,
    (
        SELECT count()
        FROM global_markets.cache_options_quotes
        WHERE toYYYYMMDD(sip_timestamp) = 20260706
    ) AS jul6_option_rows,
    (
        SELECT count()
        FROM global_markets.cache_stocks_quotes
        WHERE toYYYYMMDD(sip_timestamp) = 20260707
    ) AS jul7_stock_quote_rows
SELECT
    jul7_option_rows AS jul7_option_quote_rows,
    round(jul7_option_rows / 1e9, 2) AS jul7_option_quote_updates_bn,
    round(jul6_option_rows / 1e9, 2) AS jul6_option_quote_updates_bn,
    round((jul7_option_rows / jul6_option_rows - 1) * 100, 1) AS day_over_day_pct,
    round(jul7_stock_quote_rows / 1e6, 2) AS jul7_stock_quote_updates_m,
    round(jul7_option_rows / jul7_stock_quote_rows, 1) AS jul7_option_to_stock_ratio

選擇權報價串流規模更大，是這個資料庫中最大的數據集。7 月 7 日承載了 8.69 十億次選擇權 NBBO 更新，是同一交易時段整個股票報價紀錄（492.76 百萬次更新）規模的 17.6 倍。週一的選擇權紀錄有 6.49 十億次，因此選擇權串流環比擴張了 34%，與此同時股票串流擴張了 25.7%。

查詢SPY 7月7日選擇權報價：全系列，以及近價平當日到期合約於空氣缺口中觸及價差 對比 午間對照時段
每個數據背後的精確 SQL 語法
WITH
    (
        SELECT count()
        FROM global_markets.cache_stocks_quotes
        WHERE toYYYYMMDD(sip_timestamp) = 20260707
    ) AS jul7_stock_quote_rows,
    (
        SELECT round(toFloat64(avg(close)), 2)
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'SPY'
          AND window_start >= '2026-07-07 14:36:00' AND window_start < '2026-07-07 14:48:00'
    ) AS spy_pocket_price,
    (
        SELECT round(toFloat64(avg(close)), 2)
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'SPY'
          AND window_start >= '2026-07-07 17:30:00' AND window_start < '2026-07-07 17:42:00'
    ) AS spy_midday_price
SELECT
    round(count() / 1e6, 2) AS spy_option_updates_m,
    round(100.0 * count() / jul7_stock_quote_rows, 1) AS pct_of_whole_equity_quote_tape,
    uniqExact(ticker) AS spy_contracts_quoted,
    round(countIf(substring(ticker, -15, 6) = '260707') / 1e6, 2) AS same_day_updates_m,
    spy_pocket_price AS spy_price_in_pocket,
    spy_midday_price AS spy_price_midday,
    round(quantileExactIf(0.5)(toFloat64(ask_price - bid_price),
        bid_price > 0 AND ask_price >= bid_price
        AND substring(ticker, -15, 6) = '260707'
        AND abs(toFloat64(toUInt32OrZero(substring(ticker, -8))) / 1000 - spy_pocket_price) <= 2
        AND sip_timestamp >= '2026-07-07 14:36:00' AND sip_timestamp < '2026-07-07 14:48:00'), 3) AS pocket_atm_spread_dollars,
    round(quantileExactIf(0.5)(toFloat64(ask_price - bid_price),
        bid_price > 0 AND ask_price >= bid_price
        AND substring(ticker, -15, 6) = '260707'
        AND abs(toFloat64(toUInt32OrZero(substring(ticker, -8))) / 1000 - spy_midday_price) <= 2
        AND sip_timestamp >= '2026-07-07 17:30:00' AND sip_timestamp < '2026-07-07 17:42:00'), 3) AS midday_atm_spread_dollars,
    round(pocket_atm_spread_dollars - midday_atm_spread_dollars, 3) AS atm_spread_delta_dollars,
    round(countIf(substring(ticker, -15, 6) = '260707'
        AND abs(toFloat64(toUInt32OrZero(substring(ticker, -8))) / 1000 - spy_pocket_price) <= 2
        AND sip_timestamp >= '2026-07-07 14:36:00' AND sip_timestamp < '2026-07-07 14:48:00') / 1e3, 1) AS pocket_atm_updates_k,
    round(countIf(substring(ticker, -15, 6) = '260707'
        AND abs(toFloat64(toUInt32OrZero(substring(ticker, -8))) / 1000 - spy_midday_price) <= 2
        AND sip_timestamp >= '2026-07-07 17:30:00' AND sip_timestamp < '2026-07-07 17:42:00') / 1e3, 1) AS midday_atm_updates_k,
    round(pocket_atm_updates_k / midday_atm_updates_k, 1) AS atm_quote_rate_ratio,
    countIf(NOT (bid_price > 0 AND ask_price >= bid_price)
        AND substring(ticker, -15, 6) = '260707') AS dropped_invalid_same_day_quotes
FROM global_markets.cache_options_quotes
WHERE ticker >= 'O:SPY26' AND ticker < 'O:SPY27'
  AND sip_timestamp >= '2026-07-07 13:30:00' AND sip_timestamp < '2026-07-07 20:00:00'
HAVING pocket_atm_updates_k > 0 AND midday_atm_updates_k > 0

光是 SPY 的 2026 年到期合約，在常規交易時段內就佔了其中 327.55 百萬次更新，分佈在 9116 檔掛牌合約中。這相當於同一天整個股票市場報價量的 66.5%，僅來自一個標的。那些構成0DTE 紀錄的當日到期合約，則佔了其中的 13.61 百萬次。

在早盤那十二分鐘的真空期內，近價平的當日到期合約報出的中位數價差為 $0.01，與它們在日中對照組中的報價（$0.01，差異為 0）完全一致。同一組合約的報價頻率則達到 3 倍：低點時有 191.1 千次更新，而午餐時段為 64 千次。選擇權的價差表現與股票價差一致。在報價機制高速運轉時，它維持住了自身的寬度。近價平意指在相同時間窗口內，履約價在 SPY 平均價格的 2 美元以內，且該面板同時列印了兩個價格（真空期為 746，日中為 749.08）。

利率：殖利率曲線隨成長行情走跌

查詢美國公債殖利率曲線：7月7日 與 7月6日 數據比較（僅包含有報價的期限）
每個數據背後的精確 SQL 語法
SELECT
    t.1 AS curve_point,
    round(t.2, 2) AS jul7_yield_pct,
    round((t.2 - t.3) * 100) AS session_change_bp
FROM (
    SELECT arrayJoin([
        ('1 month',  toFloat64(d.yield_1_month),  toFloat64(p.yield_1_month)),
        ('3 month',  toFloat64(d.yield_3_month),  toFloat64(p.yield_3_month)),
        ('1 year',   toFloat64(d.yield_1_year),   toFloat64(p.yield_1_year)),
        ('2 year',   toFloat64(d.yield_2_year),   toFloat64(p.yield_2_year)),
        ('5 year',   toFloat64(d.yield_5_year),   toFloat64(p.yield_5_year)),
        ('10 year',  toFloat64(d.yield_10_year),  toFloat64(p.yield_10_year)),
        ('30 year',  toFloat64(d.yield_30_year),  toFloat64(p.yield_30_year)),
        ('2s10s spread', toFloat64(d.yield_10_year - d.yield_2_year), toFloat64(p.yield_10_year - p.yield_2_year))
    ]) AS t
    FROM (SELECT * FROM global_markets.treasury_yields WHERE date = '2026-07-07') AS d,
         (SELECT * FROM global_markets.treasury_yields WHERE date = '2026-07-06') AS p
)

美國公債隨成長行情走跌。1年期上揚 11 個基點至 4.06%，10年期 7 至 4.55%，30年期 6 至 5.05%——突破5%關卡。2年10年期利差維持在 0.36 個百分點（變動 1 個基點）：走勢大致平行，並非曲線型態重塑。

這一天背後的日曆

查詢7月7日公司行事曆與資訊流，單列摘要
每個數據背後的精確 SQL 語法
WITH
    (
        SELECT (count(), uniqExact(publisher))
        FROM global_markets.stocks_news
        WHERE toDate(toTimeZone(published_utc, 'America/New_York')) = '2026-07-07'
    ) AS news,
    (
        SELECT (argMax(t, n), max(n), max(n) - arraySort(x -> -x, groupArray(n))[2])
        FROM (
            SELECT t, count() AS n
            FROM (
                SELECT arrayJoin(tickers) AS t
                FROM global_markets.stocks_news
                WHERE toDate(toTimeZone(published_utc, 'America/New_York')) = '2026-07-07'
            )
            WHERE t != 'SPCX'
            GROUP BY t
        )
    ) AS top_news
SELECT
    (SELECT count() FROM global_markets.stocks_dividends WHERE ex_dividend_date = '2026-07-07') AS ex_dividend_records,
    (SELECT count() FROM global_markets.stocks_splits WHERE execution_date = '2026-07-07') AS splits_executed,
    (SELECT countIf(toFloat64(split_from) > toFloat64(split_to)) FROM global_markets.stocks_splits WHERE execution_date = '2026-07-07') AS reverse_splits,
    (SELECT countIf(toFloat64(split_to) > toFloat64(split_from)) FROM global_markets.stocks_splits WHERE execution_date = '2026-07-07') AS forward_splits,
    (SELECT count() FROM global_markets.stocks_ipos WHERE listing_date = '2026-07-07') AS ipos_listed,
    (SELECT uniqExact(accession_number) FROM global_markets.stocks_sec_edgar_index WHERE filing_date = '2026-07-07') AS sec_filings,
    (SELECT uniqExactIf(accession_number, form_type = '4') FROM global_markets.stocks_sec_edgar_index WHERE filing_date = '2026-07-07') AS insider_form4_filings,
    (SELECT uniqExactIf(accession_number, form_type = '8-K') FROM global_markets.stocks_sec_edgar_index WHERE filing_date = '2026-07-07') AS filings_8k,
    news.1 AS news_articles,
    news.2 AS news_publishers,
    top_news.1 AS most_covered_ticker,
    if(top_news.1 = 'NVDA', 1, 0) AS most_covered_is_nvda,
    top_news.2 AS most_covered_articles,
    top_news.3 AS most_covered_lead_over_next,
    (SELECT countIf(has(tickers, 'CRNX')) FROM global_markets.stocks_news
     WHERE toDate(toTimeZone(published_utc, 'America/New_York')) = '2026-07-07') AS crnx_articles

79 筆股息記錄進入除息日，7 筆股票分割執行——其中 5 筆為反向分割，這種低於一美元的合股操作，在未調整的行情記錄上偽造出三位數的「漲幅」，對比 2 筆為正向分割——另有 1 檔新上市股票掛牌。SEC 索引記錄了 2433 份申報文件：463 份表格 4，173 份 8-K。我們的新聞來源共刊出 215 篇文章，來自 3 家發布機構；報導量最高的個股是 NVDA，共 17 篇文章，比第二名多出 4 篇——正是那檔在其所屬板塊收跌之日仍收漲的 NVDA。新聞來源刊出 0 篇關於 CRNX 的文章，該股是當日漲幅最大的股票——這是一家新聞來源的關注，而非全球媒體的焦點。

盤勢確認

查詢交易時段檢查：SPY 觀察到的分鐘K線跨度、假期表，以及下一個日曆上的休市日
每個數據背後的精確 SQL 語法
SELECT
    formatDateTime(min(toTimeZone(window_start, 'America/New_York')), '%H:%i') AS first_spy_bar_et,
    formatDateTime(max(toTimeZone(window_start, 'America/New_York')), '%H:%i') AS last_spy_bar_et,
    count() AS spy_minute_bars,
    countIf(window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00') AS regular_session_bars,
    uniqExactIf(toDate(toTimeZone(window_start, 'America/New_York')), window_start >= '2026-07-07 13:30:00' AND window_start < '2026-07-07 20:00:00') AS day_sessions,
    (SELECT count() FROM global_markets.stocks_market_holidays WHERE date = '2026-07-07') AS jul7_holiday_rows,
    (SELECT toString(min(date)) FROM global_markets.stocks_market_holidays WHERE date > '2026-07-07' AND status = 'closed') AS next_closure_date,
    (SELECT argMin(name, date) FROM global_markets.stocks_market_holidays WHERE date > '2026-07-07' AND status = 'closed') AS next_closure_name
FROM global_markets.delayed_stocks_minute_aggs
WHERE ticker = 'SPY' AND window_start >= '2026-07-07 00:00:00' AND window_start < '2026-07-08 00:00:00'

7月7日為完整交易時段，並非提早收盤：SPY 的 K 棒從紐約時間 04:00 延續至 19:59，正規交易時段內恰好有 390 根 K 棒，且假日表對應該日期共有 0 筆資料。下一個排定的休市日為 Labor Day，2026-09-07。

資料說明

  • 成交金額為每分鐘代理指標 — 收盤價 × 成交量，按每分鐘K棒加總，僅計入一般交易時段。
  • 跳空與盤中走勢的拆分，係以各標的當日第一根一般交易時段K棒（開盤價）與前一交易日最後一筆一般交易時段收盤價的差距來衡量；其餘部分則為7月7日開盤至收盤的變動。
  • 同步出現的早盤低點，在撰寫當下已與相鄰K棒交叉比對 — 每個低點皆由其前後數分鐘的走勢所確認，而非單一孤立報價；無任何一筆需要以異常報價單處理。若極端價格出現在超過一分鐘的K棒中，面板顯示的是最早出現該價格的K棒（以確定性方式解決平手問題）。
  • 一檔重複使用代碼的掛牌標的，在實體驗證完成前，已自成交量排行榜與新聞提示中排除；其獨立文章載有相關憑證。
  • 報價價差統計數據排除無效報價（單邊報價或交叉的NBBO紀錄），且每個價差面板會以獨立欄位計數其所剔除的筆數。「更新加權平均價差」係將每次有效NBBO更新的報價寬度加以平均 — 交易活躍時段權重較高；此為報價統計指標，而非每筆交易的實際成本。
  • 7月7日的選擇權報價在此篇文章首次發布後才送達。 cache_options_quotes 承載了資料倉儲最長的接收延遲，因此第一版在目前選擇權報價面板所在位置，交付的是一份有界的零列收據。資料列送達後，零列邊界在重新生成時失效，本次修訂版讀取了該交易時段自身的 8.69 億筆更新。股票報價磁帶從一開始就是完整的：其最後一筆7月7日更新列印於美東時間分鐘 1199，在最後一個一般交易小時內有 77.13 百萬筆更新（收據欄位位於報價磁帶面板中）。
  • 選擇權報價寬度係於價平附近衡量。 對所有掛牌的同一日到期合約，取每次更新的中位數，此為一個組成統計量：深度價內合約的報價以美元計的價差很寬，深度價外合約的報價則僅有幾美分，且哪些合約重新報價會在交易時段中發生變化。開盤區間與午盤的比較，僅限於在每個時間窗口內，履約價落在SPY平均價格正負2美元以內的合約，且兩個價格皆為面板中的欄位。無效的選擇權報價（單邊或交叉）會在該面板中計數，而非直接無聲地剔除。
  • 7月7日的公債報價延遲了一個交易時段才接收 — 第一版交付的是一份已揭露延遲的收據；其觸發機制於7月8日報價送達時啟動，本次修訂版則載有實際的殖利率曲線。

方法說明

  • 期間為單一交易時段（1 時段，已透過觀察到的 K 棒與假日曆驗證——絕非假設）。時間戳記以 UTC 儲存，並在查詢內部轉換為紐約時間。「收盤」指常規交易時段最後一分鐘的 K 棒；日別比較為 7 月 7 日對比 7 月 6 日，即連續兩個交易時段。
  • 小數欄位在進行比率運算前，會先轉換為 64 位元浮點數；選擇權到期日則從 OCC 代碼重新解析（表格本身的到期日欄位有誤）。所有面板均在撰稿時，透過設有門禁的唯讀路徑讀取一次。資料倉儲狀態截至 2026 年 7 月 8 日。
  • 報價磁帶區塊於 7 月 8 日修訂版新增（該修訂版同時將公債延遲回報替換為實際列印值），其選擇權報價面板在 7 月 7 日當天的選擇權 NBBO 資料列抵達後重新編寫。此產出在設計上為批次作業：SPY 選擇權報價面板掃描數億筆 NBBO 紀錄，而磁帶規模計數則直接讀取當日分割區的資料列總數，這正是儲存每時段數十億列資料的表格能在不到一秒內回應計數查詢的原因。

每個面板都是一個儲存的查詢結果——圖表、表格與 SQL 為同一物件。將任一查詢貼入 Strasmore 終端機，即可自行運用。前一交易時段：7 月 6 日。

#recap#daily#spy#qqq#market structure