locked market — 4 answered market questions

Whole-tape census: locked, crossed, and one-sided records across every NBBO update of the session

most recentas of 2026-08-22 scalar 1×7read in context →

session date

2026-08-18

updates m

406

locked m

7.41

crossed k

43.8

locked per crossed

169

locked or crossed per 10k

183.4

one sided per 10k

3.2

the exact SQL behind every number

WITH (
    SELECT max(toDate(toTimeZone(window_start, 'America/New_York')))
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker = 'SPY' AND window_start < toDateTime(today() - 3)
) AS census_day
SELECT
    formatDateTime(census_day, '%Y-%m-%d') AS session_date,
    round(count() / 1e6, 0) AS updates_m,
    round(countIf(bid_price = ask_price AND bid_price > 0) / 1e6, 2) AS locked_m,
    round(countIf(bid_price > ask_price AND ask_price > 0) / 1e3, 1) AS crossed_k,
    round(countIf(bid_price = ask_price AND bid_price > 0) / toFloat64(greatest(countIf(bid_price > ask_price AND ask_price > 0), 1)), 0) AS locked_per_crossed,
    round(countIf(bid_price >= ask_price AND ask_price > 0 AND bid_price > 0) / toFloat64(count()) * 1e4, 1) AS locked_or_crossed_per_10k,
    round(countIf(bid_price <= 0 OR ask_price <= 0) / toFloat64(count()) * 1e4, 1) AS one_sided_per_10k
FROM global_markets.cache_stocks_quotes
WHERE sip_timestamp >= toDateTime(census_day)
  AND sip_timestamp < toDateTime(census_day + 1)

More from this analysisLocked and Crossed Markets, Explained With Data

The receipts: locked vs crossed totals, and the rate by session stretch (premarket, open, noon, close, ET) scalar 1×10 → Locked-or-crossed records per 10,000 updates by 30-minute ET bucket, checked set, extended hours included series 32×3 → The census by name: eight liquid tickers and two thin small caps, recent completed sessions table 10×7 → AAPL: the last recorded NBBO quote in our data window scalar 1×7 → See all 2,170 queries →