Market data quality checks that catch real bugs
Declared split ratio against the close ratio across the effective dateranking ·
2026-10-07 · 10×3
One symbol's quote messages against the sequence numbers they spanseries ·
2026-10-07 · 60×3
Condition codes on one hour of prints, named from the exchange reference listranking ·
2026-10-07 · 9×3
Venue stamp to consolidated stamp, minute by minuteseries ·
2026-10-07 · 60×3
Locked, crossed and one sided quote rates across six household namesranking ·
2026-10-07 · 6×4
Declared split ratio against the close ratio across the effective date
Declared split ratio against the close ratio across the effective date
| split_label | declared_ratio | close_ratio |
|---|---|---|
| DLLL Jun 26, 2026 | 8 | 1.08 |
| INTW Jun 26, 2026 | 8 | 1.09 |
| MVLL Jun 26, 2026 | 3 | 1.12 |
| MULL Jun 26, 2026 | 25 | 1.17 |
| NVDL Jun 26, 2026 | 3 | 1.04 |
| SMCL Jun 26, 2026 | 3 | 1.07 |
| LILAK Jun 17, 2026 | 1.1 | 1.35 |
| LILA Jun 17, 2026 | 1.1 | 1.34 |
| KLAC Jun 12, 2026 | 10 | 0.95 |
| SNDU Jun 8, 2026 | 3 | 0.9 |
the exact SQL behind every number
WITH splits AS
(
SELECT
ticker,
execution_date,
any(split_from) AS from_shares,
any(split_to) AS to_shares
FROM global_markets.stocks_splits
WHERE execution_date >= '2024-06-01'
AND execution_date <= '2026-06-30'
AND split_to > split_from
AND ticker NOT IN ('SPCX')
GROUP BY ticker, execution_date
)
SELECT
concat(s.ticker, ' ', formatDateTime(s.execution_date, '%b %e, %Y')) AS split_label,
round(s.to_shares / s.from_shares, 2) AS declared_ratio,
round(
argMaxIf(toFloat64(d.close), d.date, d.date < s.execution_date)
/ argMinIf(toFloat64(d.close), d.date, d.date >= s.execution_date), 2
) AS close_ratio
FROM global_markets.stocks_daily_aggs AS d
INNER JOIN splits AS s ON s.ticker = d.ticker
WHERE d.date >= '2024-05-20'
AND d.date <= '2026-07-10'
AND d.date >= s.execution_date - 7
AND d.date <= s.execution_date + 7
GROUP BY s.ticker, s.execution_date, s.from_shares, s.to_shares
HAVING countIf(d.date < s.execution_date) > 1
AND countIf(d.date >= s.execution_date) > 1
AND max(d.volume) > 2000000
ORDER BY s.execution_date DESC
LIMIT 10
More from this analysisMarket data quality checks that catch real bugs
Condition codes on one hour of prints, named from the exchange reference list
ranking 9×3
→
Locked, crossed and one sided quote rates across six household names
ranking 6×4
→
One symbol's quote messages against the sequence numbers they span
series 60×3
→
Venue stamp to consolidated stamp, minute by minute
series 60×3
→
See all 3,171 queries →