Statistics Corpus research · single-day dumps
Anatomy of the One-Day Dump
~34 months of history · 200,000+ tapes analyzed · researched at 2026-08-31 · dump = high→low fall ≥20%, low printed after the high
A stock gets promoted for weeks or months, then gives most of it back in a single session.
This report measures who those stocks are, how they were pumped, how fast the collapse runs,
and what happens afterwards — across every 10% of drop depth.
40,365
dump days found — 1 in 5 of all archived volatile tapes
3,105
deep dumps (−50% or worse) — rare: 1.5% of all tapes
62.5%
of −90% dumps are Chinese issuers, vs a 17% base rate
−72% more
median further loss 60 trading days after a −90% dump. They don't come back.
01 · Frequency
Shallow dumps are routine. Deep ones are rare and specific.
A 20–30% intraday fade off the high happens to roughly one in five volatile small-caps —
it is the ordinary texture of this market. Each extra 10% of depth cuts the population roughly in
half. By −80% only a couple hundred events exist in ~34 months of data, and at −90%
just 80. Everything below shows that these deep events are not "more of the same" — they are unique with a recognizable profile.
Dump days by depth bucket
Depth = fall from day high to day low. d20 = 20–30%, … d90 = 90%+. Full statistics corpus, Nov 2023 – Aug 2026.
02 · Who they are
The deeper the dump, the more Chinese it gets
Chinese issuers make up 16.8% of the non-dump control population. Among shallow
dumps they are barely over-represented (20%). But their share climbs monotonically with depth, and
in the −90% bucket 50 of the 80 events are Chinese names. Deep dumps also
carry the manipulation fingerprint regardless of country: no same-day news for about half of them,
trading halts on most, SSR triggered on nearly all.
Share of Chinese issuers by dump depth
Dashed line = China's share of the non-dump control population (16.8%).
The manipulation fingerprint grows with depth
Share of dump days with each trait, by depth bucket.
03 · The pump before the crash
They crash from the top of a months-long ramp
Deep dumpers do not fall out of nowhere. The median −80/−90% name enters its crash day
after a +700–730% run over the prior trading year (Chinese −90% names:
+1,246%), sitting at or near its yearly high, up 15–44% over just the last 20
days, on a rising streak of green closes. The grind is unusually orderly — 51% of the last 60
days closed green (vs 44% for shallow dumps) — and volume builds into the top (recent volume
2.6× its earlier average for d90). The final vertical leg is short: the deepest names spent a
median of only 18–38 trading days above half of their crash-day high.
Long quiet accumulation, a fast blow-off, then the rug.
What the year before the dump looked like (medians)
From split-adjusted daily bars, ~250 trading days before each dump day. "Days above ½ peak" is capped at 250.
The −90% Chinese cohort in one sentence: up ~12× in a year,
green 51% of days, 20-day gain +70%, volume building 3.5×, roughly zero same-day news —
and then it loses 90% in one session.
04 · The day itself
High in the morning, destruction by the close
Across all depths the day high is set early — median around 7:00–7:30 ET, with
72–80% of highs printed in premarket. The low lands late: for −90% events the median low
prints around 15:15–15:40 ET. The collapse is fast where it matters: a −90% dump is
already down 50% about 2.4 hours after the high and down 80% in under 3 hours;
the last leg to −90% takes longer because of repeated volatility halts (78% of d90 events
halt at least once). After the low there is essentially no bounce — the median −90%
event recovers just 3% of its fall range before the session ends.
Median minutes from the day high to each drop level
One line per final depth bucket. Read: a d90 event needed a median of 142 minutes to be down 50%.
05 · The aftermath
There is no recovery trade
Whatever the depth, the dump day is not the bottom — roughly 9 in 10 names
print a new low within the next 60 trading days. Returns keep bleeding, and the deeper the dump,
the worse the bleed: 60 days after a −90% event the median name is down another
72% from the dump-day close (Chinese names: −82%). "Recovered half the fall within 60
days" happens for 86% of shallow d20 fades — those are ordinary volatility — but for only
11% of −90% dumps, and 4% of the Chinese ones. Once the rug is pulled, the story is over.
Median return 60 trading days later
Close-to-close vs the dump-day close, split-adjusted.
Share that recovered half the fall within 60 days
Any day whose high reached the midpoint of the dump-day range.
06 · IPO & underwriters
The $4 IPO pipeline is real
For every symbol with a dump of −40% or worse (1,703 symbols; a CIK was found for all of
them), we pulled the IPO date, IPO price and underwriters from SEC EDGAR. A quarter of these
companies went public at about $4 — the Nasdaq listing minimum and the classic
small-float China-IPO price point — and one in five dumped within its first year as a
public company.
$4.13
25th-percentile IPO price (median $8.50) — the "$4 IPO" tier is a quarter of all deep dumpers
19.9%
of deep dumps happen within 365 days of the IPO
560 days
25th-percentile time from IPO to dump (median ~4 years — the tail is old listed names)
Underwriters split into two different worlds. The boutique circuit that
brings $4 micro-caps public — Revere, D. Boral, Prime Number, Bancroft, Craft, WestPark,
US Tiger, AC Sunshine — has heavy Chinese concentration and sees its IPOs dump
within months. Bulge-bracket names (Goldman, Deutsche) also appear in the list, but
their IPOs dump after a median of ~5–7 years with no China tilt — those are ordinary
biotech and growth blow-ups, not pump schemes.
Underwriters of deep dumpers (≥40% dump), cleaned list
Names are parsed heuristically from IPO prospectuses — generic cover-page phrases were removed by hand; counts are a floor, not a census. "Days to dump" = median days from IPO to the symbol's deepest dump.
07 · Trend
Dumps aren't getting more common — but they're getting more Chinese
The share of volatile tapes that end in a dump has been flat at 18–22% for ~34 months.
What moved is the mix: the Chinese share of dump days drifted from ~15–20% in 2024 to
25–32% through 2026, peaking at 32.4% in June 2026 — the same month
that set the record for deep dumps (244 events of −50% or worse, versus ~40/month when the
corpus starts in late 2023).
Monthly: dump rate and Chinese share of dumps
Both lines are shares, same axis. Dump rate = dumps ÷ all archived tapes that month.
Dump rate (share of tapes)
China share of dumps