After a steep one-day fall, does a low-priced share rebound overnight beyond its bid-ask spread?
Study 35 · Changed a rule or plan · recorded 3 October 2026 · answer: Not at the size tested · full paper with figures
The question
An earlier study found that, for shares priced 1.00 to under 10.00 in US currency, the rebound after one-day falls of 10% or more comes between the close and the next opening. We ask whether that overnight step exceeds the share's usual one after a down day plus half its bid-ask spread, the gap between buyers' and sellers' prices.
Rules fixed before the result
We fixed two tests before computing any outcome: does the overnight step exceed what the usual down-day step plus half a spread can produce (its 95% range above zero), and does it grow with the fall (a slope at least twice its standard error)? An excess that grows would mean selling pressure relieved overnight; without growth, some other cause. A range reaching zero but under 0.50 percentage points, without growth, would mean a quoting effect; anything else leaves it open.
Data and method
The events are 10,041 one-day falls of at least 10% on 2,003 days through 2022. The outcome is the overnight step to the next opening, less the market's step, the share's usual step after down days in the prior year, and half its spread, estimated from daily highs, lows and closes in the prior month. Each day's falls are averaged first so days count equally; growth is the slope on the fall's size at a fixed spread.
The answer
The overnight step averaged about 1.10 percentage points, and the share's usual overnight step after a down day (0.19) and half a spread (0.95) account for all of it: what remained was about 0.04 points below zero, 95% range 0.25 below to 0.16 above, well under the 0.50-point bar fixed in advance; we had expected more, with low confidence. Growth with the fall was not detected; the test passed its advance power check but proved far less precise than assumed, with little chance of finding the growth it targeted.
Main estimate
-0.000 fraction (95% range -0.002 to 0.002), from 10041 observations
What it changes
It changed a plan. The overnight part of the earlier reversal is now recorded as a quoting effect for these shares, and that line of work rests. Whether the rebound grows with the size of the fall stays open: none was detected, and although the test passed the power check made in advance, it proved far less precise than that check assumed, so it had little chance of detecting the growth it targeted.
What remains uncertain
The data end in 2022; the spread is estimated, not quoted, and can miss a widening that day; the opening price is a trade at either side of the quote. Two side checks decided nothing: beyond the usual step and half a spread, falls closing in the day's bottom tenth were 0.60 points above (95% range 0.34 to 0.86) and the rest 0.62 below (0.42 to 0.81 below); the smallest third of companies, 0.56 above (0.00 to 1.11).
How it was made
Run took 31 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan 8c1691614d5c5e3f79db85afb6749ca6b10eb5c8e28d5f455522ea8ef3d4f1e9 result dc917e87270cebaad180770936adb87277e302422243fe3e791a60b2b7c1df0a
Exploratory work on development data; not a performance claim.