After a large one-day fall, do shares recover more over three sessions if no company filing explains it?
Study 55 · Changed a rule or plan · recorded 14 October 2026 · answer: Yes · full paper with figures
Correction, 19 October 2026. This reading was decided by the fallback measure, which pulls in the most extreme 1% of values at each end. Its false-alarm rate was checked on this study's own design and stayed within the limit the registration allows, and the plain measure gives the same answer. The reading stands.
Rules fixed before the result
Fixed in advance: at least 2,000 falls, 300 trading days and 500 companies of each kind, 500 days holding both kinds in one price range, and a comparison able to detect one percentage point; it could detect 0.69. The plain average was to decide unless its range was wider than 1.4 points, when the measure with extremes pulled in would; the range was 2.9 points wide, so that measure decided, and no minimum or bar was changed.
The answer
Yes. Over the three sessions after a fall, against the market, falls with no explaining filing recovered 1.6 percentage points more than explained falls on the same day in the same price range (95% range 1.1 to 2.1), with the most extreme 1% at each end pulled in, as planned once the plain average's range proved too wide (2.5 points, 1.1 to 4.0); the whole range lies above the one point set in advance as the size that matters. Unexplained falls gained 0.5% (0.1 to 0.9) and explained falls lost a further 1.0% (0.7 to 1.3). The difference was 0.7 points (0.1 to 1.3) from prices of 10 or more, 2.1 (1.5 to 2.7) from between 1 and 10, and 1.6 (1.1 to 2.1) counting filings two to five days late. Of all falls, 56% had none.
Main estimate
0.016 fraction (95% range 0.011 to 0.021), from 16957 observations
What it changes
By the rule set in advance, falls that no filing explains gain a recorded expectation of a short recovery, proposed to the owner and never a trade, and a study of why, from the next opening price and by trading volume, becomes the next question. The plan expected, with low confidence, that unexplained falls would recover about one point more.
What remains uncertain
These falls were studied before, though never split by filings. Falls on the same day can share the same news, which the 95% ranges allow for by resampling whole days; one company's repeated falls are allowed for only in a check, which showed no contradiction. Leaving out the first session, the difference was 0.5 points (95% range 0.2 to 0.9), wholly below the one-point size set in advance, so most of the difference may come from that session.
How it was made
Run took 75 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan 8162b3c4c050c098a19d0ae1f9e536d5a44212d741f5d87cb081c0f391e1f21e result 66e8a01bb82ff9598828cc8fde0e9a6256965d810c04abd83f3d3c0f0af1108d
Exploratory work on development data; not a performance claim.