Run again: among a day's biggest fallers, do shares closing at a year's low rank lower the next day?
Study 91 · No reading (stopped by its own checks) · recorded 8 November 2026 · answer: Not settled · full paper with figures
The question
Among the day's biggest fallers priced at five or more in US currency with a usable next day, on days when at least five closed at or below a year's low and at least fifty did not, does the next day's return rank lower for those at the low, within fallers of similar volume, in 2018 to 2022?
Rules fixed before the result
Fixed before any outcome: the deciding measure, with twelve thousandths of the ranking as the smallest gap worth reporting; a session counted only when five or more of the day's biggest fallers closed at a year's low, with eight as the fallback; a power check on the number of such sessions; and a check that every figure of the earlier attempt's record on 2014 to 2017 came out the same, any difference stopping the study.
Data and method
Each session we took the fifth of shares falling most against the market fund, adjusted the next day's return by a straight-line fit for the day's fall and the close's place in its range, and ranked it among fallers at a similar multiple of their usual volume. The measure was the other fallers' average rank less that of those closing at a year's low. The plan rebuilt it on 2014 to 2017 and checked its false-alarm rate by resampling.
The answer
Not settled: the study stopped at its planned checks, before computing any return in 2018 to 2022, so nothing is shown for those years. It runs the earlier attempt's test again after that one stopped for want of qualifying days, with one change: the smallest gap worth reporting rose from ten to twelve thousandths of the ranking. That left enough qualifying sessions to open the test, but the plan first required every figure of the earlier attempt's checks on 2014 to 2017 to come out the same. One did not: the chance that any of the study's descriptive notes fires by luck alone, which moves with the gap because several notes are judged against it, and which matched with the old gap put back; by the rule fixed in advance, that stopped the study.
What it changes
No plan or published answer changed. The stop came from the plan's own wording: it listed as fixed a figure that the larger gap enters, and two tests on simulated sessions, of the fallback and of the stop for too few sessions, were sized for the old gap and fail at the new one. The plan says a second stop of this design ends the programme's test of the low-side condition; a rerun needs a new plan correcting all three.
What remains uncertain
No return after 2017 was measured, so neither a gap nor its absence is shown for 2018 to 2022. Those years were used by earlier studies of these shares, so even a measured answer would have been a check within our development data, not a test on fresh data. Daily prices also cannot tell news from selling pressure, so a measured gap would not have named its source.