Does a share that beat similar shares rank lower in the next day's trading hours, apart from where it closed?
Study 85 · Changed a rule or plan · recorded 4 November 2026 · answer: Partly · full paper with figures
The question
Holding the prior week's move, does the top fifth by the day's gap to similar shares (each share's fifteen closest past-year trackers) rank below the bottom fifth on next-day trading-hours returns ranked 0 to 1, by a 95% range reaching at least 0.007 below zero, and stay below with the close's place between the day's low and high held?
Rules fixed before the result
Fixed in advance: yes only if the difference's 95% range, widened for false alarms, lies wholly below zero and reaches 0.007, the lower rank survives holding where the close sat in the day's range, and checks on the prior five sessions' move and on shares missing the next day's prices pass. Below zero but short of 0.007, or with the position question unresolved, gives partly. At least 1,000 sessions and a passed power check were required.
Data and method
Each session, shares at five or more in US currency at the unadjusted close are ranked on next-day returns from opening trade to close. Within thirds of the prior week's move, the top fifth by the gap to similar shares is set against the bottom fifth, averaged over sessions with a 95% range allowing for correlated nearby sessions. Resampled 2013 to 2017 sessions with no effect built in set the range's widening against false alarms and each reading's chance rate.
The answer
Partly. Over 1,258 sessions from 2018 to 2022, shares in the top fifth by the gap to their similar shares ranked 0.0045 lower than the bottom fifth on the next day's return from the opening trade to the close, ranked against the other shares that day (95% range 0.0004 to 0.0086 on a scale of 0 to 1), a range that includes the 0.007 set in advance. With where the close sat between the day's low and the day's high held, the link stayed below zero, but on earlier years a result like it arose by chance too often to count, so that question is not resolved. The lower rank was clear in shares whose own daily returns over the past year had tended to reverse, and within 0.007 of zero in those whose returns had tended to continue.
Main estimate
-0.004 fraction (95% range -0.009 to -0.000), from 1258 observations
What it changes
Under the rule fixed in advance, the lower next-day rank is recorded as present with its source left open; no note for the record is proposed and no screen changes. Because it was clear only in shares whose returns had tended to reverse, later reversal tests hold that past-year tendency as a separate group beside where the close sat. No trading decision changes.
What remains uncertain
Earlier studies used these years, so this is a check inside our development data, not an independent test. Descriptive readings not reported here (parts of the day, trading value, similar shares alone, the sorts, years) were each read alone: with no effect beyond the earlier years' averages, at least one shows a pattern in 44 percent of resampled runs. The position and past-return readings each had a 1 percent chance limit, met by the past-return reading only.
How it was made
Run took 118 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan 5777050e4e1ffd74d1a3315ee67a28a2e0012b27c53e3c4a8fbbfec14f50d32a result f6ef43375702c825486beb12e32e772b0976472508ffcc38f246903d1b9f0fcf
Exploratory work on development data; not a performance claim.