Do shares that close near the day's high earn less the next session than shares that close near the low?
Study 83 · Changed a rule or plan · recorded 3 November 2026 · answer: Partly · full paper with figures
The question
Each session we compare shares by where the close sat in the day's range, holding the day's and prior week's moves. We test 2018 to 2022 against 0.10 percentage points fixed in advance (2013 to 2017: 0.170 less after a close at the high), and whether the gap has the pattern a passing distortion of the closing price would leave.
Rules fixed before the result
Fixed in advance: the answer is yes only if the gap's 95% range, widened as a false-alarm check required, lies wholly below zero and reaches 0.10 percentage points, and the planned test of its source reads the pattern a passing distortion of the closing price would leave. If the range lies below zero and that test reads otherwise, the answer is partly. At least 1,000 sessions were needed, and a power check had to pass first.
Data and method
Each session, among shares priced at five or more in US currency with a positive day's range, a least-squares fit of the next session's return over the market fund on where the close sat in that range, the day's own return and the prior five sessions' return gives the gap, averaged over sessions; its 95% range allows for correlation between nearby sessions. It first matched the earlier exploratory figures, and resampling earlier years set that range's widening against false alarms.
Main estimate
-0.247 percentage points (95% range -0.318 to -0.176), from 1258 observations
What it changes
Under the rule fixed in advance, every planned study of the next-day reversal holds where the close sat in the day's range as a control. A note for the record is proposed for screens that admit shares at the next open, stating only that shares closing at the top of the day's range closed the next session lower than shares closing at the bottom, with the source left open. 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. Each session counts once, but the descriptive readings (parts of the day, trading value, price and volume, later sessions, the day's own return, forms of the hold, years) were read one at a time: with no effect beyond the earlier years' averages, some reading shows a pattern in 37.2 percent of resampled runs. Daily prices cannot tell which source produced the gap.
How it was made
Run took 64 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan ece54735169fe4b43134b0f176037f7a7f8b3190abdefacdc489b40e1e0f7e98 result 67461feb6a157ec2e5bcae9c30197b7abd1851ea58731368ea3c6aa9f9f2f704
Exploratory work on development data; not a performance claim.