Do moves flagged by only one of two rules lead to returns beyond the market?
Study 28 · Not supported · recorded 1 October 2026 · answer: Not at the size tested · full paper with figures
Correction, 2 October 2026. The text of this record was rewritten to the project's editorial standard, from the study's plan, result and records rather than its earlier wording. What the study found is unchanged.
Correction, 3 October 2026. The outcome label was changed from "concluded" to "not supported", which states more exactly what the study found: the idea it tested, switching to the market-wide rule, was not supported.
The question
Our monitoring rule flags a share's move beyond what the market predicts for it when that move is in the top 5% of the share's own last 126 sessions and at least 2%; the alternative ranks it against every share's move that day. We follow five sessions after moves only one rule flags, and after moves only the other flags.
Rules fixed before the result
Before any return was computed, we fixed both rules, the five-session window and the bar. A set carries a result if its average, in either direction, is at least 0.30 percentage points and at least twice its standard error (its sampling noise); it is flat if its 95% interval lies inside plus or minus 0.30 points. Each set needed at least 1,000 moves and 300 days with a flag, and if both were flat, the current rule would stand.
Data and method
The outcome is the five-session return after a flag beyond what the market's move predicts for that share, signed so that a positive figure continues the move and a negative one reverses it. Moves only the own-history rule flags and moves only the market rule flags, among shares last closing at 5.00 or more in US currency from 2 April 2014 to 31 December 2022, are two sets, each judged alone against the 0.30-point bar, not against each other.
The answer
Neither set reached the 0.30-point bar set in advance. After moves flagged only by the market rule, the five-session return averaged a reversal of about 0.06 percentage points, but its 95% range runs from a 0.24-point reversal to a 0.12-point continuation, so no reversal was detected; after moves flagged only by the own-history rule, it reversed by about 0.06 points (95% range: a 0.13-point reversal to one just above zero), a reversal far too small to reach the bar. Both ranges lie inside plus or minus 0.30 points, so smaller effects remain compatible with the result, but not one as large as the bar. A side check, not the registered reading, split the market rule's moves by direction: rises reversed by about 0.62 points and falls continued by about 0.52.
What it changes
Nothing changed. The idea of switching to the market-wide rule was not supported, so the monitoring rule that judges a move against the share's own past stays as it is.
What remains uncertain
One sample, 2 April 2014 to 31 December 2022, was used; only the five-session window decides. Split by direction, the market rule's moves passed the bar both ways, rises reversing and falls continuing, which its flat average hides; that split is a reported check. Whether a move had news behind it was not tested, and with a lower price floor those moves did reverse beyond the bar, possibly because a cheap share's close sits at the bid or the ask.
How it was made
Run took 25 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Exploratory work on development data; not a performance claim.