At a flat market, did shares the market fund explains most beat those it explains least over the next month?
Study 69 · Inconclusive · recorded 25 October 2026 · answer: Not settled · full paper with figures
The question
Every twenty-one sessions, shares priced at five or more in the market's own currency were sorted into fifths by how much of their daily movement over the past year the market fund explains; we compare the next twenty-one sessions' return of the most explained fifth with the least explained fifth's, one gap per formation.
Rules fixed before the result
Fixed in advance from the earlier years: the plain average decided if its smallest detectable gap was at most a point and a half, else the gap at a flat market, net of the market fund's month through the fifths' exposures, if its own was, else the median after a false-alarm check. Ranges are widened for one formation carrying into the next. The study would have ended had the gap within similar largest-move groups been under half the plain one.
Data and method
The plain average's smallest detectable gap, 1.562 points, missed the bar, while the flat-market gap's, 1.139, passed, so the flat-market gap decided; the median's false-alarm check did not pass its band. Consecutive formations' gaps did not move together, so the gap's own ranges were not widened, though some side ranges were, the least explained fifth's among them. On the earlier years the gap within similar largest-move groups, 0.496, was above 0.397, half the plain gap, so the study ran.
The answer
At a flat market the gap was 0.557 points (-0.842 to 1.956), unsettled; the plain average 0.760 (-0.639 to 2.159), the market fund's month explaining 0.048 of the gaps' variation. Against the market fund neither fifth settled: most explained -0.217 (-0.912 to 0.478), least -0.977 (-2.346 to 0.391). Within similar largest-move groups the gap was 0.895 (-0.309 to 2.099), within similar firm-specific volatility 0.857 (-0.523 to 2.236), among shares traded on both end sessions 0.750 (-0.647 to 2.146), and among domestic operating companies 0.591 (-0.928 to 2.109): each too weak to settle. The least explained fifth held 21% foreign listings and 5% blank-cheque shells, priced 9.50 to 10.50 on at least nine-tenths of the year's sessions, now labelled as what they became. By year the plain average ran from -1.386 (2020) to 2.489 (2019); dropping any year kept it positive.
Main estimate
0.557 percentage points (95% range -0.842 to 1.956), from 59 observations
What it changes
By the rule set in advance, an unsettled answer leaves the range and the count as the record, and the count cannot be raised within our development data: no co-movement adjustment is proposed for later studies' comparisons with the market fund, and nothing else changes. The plan expected no gap as large as a point and a half, with low confidence, and gave an unsettled answer a chance of 0.24.
What remains uncertain
Earlier studies used these years, so this checks a pattern within our development data rather than on fresh data. The sorting share is estimated over a year and noisy, the flat-market gap rests on a straight-line fit across formations, and security types are the vendor's current labels; five years, 2020 among them, and 59 formations are few, so the allowance for one formation carrying into the next is rough, and trading value is the only size measure held.
How it was made
Run took 83 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan 2b0ded0ed3f743e87ed481810601f3844c18be1cda78cfa896bd8c502ef6fc7a result e8dfb1ada0c1e0714819888ba47c6a9870908d5a6e23ce0406ee49ded44b9937
Exploratory work on development data; not a performance claim.