How often do our two-group decision rules report a false gap when one group's extreme values reach further?
Study 88 · No reading (stopped by its own checks) · recorded 5 November 2026 · answer: Not settled · full paper with figures
The question
Earlier studies chose between two groups by rules pairing the plain average with the clipped one (which pulls each group's extreme values in), and one rule widened its ranges; the clipped average alone was shown to report false gaps when one group's extremes reach further. We ask how often each whole rule, applied step by step, reports a false gap.
Rules fixed before the result
Fixed before any test: on made-up comparisons with no true difference, a rule failed when its false-gap rate was shown above 7.5% (3.75% on either side) and held when its 95% range stayed within those limits, each rule read on 6,000 comparisons. Checks came first, any failure stopping the study: one required the range for the chance of passing study 67's advance check to start at 90% or higher at that study's recorded counts.
Data and method
The plan was to rebuild two earlier studies' ordinary trading days, redraw study 76's made-up comparisons with no true difference, and run each earlier rule on each comparison as written, with study 67's widening of its ranges set afresh on each comparison as its own record did. Study 67's advance check was to be represented by the probability of passing it, with a range from resampled counts. None of these steps ran.
The answer
Not settled: the study stopped at a check planned before any data was read, so no rate of false gaps was measured. Study 67's rule begins with an advance check, and the plan measures the chance of passing that check with a range around it. The plan required that, at the counts study 67's own advance check recorded (57 and 60 of 2,000 on the two sides), the range's low end be at least 90%; worked out by the plan's own method on made-up counts alone, it falls far below 90%, so no faithful run could pass. By the rule set in advance, a failed check stopped the study. Nothing is shown about how often any of the rules reports a false gap.
What it changes
No rule and no published answer changed; study 86 stopped when a rebuilt rule missed its earlier record's count, and this corrected re-run stopped at a test of its own plan before reading any data. The plan said a second stop would close the question; the owner set that aside, since a faulty check does not answer it. A corrected plan must first show, worked out, that it meets its own checks, and an outside reviewer reads it first.
What remains uncertain
No rate of false gaps was measured, so the question stays open for every rule studied. The stop came from the plan's own arithmetic, not from data: no day was rebuilt and no comparison drawn. These days were studied before, so even a measured rate would have been a check within our development data, not a test on fresh data.