Do shares' daily lows pile up just past ten percent below the previous close, where short selling is limited?
Study 95 · Changed a rule or plan · recorded 12 November 2026 · answer: Yes · full paper with figures
The question
Once a share falls ten percent below the previous close, a market rule limits short selling for the rest of that day and the next. For shares priced five or more at the previous close, do more daily lows sit just past that point than a smooth count of neighbouring points gives, by 0.15 or more on a log scale?
Data and method
For each share and day with a low three to twenty percent below the previous close, falls are counted in steps of half a percent around ten percent. The measure is how many more of the day's lowest prices sit just past that point than a smooth count of the neighbouring points gives, on a log scale, read quarter by quarter. Nine and eleven percent down, highs ten percent up and the count without exact-level lows are read alike.
The answer
Yes. Over the 20 calendar quarters of 2018 to 2022, the extra lows just past ten percent down came to 0.34 on the log scale (95% range 0.29 to 0.39), wholly above the 0.15 set in advance, so the excess is at least that size. It stood 0.39 (0.34 to 0.44) above the same measure at nine and eleven percent down, and 0.35 (0.27 to 0.43) above the same measure for daily highs ten percent up, where no such rule applies. Without the lows at exactly ten percent down it was 0.31 (0.25 to 0.36), so the pile is not only lows at the exact level. The next day's return was 0.34 percentage points lower just past the level than just short of it (95% range 0.66 to 0.02 lower), inside the bound of 0.75 set in advance.
Main estimate
0.341 ratio (95% range 0.292 to 0.391), from 347671 observations
What remains uncertain
Earlier studies used these years (studies 90 to 92 read returns after large one-day moves), so this is a check inside our development data, not an independent test. Some days counted were already under the short-selling limit from the day before (7.9 percent of lows near the level); without them the measure was 0.37 (0.32 to 0.42). This is a reading of these five years quarter by quarter, and the daily low only stands in for the rule's trigger.
How it was made
Run took 44 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan 9e9ac5ab26c760e6dc919367f8611c4082266b2a963d22c1319cb14c8810f96a result c456d3362e20363b64b8ebbd86e75a0752ab26132d2d48a40471ef9d03511981
Exploratory work on development data; not a performance claim.