Gap register

Do gaps fill more often than chance?

Tested 2026-11-05 on 181 stocks, 25,397 gaps, sessions 2024-07-01 to 2026-11-04. Two controls, each counted by the same rule: a price the same distance from the open of an ordinary session of the same stock (open within 1% of its prior close), on the same side; and a price the same distance from the gap session’s own open on the other side, which shares that session’s wider range. The second was added after the first run and is reported beside it.

Ranges in brackets are 95% bootstrap ranges over stocks, because gaps within one stock are not independent.

Most of a gap fill rate is the base rate. Within 20 sessions 85.0% of gaps went back to the prior close, and 82.5% of random levels the same distance from an ordinary open, and 83.2% of levels the same distance on the other side of the same open, were reached too. The prior close was reached 1.8 to 2.5 points more often over 20 sessions, more on the gap session itself; against the same-session control, gaps up showed no excess past the gap session (83.0% against 85.9% within 20).

The run, its control and its numbers are on record in the File Drawer, with the launch-level test run on the same bars. A gap is a fact about one open; nothing here says where a price will go.