# Correction note — September 22, 2026

What we found

Some prices we recorded as "closing lines" were captured after the game had already started.
A date-handling bug shifted some stored start times six hours forward, so our own check for
"is this still pre-game" was comparing against the wrong time.

How many

199 stored rows carry a shifted start time, all from July. Twenty of those stored a price we
had labelled a close. Three of the twenty are posted picks.

What it changed on this site

Nothing. Two of the three posted picks were already excluded from the published average for a
different reason. The third is a pick from July 22, on a day the Yankees and Pirates played
twice. We cannot tell which of the two games that pick was on, so we cannot honestly exclude
it, and we will not guess. Its number now carries a visible flag instead of standing as a
verified close.

The published figures are unchanged:
Average CLV 0.8 (n=52, capture 38.5%)
MLB totals gate: 86 decided, 26 with a close, mean CLV +0.76, beat-close 65.4%

What we fixed

The date handling, so a missing timezone can no longer shift a time. A single rule now decides
whether a price counts as a close, and the same rule is used everywhere a close appears: the
average, the gate table, and each individual pick's card. Before today, an individual pick's
card had no such check at all.

No row was deleted, voided or re-graded. The record is unchanged.