Earnings-week filters: a calendar is not a verified edge

A dated correction preserves selected first-party reporting schedules while withdrawing unsupported report counts, filter results and risk rules.

What changed in this article

Editorial correction · 9 September 2026. The original article was published on 22 February 2026 and presented a 23–27 February earnings cluster, a 162-report test and a rules-based filter that allegedly cut trades 58%, improved hit rate 11.4 points and reduced weekly drawdown 35%. It did not preserve the report list, call archive, rule execution, benchmark, costs or calculation output. Those sample, performance, filter and prescriptive risk-rule claims are withdrawn. The original URL and publication date remain unchanged.

An earnings calendar is a dated schedule, not evidence that one selection process outperformed another.

What the dated records establish

Source date and record	Dated fact	Boundary
2 February 2026 · Snowflake first-party announcement	Snowflake scheduled its FY26 results for 25 February 2026.	This is one company schedule record; it does not establish a 50+ report count or the article's filter test.
3 February 2026 · Workday first-party announcement	Workday scheduled its FY26 results for 24 February 2026.	A scheduled date is not an earnings surprise, a trade trigger or a return.
4 February 2026 · Salesforce first-party announcement	Salesforce scheduled its FY26 results for 25 February 2026.	The record does not contain the article's report universe, influencer calls or rule outcomes.
25 February 2026 · NVIDIA FY26 Q4 results event	NVIDIA's first-party event page identifies the scheduled results event.	This is event chronology. Any later result is not evidence for the 22 February pre-event filter claim.

The listed records can anchor a historical calendar. They do not establish that the week contained exactly 50+ reports under the article's inclusion rule, or that a filter beat reactive behavior.

What is withdrawn

The 50+ report count, 162-report sample, 58% trade reduction, 11.4-point hit-rate improvement, 35% drawdown reduction, NVDA consensus-growth framing, three-branch decision matrix, active-name limit, drawdown stop and 25% over-trading rule are not verified findings. No replacement earnings-week filter, risk threshold or stock list is published.

The neutral method prompt is retained: freeze the complete report universe, call wording, execution rule and benchmark before observing outcomes. A dated calendar alone cannot answer a performance question.

Minimum register for a future earnings-week comparison

Record the complete issuer list and release timestamps; source and data vintage; call/post URLs and wording; inclusion and deduplication rules; entry, exit and benchmark definitions; costs, spreads and trading halts; missing observations; and the full pre-specified calculation output. Keep event schedules separate from expectations and realized returns.

Original publication: 22 February 2026. Correction: 9 September 2026. AI-assisted source checking and writing; no independent human expert review. This page is educational market research, not personalized investment advice.