Daily Research Index
After each trading day, the eligible universe is ranked into a review queue. Shape retrieval finds candidates; it does not establish that the situations belong together or that a trade exists. New picks drop after market close.
Open a row to inspect the full comp set, then qualify it by situation. Rankings are review priority—not signal strength, conviction, or prediction.
How the Daily Research Index Works
After market close, eligible charts are encoded with the primary 256-dimensional V5 price-volume representation. The system searches 25M+ historical embeddings to produce shape candidates for review. That first pass is intentionally incomplete: economic situation, instrument type, events, liquidity, regime, and point-in-time context still determine whether two cases belong together.
The index orders candidates for inspection using retrieval diagnostics and distinctiveness. It is not a search for the best historical outcome, and the order must not be interpreted as expected return or confidence.
Each row exposes descriptive forward returns from its retrieved comp set. Those outcomes are a starting point for falsifiable research, not a setup, prediction, or trading signal.
What You'll Find on This Page
- Retrieval candidates — rows prioritized for human or agent review, with summaries that remain subordinate to the underlying evidence.
- Historical forward returns — 1-day, 3-day, 5-day, and 10-day return statistics from historically similar chart patterns, including average return, win rate, and distribution data.
- Recent recap — descriptive follow-up showing what the displayed rows did next. It does not validate the ranking rule or constitute an accuracy score.
- Date navigation — browse past daily picks to see what the scanner found on any previous trading day.