State map
A score is a timestamp, not an identity
Early discoveryOften rank 31+, limited visibility, and incomplete confirmation.
DevelopingRank 6-30, improving persistence, sector context, and possible accumulation.
Visible leaderTop ranks attract attention but can carry worse entry comfort.
OverheatExtension, opening fade, or weakened reward-to-risk can outweigh trend strength.
RetiredStructure, sector support, or the original hypothesis no longer survives.
Research lesson
Promotion was more interesting than static rank
The project began by asking whether the highest score identified the best future return. Repeated reviews shifted the more useful question toward trajectory: first entry into the broader pool, persistence, movement toward the visible ranks, and whether intraday demand confirmed or rejected that promotion.
This lifecycle remains a research hypothesis. The broad controlled replay did not establish benchmark-beating performance, so state labels should organize attention rather than prescribe trades.
Daily observation protocol
What to record at each transition
| Transition | Daily evidence | Intraday confirmation | Failure condition |
|---|---|---|---|
| Unseen to early | First signal date, rank band, score, sector, distance from trend anchors. | Opening gap, first 30-minute low, VWAP behavior. | Immediate loss of structure or sector divergence. |
| Early to developing | Persistence, rank improvement, score breadth, market regime. | Pullback holds with relative volume and sector support. | Promotion driven only by a one-session price burst. |
| Developing to visible | Top-10 or Top-5 entry, extension, momentum-risk state. | Opening acceptance rather than a gap-and-fade. | Visibility rises while entry quality and excess return deteriorate. |
| Visible to overheat | Extension, high RSI, weakening incremental score. | Failed VWAP reclaim or break of the first 30-minute low. | No follow-through after the visibility event. |
| Any state to retired | Absent signal dates, sector loss, thesis invalidation. | Repeated failed support or adverse event evidence. | Re-entry requires a new episode, not a rewritten old one. |
Version discipline
How to keep the lifecycle honest
- Freeze the state definition before the next observation window.
- Track every eligible candidate, not only memorable winners.
- Report raw, QQQ-relative, and SPY-relative outcomes together.
- Separate daily discovery from partial-session confirmation.
- Start a new episode after one or more absent signal dates.
Review the current model evidence | Read the original lifecycle field note