Pattern analysis — what predicted the winners
Across every scan with a recorded outcome, which signals and flags actually separated the coins that ran from the ones that didn't. The feedback loop: scan → check outcomes → see what worked.
Hold-time analysis — when do coins peak?
Each scan's price multiple recorded at fixed checkpoints after the scan. Over time this shows the historical sweet spot — how long to hold coins like these before they typically top out.
Learned scoring model
The self-refining loop adjusts each signal's weight from these outcomes (bounded 0.6×–1.7×, shrunk toward neutral until enough data). 1.00× = baseline. Runs daily; fully reversible.