Cup and handle (O'Neil 1988)

Thesis (rejected): After a rounded base of 30-100 days with a right rim within 5 % of the left rim and a 5-20 day handle that retraces no more than half the cup height, a breakout above the rim on volume continues higher over 5-60 days.

Code: research/cup_handle.py Report: research/cup_handle_report.json (n=13,124)

Why it failed: Edges are flat-to-negative across every horizon.

The breakout-on-volume above a long base is in the same family as our coil-breakout cohort (which works), but the shape of the base — specifically the U-rounded retrace required by cup-and-handle — selects a different population than coil's tight flat base. Either O'Neil's pattern has been arbitraged or it never worked as a stand-alone signal.

Verdict: Cohort skipped. Coil-breakout already captures the useful base-and-breakout structure.