Adaptive Learning

The bot learns optimal zone transition times from price data and can auto-reschedule the short/long flips.

How it works

Collects prices every 5 minutes during transition windows
Finds the time with the biggest drop (morning) or rise (evening)
Auto-reschedules daily at 6 PM ET
Persists state in .market-data.json and .zone-config.json

Sampling windows (from CLAUDE.md)

Window (ET)	Purpose
7–11 AM	Morning samples — find biggest drop (optimal short entry)
2–6 PM	Evening samples — find biggest rise (optimal long entry)

Analysis uses the last 3 days of collected data.


Daily cycle
────────────────────────────────────────
7–11 AM   sample every 5 min (morning)
2–6 PM    sample every 5 min (evening)
6 PM ET   analyze last 3 days → reschedule cron

State files

File	Contents
`.market-data.json`	Price snapshots for analysis
`.zone-config.json`	Current optimal zone times

Both are gitignored. See Architecture for the full state-file list.