Should I Trust My Price Forecast? I Measured It
As a price oracle: ~41% trustworthy. As a slot-picking map: 97.4% of perfect. My planner only uses it as the second thing — measured over 26,552 predictions.
Self-hosted solar + battery ROI tracker, motorbike trip writeups with GPS routes, and notes on building it all with Claude Code.
As a price oracle: ~41% trustworthy. As a slot-picking map: 97.4% of perfect. My planner only uses it as the second thing — measured over 26,552 predictions.
A week ago I said my battery had never breached its floor. Last night it did — not because the safety net failed to fire, but because a second safety system was running month-old code and kept turning the charger back off. Two backstops at war, and a human at 2am.
The 0-byte ghost.db fossil at content/data is a Ghost migration leftover. The one that actually runs this site lives one directory over — 13 MB, 44 tables, and the entire reporting and automation layer for GPS, energy, SEO, alerts, and the planner.
Most home batteries charge whenever the rate is cheap. Mine compares this cycle's cheapest slot against the next cycle's, the projected drain, the SOC-aware ceiling, and a forecast of tomorrow's solar — every 30 minutes.
A small wrapper around an amazing open-source UK bin-collection scraper — plus three upstream council fixes I sent along the way.
Discover how Home Assistant automates your solar energy system to maximize savings with Octopus Agile pricing and intelligent self-management.