Julian Pawlowski b1e0245a60 refactor(coordinator): use IQR% as primary flat-day metric in period relaxation ... Replace CV with IQR% as the primary indicator for flat-day detection in _compute_day_effective_min(). CV is inflated by isolated price spikes (a single spike at 2× the average pushes CV to 15-25% while the core price band stays flat), causing the flat-day adaptation to be missed. IQR% (spread of the central 50% of prices / median) is unaffected by tail outliers and correctly identifies "flat core + spike" days. Threshold: LOW_IQR_PCT_FLAT_DAY_THRESHOLD = 15.0% - IQR% ≈ 1.35 × CV for symmetric data, so 15% ≈ old CV threshold of 10% - Extra headroom catches flat days with a single outlier (IQR%~3%, CV~20%) that were previously missed CV retained as fallback for edge cases where iqr_pct is None (near-zero or negative median prices). Impact: Flat days with a single isolated price spike are now correctly identified, reducing unnecessary relaxation iterations on those days.		2026-04-12 15:31:40 +00:00
..
tibber_prices	refactor(coordinator): use IQR% as primary flat-day metric in period relaxation	2026-04-12 15:31:40 +00:00