Compare observed returns by month and weekday, with sample sizes. Historical patterns do not predict price direction.
Compare observed returns by month and weekday, with sample sizes. Historical patterns do not predict price direction. Only closed candles are used. Monthly returns compare consecutive complete calendar months; weekdays use session close-to-close returns. Hours use complete hour open-to-close returns in exchange time, including DST. Gaps are never filled. History depth depends on the provider; splits and adjustments may affect results.
Minimum samples: 5 per month, 20 per weekday, 30 per hour. Estimates are hidden below these thresholds.
GARCH estimates the size of moves, never direction. ±1σ and ±2σ are model ranges for the closing price (about 68% and 95% under a normal log-return assumption), not guaranteed path containment. The existing accuracy page scores ±1.96σ, not exactly ±2σ.
Scored track record