This research investigates platinum price seasonality using high-order autoregressive modeling. The research object is daily platinum price dynamics (LME data, 2015–2024), focusing on long-term dependencies and cyclical patterns. The method employs stepwise decomposition of a 270-day lag autoregression AR(270) into computationally efficient 15-day lag sub-models, enabling significance testing of all coefficients while minimizing resource demands. Results identify the one-day lag as the dominant predictor, with marginal effects at 6–15-day lags and MAPE (1.15%) confirm model robustness. Conclusions indicate no statistically significant weekly cycles due to the overwhelming influence of short-term lags, though the method’s applicability in low-resource environments (e.g., Microsoft Excel) facilitates accessible highorder autoregression.