Synthetic order history

Seasonality strength: 1.0× Noise level: 1.0×

Holt-Winters smoothing

Level α (alpha): 0.25 Trend β (beta): 0.05 Seasonal γ (gamma): 0.35

Forecast

Forecast horizon: 14 days

Model fit stats

Training days—

Held-out days—

Residual σ (1-step)—

Forecast MAE—

Forecast RMSE—

Level / Trend now—

The top chart shows the raw daily order series (grey), the model's fitted one-step-ahead values (accent line) over the training window, and the forecast for the held-out days with a 95% confidence band derived from the model's own residual variance (band width grows as √h with the forecast horizon h). The lower chart decomposes the series into its live level, trend and seasonal (day-of-week) components as Holt-Winters updates them additively: level += α(y − seasonal) + (1−α)(level+trend); trend += β·Δlevel + (1−β)trend; seasonal += γ(y − level) + (1−γ)seasonal.