Forecast accuracy calculator
Paste actual and forecast values in matching order to compare five widely used error measures.
Actual and forecast values
Lines that stay close indicate smaller errors for the displayed periods; the analytical goal determines which metric matters.
Errors by period
| Period | Actual | Forecast | Error | |Error| | APE (%) |
|---|---|---|---|---|---|
| Fill in the fields above — the result is computed on this page. | |||||
How to use this page
- Paste equal counts of actual and forecast values.
- Compare MAE, RMSE, MAPE, sMAPE and MASE.
- Inspect the chart and the errors for individual periods.
What this page does not do: MAPE is undefined when the actual value is zero, and MASE needs variation between successive actual values. The metrics describe the supplied sample and do not guarantee future performance.
Further reading
MAE has the same unit as the data and weights every absolute error equally. RMSE reacts more strongly to large errors. MAPE is easy to communicate but omits periods with an actual value of zero. sMAPE scales error by the sum of absolute values, while MASE compares MAE with a one-step naive forecast.
A lower value means less error under that metric. Metrics on different scales should not be compared directly.
Frequently asked questions
Which forecast metric is best?
No single metric is always best. MAE is readable in the data unit, RMSE penalises large errors more, and MASE supports comparison across series.
Why can MAPE be blank?
MAPE divides by the actual value. The percentage is undefined at zero, so the calculator excludes that period from MAPE.
See also
ACF and PACF autocorrelation calculator
Calculate ACF and PACF across time-series lags. Review the chart, reference bounds and coefficient table.
Adjusted R-squared calculator
Convert R² to adjusted R² using the number of observations and predictors in a regression model.
Beta distribution calculator
Calculate beta density, CDF, upper tail and quantile together with its mean and variance.