The errors are computed in the next column (here, column H) by subtracting the forecasts from the actual values. The root mean squared error is computed as the square root of the variance of the errors plus the square of the mean. (This follows from the mathematical identity: MSE= VARIANCE(errors)+ (AVERAGE(errors))^2.) In calculating the mean and variance of the errors in this formula, the first two periods are excluded because the model does not actually begin forecasting until the third period (row 15 on the spreadsheet). The optimal value of alpha can be found either by manually changing alpha until the minimum RMSE is found, or else you can use the "Solver" to perform an exact minimization. The value of alpha that the Solver found is shown here (alpha=0.471154)