question about predict/pe function

I have some question about Matlab predict/pe function.The sys below in the code is a transfer function model estimated by tfest using data. When I iteratively calculate the RMSE(root mean squared error) for different k-step ahead, the RMSE for different K keeps unchanged.They all equal to the same value,which made my crazy!

#code

for K = 1:20;

err = pe(sys,data,K);

error = err.OutputData;

rmse(K) = sqrt(mean(error.^2))*100; end