Forecasting using a Moving Average/Median

[cgarciga]

Maybe I'm just not seeing it in the tsa api, but is there a function that forecasts e.g. observation $T+1$ by taking an average or median of the last $T-p:T$ observations for some positive integer $p$ (and for $p=0$ just take a random walk)? I know it's trivial, but it would be a nice convenience to avoid having to write a function for this myself :)

[bashtage]

There isn't one, but I think you can easily make these using pandas rolling and then a shift.