A rolling average, also known as a moving average, is a metric used to forecast time series data by taking an average over a period of time to remove short term fluctuations.
It's calculated by averaging values over a specific period of time. The period average "rolls" or "moves" because each day the period changes to remove the oldest value and add the newest value.
The default time period used is 30 Days. This however can be changed to up to 60 days.
When tracking a calculated metric in a Google Ads KPI, the moving average is used for forecasting the KPI to the end of the period (week or month). The daily value utilises a Machine Learning prediction so that you can better understand daily performance, whilst the rolling average provides you an indication of performance in the future because the metric is measured as an average over time.