Is combining these 3 methods not possible?

​

I have a dataframe of different people and want to create a column that is a rolling average of just that person's last X data points. I keep getting various errors and haven't been able to find a stackoverflow or youtube video explaining how to do this/if its possible.

​

I've gotten errors ranging from you can't have a non-unique multi-index, to columns must be the same length as key, but am not sure why I'm getting those errors.

​

According to a stackoverflow post this should work:

df_all['3d_MA'] = df_all.groupby('Player Name')['Distance (km)'].rolling(3).mean().reset_index(drop=True)

But it gives me a column with the wrong numbers, plus the first 2 values for each person are not NaN which they should be since there isnt enough data points.