visualize the rolling averages to see if it makes sense. I don't understand why sum was used when the rolling average was requested.
df=pd.read_csv('poll.csv',parse_dates=['enddate'],dtype={'favorable':np.float,'unfavorable':np.float,'other':np.float})
df.set_index('enddate')
df=df.fillna(0)
fig, axs = plt.subplots(figsize=(5,10))
df.plot(x='enddate', ax=axs)
plt.show()
df.rolling(window=3,min_periods=3).mean().plot()
plt.show()
print("The larger the window coefficient the smoother the line will appear")
print('The min_periods is the minimum number of observations in the window required to have a value')
df.rolling(window=6,min_periods=3).mean().plot()
plt.show()