Smoking is Good

It seems such a strange conclusion.

(clean_df
  .groupby(['smokes'])
  .agg(prob=('alive', np.mean)))

Or can you figure out what went wrong here?