This is an extension to this accepted answer.

My DataFrame:

import pandas as pd
df = pd.DataFrame(
    {
        'a': [-3, -1, -2, -5, 10, -3, -13, -3, -2, 1, 2, -100],
        'b': [1, 2, 3, 4, 5, 10, 80, 90, 100, 99, 1, 12]
    }
)

Expected output:

      a    b
5    -3   10
6   -13   80
7    -3   90
8    -2  100

Logic:

a) Selecting the longest streak of negatives in a.

b) If for example there are two streaks that has same size, I want the one that has a greater sum of b. In df there are two streaks with size of 4 but I want the second one because sum of b is greater.

My Attempt:

import numpy as np
s = np.sign(df['a'])
df['g'] = s.ne(s.shift()).cumsum()
df['size'] = df.groupby('g')['g'].transform('size')
df['b_sum'] = df.groupby('g')['b'].transform('sum')