import pandas as pd
import numpy as np

Raw Data:

date       lplp_returns
2018-03-14  0.000000
2018-03-15  0.000000
2018-03-16  0.000006
2018-03-19  -0.000469
2018-03-20  -0.001312
... ...
2025-04-10  0.082415
2025-04-11  0.002901
2025-04-14  0.005738
2025-04-15  0.007664
2025-04-16  0.012883
1848 rows × 1 columns

Creating groups before using groupBy helped get me going in the right direction:

start_date = pd.Timestamp('2018-04-20')
df['group'] = ((df.index - start_date).days // 365)
grouped = df.groupby('group')
result = grouped['lplp_returns']

From there, I want cumulative returns (cumprod). This, of course, is problematic because it is a groupby/transform operation.

g = result.apply(lambda x: np.cumprod(1 + x) - 1)
g.groupby('group').tail(1)

Output:

group  Date      
-1     2018-04-19    0.003971
 0     2019-04-19   -0.077341
 1     2020-04-17   -0.068972
 2     2021-04-16    0.429971
 3     2022-04-18   -0.024132
 4     2023-04-18    0.032741
 5     2024-04-17    0.190119
 6     2025-04-16    0.131955
Name: lplp_returns, dtype: float64

This gets me 95% to where I want to be.

Needs for improvement:

(1) I don't want/need group '-1',

(2) I want each group to start on or after 'xxxx-04-20' not to proceed 'xxxx-04-20', and

(3) to stop on or before 'xxxx-04-20' not to exceed 'xxxx-04-20'. (This is to address trading days).

Suggestions on coding or approaching/solving this in a better way?