TradingSessions.ipynb

Kernel: Python 3 (ipykernel)

import pandas as pd
import numpy as np
import vectorbt as vbt
from datetime import timedelta

# Generate sample price
price_idx = pd.date_range('2018-01-01 12:00:00', periods=48, freq='H')
np.random.seed(42)
price = pd.Series(np.random.uniform(size=price_idx.shape), index=price_idx)
print(price)

2018-01-01 12:00:00    0.374540
2018-01-01 13:00:00    0.950714
2018-01-01 14:00:00    0.731994
2018-01-01 15:00:00    0.598658
2018-01-01 16:00:00    0.156019
2018-01-01 17:00:00    0.155995
2018-01-01 18:00:00    0.058084
2018-01-01 19:00:00    0.866176
2018-01-01 20:00:00    0.601115
2018-01-01 21:00:00    0.708073
2018-01-01 22:00:00    0.020584
2018-01-01 23:00:00    0.969910
2018-01-02 00:00:00    0.832443
2018-01-02 01:00:00    0.212339
2018-01-02 02:00:00    0.181825
2018-01-02 03:00:00    0.183405
2018-01-02 04:00:00    0.304242
2018-01-02 05:00:00    0.524756
2018-01-02 06:00:00    0.431945
2018-01-02 07:00:00    0.291229
2018-01-02 08:00:00    0.611853
2018-01-02 09:00:00    0.139494
2018-01-02 10:00:00    0.292145
2018-01-02 11:00:00    0.366362
2018-01-02 12:00:00    0.456070
2018-01-02 13:00:00    0.785176
2018-01-02 14:00:00    0.199674
2018-01-02 15:00:00    0.514234
2018-01-02 16:00:00    0.592415
2018-01-02 17:00:00    0.046450
2018-01-02 18:00:00    0.607545
2018-01-02 19:00:00    0.170524
2018-01-02 20:00:00    0.065052
2018-01-02 21:00:00    0.948886
2018-01-02 22:00:00    0.965632
2018-01-02 23:00:00    0.808397
2018-01-03 00:00:00    0.304614
2018-01-03 01:00:00    0.097672
2018-01-03 02:00:00    0.684233
2018-01-03 03:00:00    0.440152
2018-01-03 04:00:00    0.122038
2018-01-03 05:00:00    0.495177
2018-01-03 06:00:00    0.034389
2018-01-03 07:00:00    0.909320
2018-01-03 08:00:00    0.258780
2018-01-03 09:00:00    0.662522
2018-01-03 10:00:00    0.311711
2018-01-03 11:00:00    0.520068
Freq: H, dtype: float64

# Sessions must be equal - fill missing dates
# Fill on first date before 12:00 and on last date after 11:00
first_date = price.index[0].date()
last_date = price.index[-1].date()+timedelta(days=1)
filled_idx = pd.date_range(first_date, last_date, freq='H')
filled_price = price.reindex(filled_idx)
print(filled_price)

2018-01-01 00:00:00   NaN
2018-01-01 01:00:00   NaN
2018-01-01 02:00:00   NaN
2018-01-01 03:00:00   NaN
2018-01-01 04:00:00   NaN
                       ..
2018-01-03 20:00:00   NaN
2018-01-03 21:00:00   NaN
2018-01-03 22:00:00   NaN
2018-01-03 23:00:00   NaN
2018-01-04 00:00:00   NaN
Freq: H, Length: 73, dtype: float64

# Remove dates that are outside of trading sessions
session_price_idx = filled_price.between_time('9:00', '17:00', include_end=False).index
session_price = filled_price.loc[session_price_idx]
print(session_price)

2018-01-01 09:00:00         NaN
2018-01-01 10:00:00         NaN
2018-01-01 11:00:00         NaN
2018-01-01 12:00:00    0.374540
2018-01-01 13:00:00    0.950714
2018-01-01 14:00:00    0.731994
2018-01-01 15:00:00    0.598658
2018-01-01 16:00:00    0.156019
2018-01-02 09:00:00    0.139494
2018-01-02 10:00:00    0.292145
2018-01-02 11:00:00    0.366362
2018-01-02 12:00:00    0.456070
2018-01-02 13:00:00    0.785176
2018-01-02 14:00:00    0.199674
2018-01-02 15:00:00    0.514234
2018-01-02 16:00:00    0.592415
2018-01-03 09:00:00    0.662522
2018-01-03 10:00:00    0.311711
2018-01-03 11:00:00    0.520068
2018-01-03 12:00:00         NaN
2018-01-03 13:00:00         NaN
2018-01-03 14:00:00         NaN
2018-01-03 15:00:00         NaN
2018-01-03 16:00:00         NaN
dtype: float64

# Select first and last ticks of each trading session and split price into ranges between those ticks
start_idxs = session_price.index[session_price.index.hour == 9]
end_idxs = session_price.index[session_price.index.hour == 16]
price_per_session, _ = session_price.vbt(freq='1H').range_split(start_idxs=start_idxs, end_idxs=end_idxs)
print(price_per_session)

split_idx         0         1         2
0               NaN  0.139494  0.662522
1               NaN  0.292145  0.311711
2               NaN  0.366362  0.520068
3          0.374540  0.456070       NaN
4          0.950714  0.785176       NaN
5          0.731994  0.199674       NaN
6          0.598658  0.514234       NaN
7          0.156019  0.592415       NaN

# Run your strategy (here using random signals)
entries, exits = pd.DataFrame.vbt.signals.generate_random_both(price_per_session.shape, n=2, seed=42)
pf = vbt.Portfolio.from_signals(price_per_session, entries, exits, freq='1H')
print(pf.total_return())

0   -0.786858
1    1.466807
2   -0.529509
Name: total_return, dtype: float64