Top of Book Snapshots at One-Minute Intervals¶

This notebook demonstrates how to capture best bid and ask prices at regular intervals from ITCH 5.0 data.

Top of book snapshots are useful for:

• Creating time series of bid-ask spreads
• Analyzing price movements over time
• Calculating mid-price evolution

This example uses the S081321-v50-AAPL-SPY.itch50.gz file generated in the Extracting Symbols example.

In [ ]:

from pathlib import Path
import datetime
import pandas as pd
from meatpy.itch50 import ITCH50MessageReader, ITCH50MarketProcessor
from meatpy.event_handlers.lob_recorder import LOBRecorder
from meatpy.writers.parquet_writer import ParquetWriter

# Define paths and parameters
data_dir = Path("data")

file_path = data_dir / "S081321-v50-AAPL-SPY.itch50.gz"
outfile_path = data_dir / "spy_tob.parquet"
book_date = datetime.datetime(2021, 8, 13)


with ITCH50MessageReader(file_path) as reader, ParquetWriter(outfile_path) as writer:
    processor = ITCH50MarketProcessor("SPY", book_date)

    # We only care about the top of book
    tob_recorder = LOBRecorder(writer=writer, max_depth=1)
    # Generate a list of timedeltas from 9:30 to 16:00 (inclusive) in 1-minute increments
    market_open = book_date + datetime.timedelta(hours=9, minutes=30)
    market_close = book_date + datetime.timedelta(hours=16, minutes=0)
    record_timestamps = [
        market_open + datetime.timedelta(minutes=i)
        for i in range(int((market_close - market_open).total_seconds() // 60) + 1)
    ]
    tob_recorder.record_timestamps = record_timestamps

    # Attach the recorders to the processor
    processor.handlers.append(tob_recorder)

    for message in reader:
        processor.process_message(message)

from pathlib import Path import datetime import pandas as pd from meatpy.itch50 import ITCH50MessageReader, ITCH50MarketProcessor from meatpy.event_handlers.lob_recorder import LOBRecorder from meatpy.writers.parquet_writer import ParquetWriter # Define paths and parameters data_dir = Path("data") file_path = data_dir / "S081321-v50-AAPL-SPY.itch50.gz" outfile_path = data_dir / "spy_tob.parquet" book_date = datetime.datetime(2021, 8, 13) with ITCH50MessageReader(file_path) as reader, ParquetWriter(outfile_path) as writer: processor = ITCH50MarketProcessor("SPY", book_date) # We only care about the top of book tob_recorder = LOBRecorder(writer=writer, max_depth=1) # Generate a list of timedeltas from 9:30 to 16:00 (inclusive) in 1-minute increments market_open = book_date + datetime.timedelta(hours=9, minutes=30) market_close = book_date + datetime.timedelta(hours=16, minutes=0) record_timestamps = [ market_open + datetime.timedelta(minutes=i) for i in range(int((market_close - market_open).total_seconds() // 60) + 1) ] tob_recorder.record_timestamps = record_timestamps # Attach the recorders to the processor processor.handlers.append(tob_recorder) for message in reader: processor.process_message(message)

In [2]:

pd.read_parquet(outfile_path).head(10)

pd.read_parquet(outfile_path).head(10)

Out[2]:

	Timestamp	Type	Level	Price	Volume	N Orders
0	2021-08-13 09:30:00	Ask	1	445.22	300	3
1	2021-08-13 09:30:00	Bid	1	445.68	4200	4
2	2021-08-13 09:31:00	Ask	1	445.22	300	3
3	2021-08-13 09:31:00	Bid	1	445.68	4200	4
4	2021-08-13 09:32:00	Ask	1	445.22	300	3
5	2021-08-13 09:32:00	Bid	1	445.68	4200	4
6	2021-08-13 09:33:00	Ask	1	445.22	300	3
7	2021-08-13 09:33:00	Bid	1	445.68	4200	4
8	2021-08-13 09:34:00	Ask	1	445.22	300	3
9	2021-08-13 09:34:00	Bid	1	445.68	4200	4