Skip to content

renves/b3quant

BranchesTags

Repository files navigation

b3quant

PyPI version Python 3.10+ License: MIT Tests

Python library for downloading and parsing historical market data from B3 (Brazilian Stock Exchange).

Features

  • Download COTAHIST files - Yearly, monthly, or daily historical data
  • Parse to pandas DataFrames - Clean, typed data ready for analysis
  • Filter by instrument type - Options, stocks, or all instruments
  • Command Line Interface - Terminal-based access with rich output and progress tracking
  • Async downloads - Concurrent downloads for 5x faster multi-year data fetching
  • Parallel parsing - Multi-core processing for 3-4x faster parsing of large files
  • Parquet data lake - Efficient columnar storage with partitioning and compression
  • Simple, Pythonic API - Intuitive interface
  • Type hints - Full type annotations for better IDE support
  • Smart caching with TTL - JSON or SQLite cache backends with automatic expiration
  • Retry with exponential backoff - Automatic retry with jitter to handle network failures
  • Progress bars - Visual feedback for long-running downloads
  • Black-Scholes pricing - Option pricing and Greeks calculation
  • Implied Volatility solver - Multi-method IV calculation (Newton-Raphson, Brent)
  • Feature engineering for ML - Advanced features for options pricing models
  • No R dependencies - Pure Python implementation

Installation

From PyPI (Recommended)

pip install b3quant

From Source (Development)

If you cloned the repository:

git clone https://github.com/renves/b3quant.git
cd b3quant
pip install -e .

Or using uv:

git clone https://github.com/renves/b3quant.git
cd b3quant
uv sync
uv run python your_script.py

Quick Start

import b3quant as pyb

# Get all options traded in 2024
options = pyb.get_options(year=2024)

# Get options from November 2024
options_nov = pyb.get_options(year=2024, month=11)

# Get options from a specific day
options_day = pyb.get_options(year=2024, month=12, day=20)

# Filter by underlying asset
petr_options = options[options['underlying'] == 'PETR']

# Get specific columns
print(options[['ticker', 'strike_price', 'close_price', 'volume']].head())

Usage

Basic Usage

from b3quant import B3Quant

# Initialize
b3 = B3Quant()

# Get options for a single year
options_2024 = b3.get_options(year=2024)

# Get options for a specific month
options_nov = b3.get_options(year=2024, month=11)

# Get options for a specific day
options_day = b3.get_options(year=2024, month=12, day=20)

# Get stocks data
stocks_2024 = b3.get_stocks(year=2024)
stocks_nov = b3.get_stocks(year=2024, month=11)
stocks_day = b3.get_stocks(year=2024, month=12, day=20)

# Get all instruments
all_data = b3.get_all(year=2024)

Working with Options Data

import b3quant as pyb

# Get options
options = pyb.get_options(year=2024)

# Filter by type
calls = options[options['instrument_type'] == 'CALL']
puts = options[options['instrument_type'] == 'PUT']

# Filter by underlying
petr_options = options[options['underlying'] == 'PETR']

# Get options near expiration
short_term = options[options['days_to_maturity'] <= 30]

# Calculate moneyness (requires underlying price)
# You'll need to merge with stocks data or calculate separately

Advanced: Enrich with Underlying Prices

import b3quant as pyb

# Get options and stocks from a specific month
options = pyb.get_options(year=2024, month=11)
stocks = pyb.get_stocks(year=2024, month=11)

# Merge to get underlying prices
options_enriched = options.merge(
    stocks[['underlying', 'trade_date', 'close_price']],
    left_on=['underlying', 'trade_date'],
    right_on=['underlying', 'trade_date'],
    how='left',
    suffixes=('', '_underlying')
)

# Calculate moneyness
options_enriched['moneyness'] = (
    options_enriched['close_price_underlying'] /
    options_enriched['strike_price']
)

# Filter ATM options (at-the-money)
atm_options = options_enriched[
    (options_enriched['moneyness'] >= 0.95) &
    (options_enriched['moneyness'] <= 1.05)
]

Custom Cache Directory

from b3quant import B3Quant

# Use custom cache directory
b3 = B3Quant(cache_dir="./my_data_cache")
options = b3.get_options(year=2024)

Force Re-download

# Force re-download even if file exists in cache
options = b3.get_options(year=2024, force_download=True)

Cache Configuration

b3quant supports two cache backends with automatic TTL (Time-To-Live) expiration:

from b3quant import B3Quant
from b3quant.downloaders.cotahist import COTAHISTDownloader

# Use JSON cache (default, human-readable)
downloader = COTAHISTDownloader(
    cache_dir="./data/raw",
    use_metadata_cache=True  # Default: True
)

# Files are cached for 30 days by default
# Configure in b3quant.config.CACHE_TTL_DAYS

# Disable progress bars if needed
downloader = COTAHISTDownloader(show_progress=False)

# Download multiple years with progress bar
paths = downloader.download_range(2020, 2024)

Cache Backends:

  • JSON (default): Simple, human-readable, good for small datasets
  • SQLite: More efficient for large datasets and concurrent access

Configure in b3quant/config.py:

CACHE_BACKEND = "json"  # or "sqlite"
CACHE_TTL_DAYS = 30     # Cache expiration in days

Automatic Retry

Downloads automatically retry on failure using exponential backoff with jitter:

# Automatic retry is built-in
# Default: 3 attempts with exponential backoff
downloader = COTAHISTDownloader()
path = downloader.download_yearly(2024)  # Retries automatically on failure

# Customize retry attempts
path = downloader.download_yearly(2024, max_retries=5)

Retry Strategy:

  • Exponential backoff: delays increase exponentially (1s, 2s, 4s, ...)
  • Jitter: random delay to prevent thundering herd problem
  • Configurable in b3quant.config: MAX_RETRY_ATTEMPTS, RETRY_BASE_DELAY, RETRY_MAX_DELAY

Command Line Interface (CLI)

b3quant provides a full-featured CLI for terminal-based workflows:

# Download 2024 options data
b3quant download --year 2024

# Download November 2024 stocks
b3quant download --year 2024 --month 11 --instrument stocks

# Download multiple years
b3quant download --start-year 2020 --end-year 2024

# Export to Parquet
b3quant download --year 2024 --output-format parquet --output-dir ./data

# Show configuration
b3quant info

CLI Features:

  • Rich terminal output with tables and colors
  • Progress bars for downloads
  • Data summaries after download
  • Export to CSV or Parquet
  • Configurable cache directory

Async Downloads

Download multiple years concurrently for 5x performance improvement:

import asyncio
from b3quant.downloaders.async_cotahist import AsyncCOTAHISTDownloader

async def main():
    downloader = AsyncCOTAHISTDownloader(max_concurrent=5)

    # Download multiple years concurrently
    paths = await downloader.download_range(2020, 2024)
    print(f"Downloaded {len(paths)} files")

asyncio.run(main())

# Or use synchronous wrapper
from b3quant.downloaders.async_cotahist import download_range_sync
paths = download_range_sync(2020, 2024)

Performance: 5x faster for downloading 5 years of data (concurrent vs sequential)

Parallel Parser

Parse large files using multiple CPU cores for 3-4x speed improvement:

from b3quant.parsers.parallel_parser import ParallelCOTAHISTParser

# Use all CPU cores
parser = ParallelCOTAHISTParser()
df = parser.parse_file('COTAHIST_A2024.TXT', instrument_filter='options')

# Specify number of workers
parser = ParallelCOTAHISTParser(n_workers=4)
df = parser.parse_file('COTAHIST_A2024.TXT')

# Parse multiple files
files = ['COTAHIST_A2023.TXT', 'COTAHIST_A2024.TXT']
df = parser.parse_multiple_files(files, instrument_filter='options')

Performance: 3-4x faster on multi-core CPUs for files with millions of records

Parquet Data Lake

Store and query data efficiently using Parquet format:

from b3quant.storage import ParquetStorage

# Initialize storage
storage = ParquetStorage(base_path='./data/lake')

# Write partitioned data
storage.write_options(options_df, year=2024, month=11)

# Read data
df = storage.read_options(year=2024, month=11)

# Read specific columns (column pruning)
df = storage.read_options(year=2024, columns=['ticker', 'close_price'])

# Read with filters (predicate pushdown)
df = storage.read_options(
    year=2024,
    filters=[('underlying', '=', 'PETR')]
)

# Get storage statistics
stats = storage.get_stats('options')
print(f"Partitions: {stats['partitions']}")
print(f"Total size: {stats['total_size_mb']:.2f} MB")
print(f"Row count: {stats['row_count']:,}")

Benefits:

  • 10-20x smaller file size vs CSV
  • Faster queries with column pruning and predicate pushdown
  • Partitioned by year/month/day for efficient access
  • Compression options: snappy (default), gzip, zstd

DataFrame Schema

Options Data

Column Type Description
record_type str Record type code
trade_date date Trading date
ticker str Option ticker (e.g., PETRL255)
instrument_type str CALL or PUT
underlying str Underlying asset code (e.g., PETR)
company_name str Company name
strike_price float Strike price in BRL
maturity_date date Expiration date
open_price float Opening premium
high_price float Highest premium
low_price float Lowest premium
close_price float Closing premium
avg_price float Average premium
volume float Trading volume in BRL
trades_count int Number of trades
quantity int Contracts traded
days_to_maturity int Days until expiration
time_to_maturity float Years until expiration

Stocks Data

Similar schema but without strike_price, maturity_date, and option-specific fields.

Options Pricing

b3quant includes a Black-Scholes pricing model for European options with full Greeks calculation.

Black-Scholes Pricing

from b3quant.models.black_scholes import BlackScholes
import numpy as np

# Initialize model
bs = BlackScholes()

# Price a call option
call_price = bs.price(
    S=100,           # Spot price
    K=100,           # Strike price
    T=1.0,           # Time to maturity (years)
    r=0.05,          # Risk-free rate
    sigma=0.2,       # Volatility
    option_type='call'
)
print(f"Call price: {call_price:.2f}")  # 10.45

# Calculate Greeks
delta = bs.delta(S=100, K=100, T=1.0, r=0.05, sigma=0.2, option_type='call')
gamma = bs.gamma(S=100, K=100, T=1.0, r=0.05, sigma=0.2)
vega = bs.vega(S=100, K=100, T=1.0, r=0.05, sigma=0.2)
theta = bs.theta(S=100, K=100, T=1.0, r=0.05, sigma=0.2, option_type='call')
rho = bs.rho(S=100, K=100, T=1.0, r=0.05, sigma=0.2, option_type='call')

print(f"Delta: {delta:.4f}")    # 0.6368
print(f"Gamma: {gamma:.4f}")    # 0.0199
print(f"Vega: {vega:.4f}")      # 39.79
print(f"Theta: {theta:.4f}")    # -6.41
print(f"Rho: {rho:.4f}")        # 53.23

Vectorized Pricing

The Black-Scholes model supports vectorized operations for efficient bulk calculations:

import numpy as np
from b3quant.models.black_scholes import BlackScholes

bs = BlackScholes()

# Price multiple strikes at once
strikes = np.array([95, 100, 105])
prices = bs.price(S=100, K=strikes, T=1.0, r=0.05, sigma=0.2, option_type='call')
print(prices)  # [13.04, 10.45, 8.24]

# Calculate Greeks for entire option chain
deltas = bs.delta(S=100, K=strikes, T=1.0, r=0.05, sigma=0.2, option_type='call')
print(deltas)  # [0.7112, 0.6368, 0.5596]

Dividends Support

# Price with continuous dividend yield
call_price = bs.price(
    S=100, K=100, T=1.0, r=0.05, sigma=0.2,
    q=0.02,  # 2% dividend yield
    option_type='call'
)

Implied Volatility Calculation

b3quant includes a robust multi-method IV solver:

from b3quant.models.iv_solver import ImpliedVolatilitySolver

solver = ImpliedVolatilitySolver()

# Calculate IV for a single option
iv = solver.solve(
    price=5.0,
    S=100,
    K=105,
    T=0.5,
    r=0.05,
    option_type='call',
    method='newton'  # or 'brent', 'auto'
)
print(f"Implied Volatility: {iv:.2%}")

# Vectorized IV calculation
import numpy as np
prices = np.array([5.0, 3.5, 2.0])
strikes = np.array([95, 100, 105])

ivs = solver.solve_vectorized(
    price=prices,
    S=100,
    K=strikes,
    T=0.5,
    r=0.05,
    option_type='call'
)

Available methods:

  • newton: Newton-Raphson (fast, requires good initial guess)
  • brent: Brent's method (robust, always converges)
  • auto: Automatically selects best method

Feature Engineering for ML

b3quant provides comprehensive feature engineering for machine learning models:

from b3quant.features import OptionFeatureEngineer
import b3quant as pyb

# Get data
options = pyb.get_options(year=2024, month=11)
stocks = pyb.get_stocks(year=2024, month=11)

# Initialize feature engineer
fe = OptionFeatureEngineer(lookback_windows=[10, 30, 60])

# Add all features
options_with_features = fe.add_all_features(options, stocks)

Feature Categories

1. Moneyness Features

# Add moneyness features
options = fe.add_moneyness_features(options)

# Available features:
# - moneyness (S/K)
# - log_moneyness
# - is_itm, is_atm, is_otm (binary flags)

2. Time Features

# Add time-related features
options = fe.add_time_features(options)

# Available features:
# - day_of_week, day_of_month, month, quarter
# - is_month_end, is_quarter_end
# - sqrt_time, inv_sqrt_time
# - is_short_term, is_medium_term, is_long_term

3. Volatility Surface Features

# Add IV surface features
options = fe.add_volatility_features(options)

# Available features (for each lookback window):
# - iv_rank_{10,30,60}d: IV percentile rank
# - iv_percentile_{10,30,60}d: Rolling percentile
# - iv_skew: Put-Call IV difference

4. Market Microstructure Features

# Add market features (requires stocks data)
options = fe.add_market_features(options, stocks)

# Available features (for each lookback window):
# - realized_vol_{10,30,60}d: Historical volatility
# - momentum_{10,30,60}d: Price momentum
# - volume_ratio_{10,30,60}d: Volume vs average

5. Option-Specific Metrics

# Calculate option metrics
options = fe.calculate_option_metrics(options)

# Available metrics:
# - volume_pct: % of total underlying volume
# - put_call_ratio: Put volume / Call volume

# IV statistics by group
iv_stats = fe.calculate_iv_metrics(options)
# Returns: iv_mean, iv_std, iv_min, iv_max, iv_median, iv_range, iv_cv

Advanced Features

For sophisticated ML models, use AdvancedFeatureEngineer:

from b3quant.features import AdvancedFeatureEngineer
import b3quant as pyb

# Get data with Greeks (calculate using Black-Scholes first)
options = pyb.get_options(year=2024, month=11)
stocks = pyb.get_stocks(year=2024, month=11)

# Initialize advanced feature engineer
afe = AdvancedFeatureEngineer(
    lookback_windows=[10, 30, 60],
    regime_windows=[20, 50, 100]
)

# Add all advanced features
options_advanced = afe.add_all_advanced_features(
    options,
    stocks_df=stocks,
    benchmark_df=None  # Optional: IBOV or other benchmark
)

Available Advanced Features:

Greeks Exposure:

options = afe.add_greeks_exposure(options, stocks)
# - total_gamma_exposure, total_vega_exposure
# - max_gamma_strike
# - delta_weighted_volume
# - delta_hedged_value

Volatility of Volatility:

options = afe.add_volatility_of_volatility(options)
# - vol_of_vol_{10,30,60}d
# - iv_skewness_{10,30,60}d

Bollinger Bands:

options = afe.add_bollinger_bands(options, num_std=2.0)
# - bb_width_{10,30,60}d
# - bb_position_{10,30,60}d (0-1, position within bands)

RSI (Relative Strength Index):

options = afe.add_rsi(options, period=14)
# - rsi_{period}d (0-100)

Regime Detection:

options = afe.add_regime_features(options, benchmark_df)
# - regime_volatility_{20,50,100}d
# - regime_trend_strength_{20,50,100}d
# - regime_autocorr_{20,50,100}d
# - is_trending_{20,50,100}d (binary)
# - is_ranging_{20,50,100}d (binary)
# - is_volatile_{20,50,100}d (binary)
# - benchmark_corr_{20,50,100}d (if benchmark provided)

ML-Ready Dataset Example

from b3quant.features import OptionFeatureEngineer, AdvancedFeatureEngineer
import b3quant as pyb

# Fetch data
options = pyb.get_options(year=2024)
stocks = pyb.get_stocks(year=2024)

# Merge underlying prices
options = options.merge(
    stocks[['underlying', 'trade_date', 'close_price']],
    on=['underlying', 'trade_date'],
    suffixes=('', '_underlying')
)

# Engineer core features
fe = OptionFeatureEngineer()
options_ml = fe.add_all_features(options, stocks)

# Add advanced features
afe = AdvancedFeatureEngineer()
options_ml = afe.add_all_advanced_features(options_ml, stocks)

# Select features for modeling
feature_cols = [
    # Core features
    'moneyness', 'log_moneyness', 'time_to_maturity',
    'iv_rank_30d', 'iv_percentile_30d', 'iv_skew',
    'realized_vol_30d', 'momentum_30d',
    'volume_pct', 'put_call_ratio',
    # Advanced features
    'vol_of_vol_30d', 'bb_width_30d', 'rsi_14d',
    'regime_volatility_50d', 'is_trending_50d'
]

target = 'close_price'  # or 'implied_volatility'

X = options_ml[feature_cols].dropna()
y = options_ml.loc[X.index, target]

# Ready for sklearn, xgboost, pytorch, tensorflow, etc.

Examples

Example 1: Calculate Implied Volatility Surface

import b3quant as pyb
import pandas as pd

# Get PETR4 options
options = pyb.get_options(year=2024)
petr_opts = options[options['underlying'] == 'PETR'].copy()

# Filter valid data
petr_opts = petr_opts[
    (petr_opts['close_price'] > 0) &
    (petr_opts['volume'] > 0) &
    (petr_opts['days_to_maturity'] > 0)
]

# You would then calculate IV using Black-Scholes
# (requires additional libraries like scipy)
# ... your IV calculation here ...

Example 2: Analyze Option Volume by Strike

import b3quant as pyb
import matplotlib.pyplot as plt

options = pyb.get_options(year=2024)

# Filter PETR4 calls expiring in January 2025
petr_calls = options[
    (options['underlying'] == 'PETR') &
    (options['instrument_type'] == 'CALL') &
    (options['maturity_date'] >= '2025-01-01') &
    (options['maturity_date'] < '2025-02-01')
]

# Group by strike
volume_by_strike = petr_calls.groupby('strike_price')['volume'].sum()

# Plot
volume_by_strike.plot(kind='bar', figsize=(12, 6))
plt.title('PETR4 Call Options Volume by Strike (Jan 2025)')
plt.xlabel('Strike Price')
plt.ylabel('Volume (BRL)')
plt.show()

Development

See docs/DEVELOPMENT.md for detailed development setup and guidelines.

Quick start:

# Clone and setup
git clone https://github.com/renves/b3quant.git
cd b3quant
uv sync

# Run tests
uv run pytest -v

# Lint code
uv run ruff check b3quant/

CAPTCHA Handling

B3 sometimes requires CAPTCHA for downloads. If automatic download fails:

  1. Download manually from B3 website
  2. Save to cache directory (default: ./data/raw/)
  3. Parse the file directly:
from b3quant.parsers.cotahist import COTAHISTParser

parser = COTAHISTParser()
options = parser.parse_file('path/to/COTAHIST_A2024.TXT', instrument_filter='options')

Data Source

All data comes from B3 (Brasil, Bolsa, Balcão) official historical data files.

Official B3 Data Page: https://www.b3.com.br/en_us/market-data-and-indices/data-services/market-data/historical-data/equities/historical-quotes/

Available Data:

  • Yearly series: 1986 to current year (COTAHIST_A{YEAR}.ZIP)
  • Monthly series: Last 12 months (COTAHIST_M{MM}{YEAR}.ZIP)
  • Daily series: Current year (COTAHIST_D{DDMMYYYY}.ZIP)

Format Details:

  • Format: COTAHIST (fixed-width text format, 245 bytes per line)
  • Update frequency: Daily
  • Historical depth: Since 1986
  • License: Data is publicly available from B3
  • Encoding: Latin-1

Contributing

Contributions are welcome! See docs/CONTRIBUTING.md for detailed guidelines.

Quick summary:

  1. Fork the repository
  2. Create a feature branch (git checkout -b feat/amazing-feature)
  3. Make your changes and add tests
  4. Run tests and linter
  5. Commit with conventional commits format
  6. Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Disclaimer

This library is not affiliated with or endorsed by B3. It is an independent project for educational and research purposes.

Market data provided by B3 is subject to their terms of use. Please review B3's data policies before using this library for commercial purposes.

Credits

This library was inspired by and builds upon the work of:

  • rb3 by Wilson Freitas - R package for downloading B3 data
  • b3fileparser by Carlos Oliveira - Python parser for COTAHIST files

b3quant combines the functionality of both libraries into a unified, Pythonic interface with additional features and optimizations.

Citation

If you use this library in your research, please cite:

@software{b3quant2024,
  author = {Renan Alves},
  title = {b3quant: Python library for B3 market data},
  year = {2024},
  url = {https://github.com/renves/b3quant}
}

Documentation

Support

About

No description, website, or topics provided.

Resources

Readme

License

MIT license

Contributing

Contributing
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases 18

b3quant: v0.1.18 Latest
Jan 3, 2026
+ 17 releases

Packages

 
 
 

Contributors

Languages