fastbt 0.9.0

A simple framework for fast and dirty backtesting

Introduction

fastbt is a simple and dirty way to do backtests based on end of day data, especially for day trading. The main purpose is to provide a simple framework to weed out bad strategies so that you could test and improve your better strategies further.

It is based on the assumption that you enter into a position based on some pre-defined rules for a defined period and exit either at the end of the period or when stop loss is triggered. See the rationale for this approach and the built-in assumptions. fastbt is rule-based and not event-based.

If your strategy gets you good results, then check them with a full featured backtesting framework such as zipline or backtrader to verify your results. If your strategy fails, then it would most probably fail in other environments.

This is alpha

Most of the modules are stand alone and you could use them as a single file. See embedding for more details

Features

• Create your strategies in Microsoft Excel
• Backtest as functions so you can parallelize
• Try different simulations
• Run from your own datasource or a database connection.
• Run backtest based on rules
• Add any column you want to your datasource as formulas

Installation

fastbt requires python >=3.6 and can be installed via pip

pip install fastbt

Quickstart

Fastbt assumes your data have the following columns (rename them in case of other names)

• timestamp
• symbol
• open
• high
• low
• close
• volume

from fastbt.rapid import *
backtest(data=data)

would return a dataframe with all the trades.

And if you want to see some metrics

metrics(backtest(data=data))

You now ran a backtest without a strategy! By default, the strategy buys the top 5 stocks with the lowest price at open price on each period and sells them at the close price at the end of the period.

You can either specify the strategy by way of rules (the recommended way) or create your strategy as a function in python and pass it as a parameter

backtest(data=data, strategy=strategy)

If you want to connect to a database, then

from sqlalchemy import create_engine
engine = create_engine('sqlite:///data.db')
backtest(connection=engine, tablename='data')

And to SELL instead of BUY

backtest(data=data, order='S')

Let's implement a simple strategy.

BUY the top 5 stocks with highest last week returns

Assuming we have a weeklyret column,

backtest(data=data, order='B', sort_by='weeklyret', sort_mode=False)

We used sort_mode=False to sort them in descending order.

If you want to test this strategy on a weekly basis, just pass a dataframe with weekly frequency.

See the Introduction notebook in the examples directory for an in depth introduction.

Embedding

Since fastbt is a thin wrapper around existing packages, the following files can be used as standalone without installing the fastbt package

• datasource
• utils
• loaders

Copy these files and just use them in your own modules.

========= History

v0.9.0

• Code formatting and whitespace cleanup with ruff/black
• Removed label field from Leg dataclass
• Bumped collation size limit from 100MB to 1000MB
• Dropped Vortex loader test coverage (see #13)

v0.8.0

• NEW: Added statistical testing skill v1.0-alpha with comprehensive test scripts
  • Includes hypothesis testing for trading strategy validation
  • Supports benchmark comparison, conditional analysis, and temporal validation
  • Added out-of-sample testing capabilities
• ENHANCED: Simulation module improvements
  • Refactored generators into Time and Sequence engines
  • Added IID lognormal distribution support
  • New generator modes and parameters for more flexible simulation
  • Fixed price stagnation in tick/quote generators
• Added regression tests for tick generator price movement
• Updated documentation for simulation module with new examples
• Code quality improvements with pre-commit hooks

v0.7.0 (BREAKING)

• BREAKING: Migrated to Pydantic v2.0 compatibility (requires pydantic>=2.0.0)
• Updated all BaseModel classes to use model_config instead of deprecated Config class
• Replaced root_validator with model_validator(mode='before')
• Added default values for Optional fields as required by Pydantic v2
• Enhanced option chain simulation capabilities
• Added correlated data simulation features
• Improved URL pattern matching functionality
• Code formatting and linting improvements with ruff and black
• Added load-data skill for efficient data discovery and loading (includes peek_file, efficient_load, collate_data, and normalize_json)

v0.6.0

• New methods added to TradeBook object
• mtm - to calculate mtm for open positions
• clear - to clear the existing entries
• helper attributes for positions
• order_fill_price method added to utils to simulate order quantity

v0.5.1

• Simple bug fixes added

v0.5.0

• OptionExpiry class added to calculate option payoffs based on expiry

v0.4.0

• Brokers module deprecation warning added
• Options module revamped

v0.3.0 (2019-03-15)

• More helper functions added to utils
• Tradebook class enhanced
• A Meta class added for event based simulation

v0.2.0 (2018-12-26)

• Backtest from different formats added
• Rolling function added

v0.1.0. (2018-10-13)

• First release on PyPI