vectorbt

The Candlestick Patterns app is a Dash-based application designed to visualize and backtest candlestick patterns. Key capabilities include:

• Data & Detection: Fetches market data via yfinance and detects patterns using TA-Lib.
• Strategy Customization: Choose entry/exit patterns, override candle settings, or specify signals manually.
• Backtesting: Uses VectorBT to backtest signals.
• Visualization: Uses Plotly to display OHLCV, signals, orders, trades, and portfolio value.
• Analysis: Displays performance metrics (e.g., Sharpe ratio) and compares strategies against 'buy & hold' and random trading.
• UI: Responsive interface built with Dash Bootstrap Components.

vectorbt is distributed under a fair-code model using the Apache 2.0 with Commons Clause license.

• Permitted: Individuals and organizations can use the source code and software for free.
• Restricted: You are not allowed to sell products or services that consist primarily of this software.

If you require a license exception, you must contact the author at olegpolakow@vectorbt.pro.

Note: Installing optional dependencies may be subject to more restrictive licenses than the core vectorbt package.

The vectorbt software is licensed under the Apache 2.0 license, but it is subject to the Commons Clause License Condition v1.0.

Under this condition, you are granted the rights provided by the Apache 2.0 license except for the right to Sell the software.

Definition of "Sell": Practicing any or all of the rights granted under the license to provide to third parties, for a fee or other consideration (including hosting or consulting/support services), a product or service whose value derives, entirely or substantially, from the functionality of the software.

If you use the software, any license notice or attribution required by the Apache 2.0 license must also include this Commons Clause License Condition notice.

This project is licensed under the GNU Affero General Public License (AGPL) version 3.

Key implications for users and developers:

• Network Interaction Requirement: If you modify the program and run it on a network server, you must provide the modified source code to the users of that server.
• Copyleft: Any modified versions or works based on this program must also be licensed under the AGPL v3.
• No Warranty: The software is provided "as is" without any warranty of any kind.
• Limitation of Liability: The copyright holders and contributors are not liable for any damages arising from the use or inability to use the program.

Official project resources:

• Source repository: The GitHub repository contains the source code, examples, issues, discussions, and release history.
• VectorBT PRO: For advanced features beyond the open-source version, visit vectorbt.pro.

You can perform retrospective analysis on trades (entry/exit) and positions. Additionally, you can perform drawdown analysis on any numeric time series using .vbt.drawdowns.

# Get projected returns of buy orders
>>> entry_trades = vbt.Portfolio.from_random_signals(price, n=5).entry_trades
>>> returns_pd = entry_trades.returns.to_pd()

# Plot deepest price dips
>>> price.vbt.drawdowns.plot(top_n=3).show()

VectorBT provides a custom accessor .vbt on top of Pandas objects to access compiled, high-performance versions of common operations like mapping, reducing, rolling, grouping, and resampling. These operations use NumPy, Numba, and optional Rust kernels for significant speedups over standard Pandas.

import vectorbt as vbt
import pandas as pd
import numpy as np
from numba import njit

big_ts = pd.DataFrame(np.random.uniform(size=(1000, 1000)))

@njit
def vbt_zscore_nb(i, col, x):
    return (x[-1] - np.mean(x)) / np.std(x)

# Use the .vbt accessor for accelerated rolling apply
%timeit big_ts.vbt.rolling_apply(2, vbt_zscore_nb)

To generate comprehensive Markdown reports, run bench_matrix.py. This produces three files by default: BENCHMARKS.md, BENCHMARKS_NUMBA.md, and BENCHMARKS_RUST.md. You can specify a custom output path for the speedup report using --output.

python benchmarks/bench_matrix.py --output benchmarks/BENCHMARKS_LOCAL.md

To add new benchmarks, add cases to make_cases in bench_engine.py after ensuring Rust implementations and dispatch tests are stable.

Requirements for new cases:

• Must be deterministic.
• Must be representative of public dispatch behavior.
• Must be cheap enough to run across the full matrix.
• Must be explicit about cases where parity cannot be exact.
• Must be tagged if they should be excluded from the core suite.

After adding cases, run a targeted check:

python benchmarks/bench_engine.py --pattern <subpackage-or-function> --check

Then regenerate reports:

python benchmarks/bench_matrix.py