HarmonicPatterns

HarmonicPatterns

A modern, high-performance Python library for detecting and predicting Harmonic Trading Patterns in financial charts. This library filters ZigZag movements and checks if their relationships (retracements/projections) align with standard Fibonacci ratios.

---

🌟 Features

• 9 Supported Patterns: Detects ABCD, Gartley, Bat, Alt-Bat, Butterfly, Crab, Deep Crab, Shark, and Cypher.
• Both Completed & Predicting Scanning: Not only matches historical patterns but also calculates potential D-point targets for incomplete patterns (predicting).
• Visualization: Seamlessly plots K-lines, moving averages, and detected pattern shapes using mplfinance and saves them to images.
• Modern Package Management: Powered by uv for reproducible builds and lightning-fast package resolution.

---

🛠️ Installation

1. Prerequisites (TA-Lib C Library)

This library depends on the python wrapper TA-Lib, which requires the underlying C library to be installed on your machine.

macOS

brew install ta-lib

Linux (Debian/Ubuntu)

Download and compile the source:

wget http://prdownloads.sourceforge.net/ta-lib/ta-lib-0.4.0-src.tar.gz
tar -xzf ta-lib-0.4.0-src.tar.gz
cd ta-lib/
./configure --prefix=/usr
make
sudo make install

---

2. Setting Up the Environment with uv

We recommend using uv (a fast alternative to pip/virtualenv) to manage the virtual environment and dependencies.

First, install uv if you haven’t already:

curl -LsSf https://astral.sh/uv/install.sh | sh

Initialize the virtual environment and install the package in editable mode:

# Create a virtual environment
uv venv

# Lock dependencies and install in editable mode (including requirements)
uv lock
uv pip install -e .

---

🚀 Quickstart

Here is a simple example to fetch BTC/USDT data and scan for patterns:

import ccxt
from harmonic_patterns import HarmonicDetector, kline_to_df

# 1. Initialize exchange & fetch data
exchange = ccxt.binance()
raw_candles = exchange.fetch_ohlcv('BTC/USDT', '1h', limit=200)
df = kline_to_df(raw_candles)

# 2. Instantiate the detector
# error_allowed=0.05 allows a 5% deviation from standard ratios
detector = HarmonicDetector(error_allowed=0.05, strict=True)

# 3. Detect patterns
# return_dt_idx=True returns timestamps instead of integer index points
patterns, predict_patterns = detector.search_patterns(
    df, 
    only_last=False, 
    last_n=4, 
    plot=True, 
    predict=True,
    return_dt_idx=True,
    save_fig_name='btc_patterns.png'
)

# 4. View results
print(f"Completed patterns: {len(patterns)}")
print(f"Predicting patterns: {len(predict_patterns)}")

---

📊 Supported Patterns & Fibonacci Ratios

Pattern	XA to AB (XAB)	AB to BC (ABC)	BC to CD (BCD)	XA to AD (XAD) / XC to CD (XCD)
ABCD	-	$0.382 \sim 0.886$	$1.13 \sim 1.168$	$AB = CD$ ($BCD / (1/ABC) = 1$)
Gartley	$0.618$	$0.382 \sim 0.886$	$1.13 \sim 1.168$	$XAD = 0.786$
Bat	$0.382 \sim 0.5$	$0.382 \sim 0.886$	$1.618 \sim 2.168$	$XAD = 0.886$
Alt-Bat	$0.382$	$0.382 \sim 0.886$	$2.0 \sim 3.168$	$XAD = 1.13$
Butterfly	$0.786$	$0.382 \sim 0.886$	$1.618 \sim 2.24$	$XAD = 1.27$
Crab	$0.382 \sim 0.618$	$0.382 \sim 0.886$	$2.618 \sim 3.618$	$XAD = 1.618$
Deep Crab	$0.886$	$0.382 \sim 0.886$	$2.24 \sim 3.618$	$XAD = 1.618$
Shark	$0.5 \sim 0.886$	$1.13 \sim 1.618$	$1.618 \sim 2.24$	$XAD = 0.886 \sim 1.13$
Cypher	$0.382 \sim 0.786$	$1.272 \sim 1.414$	-	$XCD = 0.786$ or $XAD = 0.786$

---

🏃 Running Examples

We provide a complete, real-time scanning example using OKX market data:

# Run real-time OKX scanning for BTC & ETH
python examples/run_okx_example.py

This will output scan results directly to the terminal, and save visual charts (K-lines with harmonic lines) under the ./data folder.

---

📂 Project Structure

• src/: Core library code.
  • harmonic_patterns.py: Core detector logic, ZigZag algorithms, and plot functions.
  • harmonic_functions.py: Wrapper module matching legacy imports.
  • settings.py: Scanning variables and default parameters.
• examples/: Handy runnable scripts and notebooks.
  • run_okx_example.py: Real-time scanner for OKX.
  • run_detect.py: Multi-threaded polling scanning script.
• pyproject.toml: Modern PEP-518 project definition.