ta-lib

ta-lib — C-Optimized Technical Analysis

TA-Lib (Technical Analysis Library) is a C library with a Python wrapper providing 150+ technical analysis functions and 61 candlestick pattern recognition functions. It is the industry standard for performance-critical indicator computation, used in production trading systems where pandas-ta or pure-Python alternatives are too slow.

What TA-Lib Is

TA-Lib was originally written in C for financial market data analysis. The Python wrapper (TA-Lib on PyPI, imported as talib) provides:

• 150+ indicator functions across overlap, momentum, volume, volatility, cycle, and math categories
• 61 candlestick pattern recognition functions — the most comprehensive pattern library available
• C-speed computation — 10-100x faster than pure-Python equivalents on large datasets
• Two APIs: a function API (pass arrays directly) and an abstract API (pass dict of arrays)
• NumPy native — all inputs and outputs are NumPy arrays

Installation

TA-Lib requires the underlying C library to be installed first:

# macOS
brew install ta-lib
uv pip install TA-Lib numpy pandas

# Ubuntu/Debian
sudo apt-get install -y ta-lib
uv pip install TA-Lib numpy pandas

# From source (any platform)
wget https://github.com/ta-lib/ta-lib/releases/download/v0.6.4/ta-lib-0.6.4-src.tar.gz
tar -xzf ta-lib-0.6.4-src.tar.gz
cd ta-lib-0.6.4
./configure --prefix=/usr/local
make && sudo make install
uv pip install TA-Lib numpy pandas

If the C library is not installed, import talib will fail with an ImportError. The scripts in this skill include fallback logic for environments without TA-Lib installed.

When to Use TA-Lib vs pandas-ta

Criterion	TA-Lib	pandas-ta
Speed	C-optimized, 10-100x faster	Pure Python, slower on large data
Candlestick patterns	61 built-in patterns	Limited pattern support
Installation	Requires C library	pip install only
API style	NumPy arrays	DataFrame .ta accessor
Indicator count	150+	130+
Streaming	Single-value update possible	Recompute entire series
Dependencies	C lib + numpy	pandas only

Use TA-Lib when:

• Processing millions of bars or running backtests at scale
• You need candlestick pattern recognition (TA-Lib is unmatched here)
• You are building a production pipeline where latency matters
• You need cycle indicators (Hilbert Transform family)

Use pandas-ta when:

• You want DataFrame-native convenience
• Installation simplicity matters (no C dependency)
• You need indicators not in TA-Lib (pandas-ta has some extras)

Quick Start

import numpy as np
import talib

# Create sample data
close = np.random.randn(100).cumsum() + 50
high = close + np.abs(np.random.randn(100))
low = close - np.abs(np.random.randn(100))
open_ = close + np.random.randn(100) * 0.5
volume = np.random.randint(1000, 10000, 100).astype(float)

# Function API — pass arrays directly
rsi = talib.RSI(close, timeperiod=14)
macd, signal, hist = talib.MACD(close, fastperiod=12, slowperiod=26, signalperiod=9)
upper, middle, lower = talib.BBANDS(close, timeperiod=20, nbdevup=2, nbdevdn=2)
atr = talib.ATR(high, low, close, timeperiod=14)

# Candlestick patterns — return +100 (bullish), -100 (bearish), or 0
doji = talib.CDLDOJI(open_, high, low, close)
hammer = talib.CDLHAMMER(open_, high, low, close)
engulfing = talib.CDLENGULFING(open_, high, low, close)

Function API vs Abstract API

Function API (Recommended)

Call functions directly with NumPy arrays:

import talib

rsi = talib.RSI(close, timeperiod=14)
sma = talib.SMA(close, timeperiod=20)
upper, mid, lower = talib.BBANDS(close)

Abstract API

Pass a dictionary of arrays and get results by name:

from talib import abstract

inputs = {"open": open_, "high": high, "low": low, "close": close, "volume": volume}

# Call by function name
rsi = abstract.RSI(inputs, timeperiod=14)
macd = abstract.MACD(inputs)  # returns (macd, signal, hist)

The abstract API is useful for dynamic indicator selection (e.g., looping over a list of indicator names).

Function Groups

TA-Lib organizes functions into these groups:

Overlap Studies

Moving averages and envelope indicators that overlay price charts.

sma = talib.SMA(close, timeperiod=20)
ema = talib.EMA(close, timeperiod=12)
upper, mid, lower = talib.BBANDS(close, timeperiod=20, nbdevup=2, nbdevdn=2)
sar = talib.SAR(high, low, acceleration=0.02, maximum=0.2)
mama, fama = talib.MAMA(close, fastlimit=0.5, slowlimit=0.05)

Momentum Indicators

Oscillators and trend-strength measures.

rsi = talib.RSI(close, timeperiod=14)
macd, signal, hist = talib.MACD(close, fastperiod=12, slowperiod=26, signalperiod=9)
slowk, slowd = talib.STOCH(high, low, close)
cci = talib.CCI(high, low, close, timeperiod=14)
willr = talib.WILLR(high, low, close, timeperiod=14)
adx = talib.ADX(high, low, close, timeperiod=14)
mfi = talib.MFI(high, low, close, volume, timeperiod=14)

Volume Indicators

Volume-based analysis functions.

obv = talib.OBV(close, volume)
ad = talib.AD(high, low, close, volume)
adosc = talib.ADOSC(high, low, close, volume, fastperiod=3, slowperiod=10)

Volatility Indicators

Measures of price variability.

atr = talib.ATR(high, low, close, timeperiod=14)
natr = talib.NATR(high, low, close, timeperiod=14)
trange = talib.TRANGE(high, low, close)

Pattern Recognition (Candlestick)

61 functions that detect candlestick patterns. All return integer arrays:

• +100 = bullish pattern detected
• -100 = bearish pattern detected
• 0 = no pattern

# Single patterns
doji = talib.CDLDOJI(open_, high, low, close)
hammer = talib.CDLHAMMER(open_, high, low, close)
engulfing = talib.CDLENGULFING(open_, high, low, close)

# Scan all 61 patterns at once
candle_names = talib.get_function_groups()["Pattern Recognition"]
for name in candle_names:
    func = getattr(talib, name)
    result = func(open_, high, low, close)
    hits = np.nonzero(result)[0]
    if len(hits) > 0:
        print(f"{name}: {len(hits)} detections")

See references/candlestick_patterns.md for the full list of 61 patterns with reliability ratings and crypto relevance.

Math Transform & Math Operators

Mathematical functions (sin, cos, ln, etc.) and operators (add, sub, mult, div) on arrays. Rarely used directly but available.

Crypto Considerations

24/7 Markets

• Candlestick patterns designed for traditional markets with opening/closing gaps may behave differently on crypto's continuous markets
• Gap-based patterns (morning star, evening star) are less reliable without session gaps
• Body-ratio patterns (doji, hammer, engulfing) still work well on any timeframe

Timeframe Selection

• 1m-5m: Patterns are noisy; combine with volume confirmation
• 15m-1h: Good for intraday signals on high-cap tokens
• 4h-1d: Most reliable for pattern recognition
• Tip: Higher timeframes produce fewer but more reliable pattern signals

NaN Handling

TA-Lib returns NaN for the initial lookback period of each indicator. Always account for this:

rsi = talib.RSI(close, timeperiod=14)
# First 14 values will be NaN
valid_rsi = rsi[~np.isnan(rsi)]

Solana Token Data

When using TA-Lib with Solana token OHLCV data:

• Ensure arrays are float64 dtype — TA-Lib requires this
• Sort by timestamp ascending before passing to TA-Lib
• Handle gaps in low-liquidity token data before computing indicators

# Convert to float64 for TA-Lib compatibility
close = df["close"].values.astype(np.float64)
high = df["high"].values.astype(np.float64)
low = df["low"].values.astype(np.float64)

Integration with Other Skills

With pandas-ta

pandas-ta can use TA-Lib as a backend when installed, getting C-speed through the pandas-ta API: