Why polars-anta?

Quantitative researchers and financial engineers have historically faced a trade-off between installation complexity and computation speed:

polars-anta eliminates this compromise by implementing financial indicators directly as Polars Expressions (pl.Expr). Computations are pushed into Polars' multi-threaded Rust engine, eliminating Python GIL bottlenecks without requiring any C compilers.

Performance Benchmark (1,000,000 Rows)

Benchmarks measured across 1,000,000 bars (single-asset random walk) and a 500 tickers x 2,000 bars universe. Computations executed on Polars' multi-threaded Rust query engine vs single-threaded Pandas:

Reproduce benchmarks locally using pip install -e .[benchmark] followed by python benchmarks/benchmark_speed.py.

Key Features

• Dual Expression Namespaces: Call indicators via pl.col("close").anta.<indicator>() or the classic shorthand pl.col("close").ta.<indicator>().
• Expression-First Design: Compose indicators directly within .select(), .with_columns(), and lazy query plans.
• Parallel Multi-Asset Computation: Compute indicators across thousands of tickers in parallel using Polars' native window context:

  pl.col("close").anta.rsi(14).over("symbol")
  

• Structured Outputs: Multi-component indicators (MACD, Bollinger Bands, Stochastic, Keltner Channels) return typed pl.Struct expressions that maintain clean schemas.
• Zero-C Compilation: pip install polars-anta works out of the box across Linux, macOS, and Windows.

Installation

pip install polars-anta

Requirements: Python >= 3.10, Polars >= 1.0.0

Quickstart

1. Modern Chained Expression Syntax (.anta or .ta)

import polars as pl
import polars_anta  # Registers both `.anta` and `.ta` expression namespaces

df = pl.DataFrame(
    {
        "open": [100.0, 102.0, 101.5, 103.0, 105.0],
        "high": [101.5, 103.5, 103.0, 104.5, 106.5],
        "low": [98.8, 100.8, 100.3, 101.8, 103.8],
        "close": [100.0, 102.0, 101.5, 103.0, 105.0],
        "volume": [1000, 1050, 1100, 1150, 1200],
    }
)

# Compute indicators directly on expressions (using .anta or .ta)
result = df.with_columns(
    pl.col("close").anta.sma(period=3).alias("sma_3"),
    pl.col("close").anta.rsi(period=3).alias("rsi_3"),
    pl.col("close").anta.bollinger_bands(period=3).alias("bb"),
)

# Unnest structured outputs as needed
print(result.unnest("bb"))

2. Market-Wide Multi-Asset Parallel Execution

Running technical indicators across an entire investment universe is a common bottleneck. In polars-anta, Polars distributes evaluation across all CPU cores automatically:

# Compute indicators across thousands of symbols simultaneously in Rust
universe.with_columns(
    [
        pl.col("close").anta.sma(20).over("ticker").alias("sma_20"),
        pl.col("close").anta.rsi(14).over("ticker").alias("rsi_14"),
        pl.col("close").anta.macd().over("ticker").alias("macd"),
        pl.col("close").anta.obv("volume").over("ticker").alias("obv"),
    ]
)

3. Interactive Jupyter Notebooks

Run our interactive tutorials directly in your browser with zero setup:

Notebook	Description	Interactive Demo
01 - Quickstart Tutorial	Comprehensive walk-through of expression syntax, Struct unnesting, and plotting.
02 - Multi-Asset Portfolio Scan	Real-world 500-ticker breakout screening in 50 milliseconds using .over("ticker").

Supported Indicators (100 Indicators)

Trend & Overlays (25 Indicators)

• sma(close, period=10): Simple Moving Average
• ema(close, period=10): Exponential Moving Average
• wma(close, period=10): Linearly-Weighted Moving Average
• hma(close, period=10): Hull Moving Average
• dema(close, period=10): Double Exponential Moving Average
• tema(close, period=10): Triple Exponential Moving Average
• zlema(close, period=10): Zero-Lag Exponential Moving Average
• vortex(high, low, close, period=14): Vortex Indicator (vi_plus, vi_minus)
• adx(high, low, close, period=14): Average Directional Index (adx, plus_di, minus_di)
• aroon(high, low, period=25): Aroon Oscillator & Indicator (up, down, oscillator)
• chop(high, low, close, period=14): Choppiness Index (Trend vs Consolidation)
• t3(close, period=5, vfactor=0.7): Tillson T3 Moving Average
• alma(close, period=9, offset=0.85, sigma=6.0): Arnaud Legoux Moving Average (Gaussian-filtered)
• mass_index(high, low, fast_period=9, slow_period=25): Mass Index (Reversal predictor)
• supertrend(high, low, close, period=10, multiplier=3.0): SuperTrend Volatility Stop (supertrend, direction)
• psar(high, low, af_step=0.02, af_max=0.2): Parabolic Stop and Reverse (Trailing SAR)
• kama(close, period=10, fast_span=2, slow_span=30): Kaufman's Adaptive Moving Average
• mcginley_dynamic(close, period=14, k=0.6): McGinley Dynamic Average
• vidya(close, period=9, cmo_period=9): Variable Index Dynamic Average (Chande's VIDYA)
• center_of_gravity(close, period=10): Ehlers Center of Gravity Oscillator
• ichimoku(high, low, close, tenkan=9, kijun=26, senkou_b=52, displacement=26): Ichimoku Kinko Hyo Cloud
• pivot_points(high, low, close): Floor Trader Standard Pivot Points (pivot, r1-r3, s1-s3)
• fibonacci_pivots(high, low, close): Fibonacci Pivot Points (38.2%, 61.8%, 100%)
• camarilla_pivots(high, low, close): Camarilla Intraday Pivot Points (r1-r4, s1-s4)
• woodie_pivots(open, high, low, close): Woodie's Heavy-Close Pivot Points

Momentum (19 Indicators)

• rsi(close, period=14): Relative Strength Index (Wilder's smoothing)
• macd(close, fast=12, slow=26, signal=9): Moving Average Convergence Divergence (macd, signal, histogram)
• stochastic(high, low, close, k=14, d=3): Stochastic Oscillator (%K, %D)
• williams_r(high, low, close, period=14): Williams %R
• cci(high, low, close, period=20): Commodity Channel Index
• roc(close, period=10): Rate of Change (Momentum %)
• stoch_rsi(close, rsi_period=14, stoch_period=14): Stochastic RSI (%K, %D)
• awesome_oscillator(high, low, fast=5, slow=34): Awesome Oscillator (AO)
• tsi(close, long_period=25, short_period=13): True Strength Index (TSI)
• cmo(close, period=14): Chande Momentum Oscillator
• ppo(close, fast=12, slow=26, signal=9): Percentage Price Oscillator (ppo, signal, histogram)
• coppock(close, wma_period=10, fast_roc=11, slow_roc=14): Coppock Curve
• ultimate_oscillator(high, low, close, p1=7, p2=14, p3=28): Ultimate Oscillator (UO)
• fisher_transform(high, low, period=10): Ehlers Fisher Transform (fisher, signal)
• rvi(open, high, low, close, period=10): Relative Vigor Index (rvi, signal)
• kst(close): "Know Sure Thing" Oscillator (kst, signal)
• dpo(close, period=20): Detrended Price Oscillator
• bop(open, high, low, close): Balance of Market Power
• cfo(close, period=14): Chande Forecast Oscillator

Volatility & Channels (18 Indicators)

• true_range(high, low, close): True Range
• atr(high, low, close, period=14): Average True Range
• bollinger_bands(close, period=20, num_std=2.0): Bollinger Bands (Upper, Middle, Lower, Bandwidth, %B)
• keltner_channels(high, low, close, ema=20, atr=10, multiplier=2.0): Keltner Channels
• donchian_channels(high, low, period=20): Donchian Channels (Upper, Middle, Lower)
• stddev(close, period=20): Rolling Standard Deviation
• ulcer_index(close, period=14): Ulcer Index (Downside Drawdown Risk)
• chandelier_exit(high, low, close, period=22, multiplier=3.0): Chandelier Exit (long_stop, short_stop)
• natr(high, low, close, period=14): Normalized Average True Range (%)
• historical_volatility(close, period=20, trading_days=252): Annualized Log Historical Volatility
• parkinson_volatility(high, low, period=20, trading_days=252): Parkinson Extreme-Value Volatility
• garman_klass_volatility(open, high, low, close, period=20, trading_days=252): Garman-Klass Efficient Volatility
• chaikin_volatility(high, low, period=10, roc_period=10): Chaikin Volatility
• donchian_width(high, low, period=20): Donchian Channel Width (%)
• envelopes(close, period=20, percentage=0.025): Moving Average Envelopes (upper, middle, lower)
• keltner_width(high, low, close, ema=20, atr=10, multiplier=2.0): Normalized Keltner Channel Width (%)
• bollinger_delta(close, period=20, num_std=2.0): Bollinger Band Spread ($2 \cdot k \cdot \sigma$)
• starc_bands(high, low, close, period=20, atr_period=15, multiplier=2.0): STARC Bands (Stoller Average Range Channels)

Volume (14 Indicators)

• obv(close, volume): On-Balance Volume
• vwap(high, low, close, volume, period=None): Volume Weighted Average Price (Cumulative or Rolling)
• vwma(close, volume, period=20): Volume Weighted Moving Average
• cmf(high, low, close, volume, period=20): Chaikin Money Flow
• adl(high, low, close, volume): Accumulation/Distribution Line
• mfi(high, low, close, volume, period=14): Money Flow Index
• eom(high, low, volume, period=14): Ease of Movement (EMV)
• klinger(high, low, close, volume, fast=34, slow=55, signal=13): Klinger Volume Oscillator (klinger, signal)
• nvi(close, volume): Negative Volume Index
• pvi(close, volume): Positive Volume Index
• pvt(close, volume): Price Volume Trend
• force_index(close, volume, period=13): Elder's Force Index
• vpci(close, volume, short_period=5, long_period=20): Volume Price Confirmation Indicator
• vr(close, volume, period=26): Volume Ratio (Bullish vs Bearish volume)

Quantitative Statistics & Regression (12 Indicators)

• zscore(close, period=20): Rolling Z-Score standardized anomaly score
• percentile_rank(close, period=20): Rolling Percentile Rank $[0.0, 1.0]$
• skewness(close, period=20): Rolling Sample Skewness (asymmetry metric)
• kurtosis(close, period=20): Rolling Excess Kurtosis (tail-risk metric)
• rolling_cov(x, y, period=20): Rolling Sample Covariance between two series
• rolling_corr(x, y, period=20): Rolling Pearson Correlation $[-1.0, 1.0]$
• rolling_beta(asset_ret, bench_ret, period=60): Rolling CAPM Market Beta ($\text{Cov} / \text{Var}$)
• rolling_alpha(asset_ret, bench_ret, rf=0.0, period=60): Rolling Jensen's Risk-Adjusted Alpha
• linear_regression_slope(close, period=14): Vectorized OLS Linear Regression Slope
• linear_regression_intercept(close, period=14): Vectorized OLS Linear Regression Intercept
• linear_regression_angle(close, period=14): Linear Regression Trend Angle in degrees
• linear_regression_forecast(close, period=14): Linear Regression Current/Next-Bar Value Forecast

Alpha Factor Primitives (12 Indicators)

• ts_rank(close, period=10): Time-Series Rank (WorldQuant Alpha 101 primitive)
• ts_argmax(close, period=10): Time-Series ArgMax (lookback bars ago of window peak)
• ts_argmin(close, period=10): Time-Series ArgMin (lookback bars ago of window trough)
• decay_linear(close, period=10): Linearly Weighted Decay moving transform
• ts_zscore(close, period=20): Time-Series Standardized Normalization
• returns_divergence(close, volume, period=14): Price vs Volume Momentum Divergence
• hl_ratio(high, low): Log Intraday Range Expansion $\ln(\text{High} / \text{Low})$
• co_ratio(close, open): Log Intraday Sentiment Drift $\ln(\text{Close} / \text{Open})$
• volume_weighted_momentum(close, volume, period=10): Momentum amplified by relative volume
• relative_spread(high, low, close): Normalized Bar Range $(\text{High} - \text{Low}) / \text{Close}$
• twap(high, low, close, period=20): Rolling Time-Weighted Average Price
• entropy(close, period=20, bins=5): Rolling Shannon Information Entropy (Regime detector)

Architecture & Performance

polars-anta is designed around three architectural pillars:

1. Arrow Memory Layout: Avoids memory copying between indicator transformations.
2. SIMD-Accelerated Vectorization: Executes through Polars' optimized analytical kernels.
3. Query Optimization: Integrates with Polars' expression engine for predicate pushdown and dead-code elimination.

Development & Testing

# Clone the repository
git clone https://github.com/huuan26/polars-anta.git
cd polars-anta

# Install in editable mode with development dependencies
pip install -e .[dev]

# Run the test suite
pytest -v

# Run linter and formatter checks
ruff check .
ruff format --check .

Author & Community

• Author: An Phan (Phan Bùi Hữu Ân) - phanbuihuuan261103@gmail.com
• GitHub: @huuan26
• Discussion & Feature Requests: GitHub Discussions

License

Licensed under the Apache License, Version 2.0. See LICENSE for details.