signalflow-ta - Technical Analysis¶

signalflow-ta extends SignalFlow with 199+ technical analysis indicators organized into 8 modules. Each indicator is a standard Feature class that integrates directly with FeaturePipeline and the component registry.

Installation¶

pip install signalflow-ta

Requires signalflow-trading >= 0.5.0.

Indicator Modules¶

Module	Count	Description
Momentum	18	RSI, MACD, Stochastic, ROC, CCI, Williams %R, and kinematics analogs
Overlap	26	SMA, EMA, DEMA, TEMA, HMA, KAMA, ALMA, JMA, and price transforms
Volatility	40	ATR, Bollinger, Keltner, Donchian bands, and energy-based indicators
Volume	16	OBV, A/D, MFI, CMF, KVO, and market dynamics analogs
Trend	22	ADX, Aroon, Supertrend, PSAR, Ichimoku, and physics-based strength
Statistics	73	Dispersion, distribution, memory, cycles, complexity, DSP, regression
Performance	2	Log return, percent return
Divergence	2	RSI divergence, MACD divergence detectors

Quick Start¶

Using Individual Indicators¶

import signalflow.ta as ta
from signalflow.feature import FeaturePipeline

# Create indicators
rsi = ta.RsiMom(period=14)
bbands = ta.BollingerVol(period=20, num_std=2.0)
atr = ta.AtrVol(period=14)

# Use in pipeline
pipeline = FeaturePipeline(features=[rsi, bbands, atr])
features_df = pipeline.compute(df)

Using Preset Factories¶

Preset factories provide curated indicator sets:

from signalflow.ta.pipes import (
    momentum_core_pipe,
    volatility_bands_pipe,
    all_ta_pipe,
)
from signalflow.feature import FeaturePipeline

# Compose a custom set
pipeline = FeaturePipeline(features=[
    *momentum_core_pipe(normalized=True),
    *volatility_bands_pipe(),
])

# Or use all indicators at once
full_pipeline = FeaturePipeline(features=all_ta_pipe(normalized=True))

Available factories:

Factory	Indicators
`smoothers_pipe()`	SMA, EMA, WMA, HMA, DEMA, TEMA, ...
`momentum_core_pipe()`	RSI, ROC, MOM, CMO
`momentum_oscillators_pipe()`	Stochastic, StochRSI, Williams %R, CCI, UO, AO
`momentum_pipe()`	All momentum indicators
`volatility_range_pipe()`	True Range, ATR, NATR
`volatility_bands_pipe()`	Bollinger, Keltner, Donchian, AccBands
`volatility_pipe()`	All volatility indicators
`volume_cumulative_pipe()`	OBV, A/D, PVT, NVI, PVI
`volume_oscillators_pipe()`	MFI, CMF, EFI, EOM, KVO
`volume_pipe()`	All volume indicators
`trend_strength_pipe()`	ADX, Aroon, Vortex, VHF, CHOP
`trend_stops_pipe()`	PSAR, Supertrend, Chandelier, HiLo, CKSP
`trend_pipe()`	All trend indicators
`stat_dispersion_pipe()`	Variance, Std, MAD, Z-Score, CV, Range
`stat_cycle_pipe()`	Instantaneous Amplitude, Phase, Frequency
`stat_pipe()`	All statistical indicators
`performance_pipe()`	LogReturn, PctReturn
`divergence_pipe()`	RSI divergence, MACD divergence
`all_ta_pipe()`	All 199+ indicators

Indicator Architecture¶

Every indicator follows a consistent pattern:

from dataclasses import dataclass
import signalflow as sf
from signalflow.feature.base import Feature

@dataclass
@sf.feature("momentum/rsi")
class RsiMom(Feature):
    period: int = 14

    requires = ["close"]                # input columns
    outputs = ["rsi_{period}"]          # output column template

    def compute_pair(self, df: pl.DataFrame) -> pl.DataFrame:
        # Pure NumPy computation on single pair
        ...

    @property
    def warmup(self) -> int:
        return self.period * 10         # minimum stable bars

Key design principles:

Deterministic: Pure lookback computation, always reproducible
Polars-native: DataFrame in/out, native rolling operations
Warmup support: Each indicator declares minimum bars for stable output
Normalization: Optional normalized=True parameter for value rescaling

Normalization¶

Indicators support two normalization strategies:

Bounded indicators (RSI, Williams %R, CCI):

rsi = ta.RsiMom(period=14, normalized=True)
# Rescales [0, 100] → [0, 1]

Unbounded indicators (MACD, SMA, ROC):

sma = ta.SmaSmooth(period=20, normalized=True)
# Rolling z-score normalization

AutoFeatureNormalizer¶

For automatic per-feature normalization:

from signalflow.ta.auto_norm import AutoFeatureNormalizer

normalizer = AutoFeatureNormalizer(window=256, warmup=256)
df_normalized = normalizer.fit_transform(df)

# Save for production use
normalizer.artifact.save("normalizer.json")

The normalizer analyzes each feature's distribution (skewness, outlier ratio, CV) and selects the best method: rolling robust, rolling z-score, rolling rank, rolling winsorization, or signed log1p.

Physics-Based Indicators¶

signalflow-ta includes novel indicators that model market dynamics using physics analogs:

Volatility Energy¶

Kinetic Energy: KE = 0.5 * v² - measures price velocity
Potential Energy: PE = (ln(Close) - ln(MA))² - measures deviation
Total Energy: TE = KE + PE - overall market energy
Temperature: T = KE / degrees_of_freedom - thermal regime
Free Energy: F = E - T*S - actionable energy (entropy-adjusted)

Volume Dynamics¶

Market Force: F = volume × acceleration - buying/selling pressure
Impulse: J = Σ(F × dt) - accumulated force
Market Momentum: p = volume × velocity - directional volume
Market Capacitance: C = volume / ΔPrice - liquidity measure

Trend Physics¶

Viscosity: Resistance to velocity change
Reynolds Number: Laminar vs turbulent market regime
Impedance: Z = V/I electrical analogy for trend resistance

Module Reference¶

Momentum¶

Class	Component Name	Description
`RsiMom`	`momentum/rsi`	Relative Strength Index
`RocMom`	`momentum/roc`	Rate of Change
`MomMom`	`momentum/mom`	Momentum
`CmoMom`	`momentum/cmo`	Chande Momentum Oscillator
`StochMom`	`momentum/stoch`	Stochastic Oscillator
`StochRsiMom`	`momentum/stoch_rsi`	Stochastic RSI
`WillrMom`	`momentum/willr`	Williams %R
`CciMom`	`momentum/cci`	Commodity Channel Index
`UoMom`	`momentum/uo`	Ultimate Oscillator
`AoMom`	`momentum/ao`	Awesome Oscillator
`MacdMom`	`momentum/macd`	MACD
`PpoMom`	`momentum/ppo`	Percentage Price Oscillator
`TsiMom`	`momentum/tsi`	True Strength Index
`TrixMom`	`momentum/trix`	Triple EMA Rate of Change
`AccelerationMom`	`momentum/acceleration`	Price Acceleration
`JerkMom`	`momentum/jerk`	Rate of Acceleration Change
`AngularMomentumMom`	`momentum/angular_momentum`	Rotational Momentum
`TorqueMom`	`momentum/torque`	Rate of Angular Momentum

Overlap (Smoothers & Price Transforms)¶

Class	Component Name	Description
`SmaSmooth`	`overlap/sma`	Simple Moving Average
`EmaSmooth`	`overlap/ema`	Exponential Moving Average
`WmaSmooth`	`overlap/wma`	Weighted Moving Average
`RmaSmooth`	`overlap/rma`	Running Moving Average
`DemaSmooth`	`overlap/dema`	Double EMA
`TemaSmooth`	`overlap/tema`	Triple EMA
`HmaSmooth`	`overlap/hma`	Hull Moving Average
`KamaSmooth`	`overlap/kama`	Kaufman Adaptive MA
`AlmaSmooth`	`overlap/alma`	Arnaud Legoux MA
`JmaSmooth`	`overlap/jma`	Jurik MA
`T3Smooth`	`overlap/t3`	Tillson T3
`ZlmaSmooth`	`overlap/zlma`	Zero-Lag MA
`McGinleySmooth`	`overlap/mcginley`	McGinley Dynamic
`FramaSmooth`	`overlap/frama`	Fractal Adaptive MA

Volatility¶

Class	Component Name	Description
`AtrVol`	`volatility/atr`	Average True Range
`NatrVol`	`volatility/natr`	Normalized ATR
`BollingerVol`	`volatility/bollinger`	Bollinger Bands
`KeltnerVol`	`volatility/keltner`	Keltner Channels
`DonchianVol`	`volatility/donchian`	Donchian Channels
`MassIndexVol`	`volatility/mass_index`	Mass Index
`UlcerIndexVol`	`volatility/ulcer_index`	Ulcer Index
`KineticEnergyVol`	`volatility/kinetic_energy`	Market Kinetic Energy
`PotentialEnergyVol`	`volatility/potential_energy`	Market Potential Energy
`TemperatureVol`	`volatility/temperature`	Market Temperature

Trend¶

Class	Component Name	Description
`AdxTrend`	`trend/adx`	Average Directional Index
`AroonTrend`	`trend/aroon`	Aroon Indicator
`SupertrendTrend`	`trend/supertrend`	Supertrend
`PsarTrend`	`trend/psar`	Parabolic SAR
`IchimokuTrend`	`trend/ichimoku`	Ichimoku Cloud
`VortexTrend`	`trend/vortex`	Vortex Indicator

Statistics¶

Submodule	Count	Indicators
Dispersion	9	Variance, Std, MAD, Z-Score, CV, Range, IQR
Distribution	11	Median, Quantile, PctRank, Skew, Kurtosis, Entropy
Memory	12	Hurst, Autocorrelation, Variance Ratio, Diffusion Coefficient
Cycle	9	Instantaneous Amplitude, Phase, Frequency, Spectral Centroid
Complexity	5	Permutation Entropy, Sample Entropy, Lempel-Ziv, DFA
Information	5	KL Divergence, JS Divergence, Renyi Entropy, Mutual Info
Regression	6	Correlation, Beta, R-Squared, Linear Regression Slope
Realized Vol	5	Realized, Parkinson, Garman-Klass, Rogers-Satchell, Yang-Zhang
DSP	10	Spectral Flux, Zero Crossing Rate, Spectral Rolloff, MFCC
Control	5	Kalman Innovation, AR Coefficient, Lyapunov Exponent