Crate algotrading 

Algotrading Library

High-performance quantitative trading primitives optimized for AWS EC2.

Design Philosophy

• f64 only: Financial precision is non-negotiable
• Stack allocated: Const generics, zero heap allocations
• SIMD optimized: AVX2 baseline (all modern EC2 instances)
• Cache-friendly: 64-byte alignment, structure-of-arrays layout

Modules

• stats: Rolling statistics, PDF estimation
• probability: Mahalanobis distance, kernel density, empirical CDF
• filters: Kalman filters for state estimation
• regime: Markov switching, HMM
• data: Data quality checks and validation
• matrix: Linear algebra operations (Cholesky, Eigen, SVD, etc.)
• options: Black-Scholes pricing, Greeks, implied volatility
• risk: VaR, Sharpe ratio, drawdown, Kelly criterion
• ta: Technical indicators (MACD, RSI, Bollinger Bands, etc.)

Performance Targets (c6i.xlarge)

• Rolling stats: ~2.5ns per update
• KDE density eval: ~50ns
• Mahalanobis distance: ~20ns
• Markov update: ~15ns
• Batch (4-wide SIMD): ~8ns per series

Example

use algotrading::prelude::*;

// Stack-allocated, 300-sample window
let mut stats = RollingStats::<f64, 300>::new();
let (mean, std) = stats.update(0.001);

// Regime detection
let mut regime = MarkovSwitching::spy_default();
regime.update(mean);
 
if regime.is_volatile(0.8) {
    println!("Market is volatile!");
}

Modules

core

Core data structures and primitives

data

filters

matrix

numeric

options

prelude

Common imports

probability

regime

risk

stats

Statistical computations with optional SIMD acceleration

ta

Technical indicators with optional SIMD acceleration

Macros

impl_default_numeric

Macro to implement Default for generic numeric structs

init_array

Create a new generic instance with proper initialization

numeric_aliases

Macro to create type aliases for scalar and SIMD versions

numeric_struct

Macro to create a generic version of an existing struct with numeric fields