Blog

Engineering notes on market simulation, trading mechanics, and building better practice tools.

GPU-Accelerated Regime-Switching Market Simulation

How we generate statistically realistic OHLCV data at 500K+ candles per second using a seven-engine pipeline, a Markov regime model, geometric random walks with mean-reverting tethers, and CuPy vectorization.

Agent-Based Order Flow Simulation

How we simulate realistic tick-by-tick price action using four agent types, a self-exciting Hawkes process, and a limit order book with depth.

The Problem with Backtesting on Historical Data

Why replaying recorded market data creates blind spots in your development as a trader, and what a better practice model looks like.

Simulating Realistic Market Regimes

How we generate price sessions that feel like real markets instead of random noise.