Deterministic Backtesting Engines in C++

Your backtest says Sharpe 4. The market says otherwise.

Most strategies don’t fail because the idea was wrong. They fail because the backtest was lying, and nobody built an engine capable of catching it. A fill that could never have happened. A price the strategy saw before it existed. A latency the simulator pretended away. A sweep of four thousand configurations that found exactly what noise finds.

Deterministic Backtesting Engines in C++ shows you how to build the engine that catches all of it. This is not a book of strategies. It is a book about the machinery that decides whether a strategy’s numbers can be believed. You will build a production-grade, event-driven backtester in modern C++ that treats every result as a claim and tries to break it, then shows you exactly how much of that result rests on assumptions.

Key Features

• Build a fully deterministic engine: same inputs, same outputs, bit for bit, across runs, machines, and thread counts
• Make lookahead bias impossible by construction, with a point-in-time data API the compiler enforces
• Reconstruct full order books from ITCH-style feeds and validate them against exchange snapshots
• Model queue position, partial fills, cancel races, hidden liquidity, and icebergs with every assumption stated
• Simulate latency as four separate delays, built from production measurements, including the tails that hit at the open
• Run one strategy codebase in both backtest and live trading, proven identical by parity testing
• Scale to terabytes of tick data with memory-mapped pipelines and deterministic parallel sweeps
• Every chapter ends with a Fidelity and Throughput Ledger: what it costs, what it is checked against, and how determinism is proven

What You Will Learn

• Why backtests lie, and the fingerprint each leak leaves in the trade log
• Virtual clocks, allocation-free event dispatch, and deterministic tie-breaking for simultaneous events
• Columnar, memory-mapped tick storage, survivorship-free universes, and point-in-time corporate actions
• Price-time matching against recorded liquidity, with honest depletion and level walking
• Fill probability, slippage, temporary and permanent market impact, fees, rebates, and venue rules
• Latency sensitivity analysis: how much of your edge survives each added microsecond
• CRTP and zero-virtual-call strategy harnesses, state isolation, and multi-strategy simulation
• Integer-exact accounting, mark-to-market rules, margin, and reconciliation against exchange fill reports
• Walk-forward testing, the deflated Sharpe ratio, block bootstrap resampling, and capacity estimation
• Reproducibility manifests, regression suites, sim-to-live parity, and CI for a backtesting platform

Who This Book Is For

Quantitative developers, quant researchers, and trading infrastructure engineers who build or depend on backtesting systems for intraday strategies: market making, short-horizon signals, and execution. You should be comfortable with modern C++ (templates, concepts, the memory model) and know what an order book is. Everything else is built from the ground up in C++23, with complete listings that compile as shown.

What You Get

• 12 chapters across 5 parts, from foundations to production deployment
• 4 appendices: data formats, model parameters, statistical formulas, and a 60-point backtest validation checklist
• A complete reference architecture with CMake build, project layout, and tests

Part of Weston Ashgrove’s Field Guide to Production Trading Systems, alongside Deterministic Matching Engines, Deterministic Risk Engines, and Deterministic Pricing Engines.

A backtest is a claim. This book gives you the engine to test it.