Overview
A trading strategy that has never been tested against historical data is a hypothesis, not a system. Backtesting is the process of running a strategy against past market data to understand how it would have performed — identifying strengths, exposing weaknesses, and giving you the statistical foundation to decide whether a strategy is worth trading live.
Building a backtesting platform that produces results you can actually trust requires more engineering rigour than most traders expect. Overfitting, look-ahead bias, survivorship bias, unrealistic fill assumptions, and inadequate historical data coverage all produce backtests that look compelling on a chart and fail immediately in live markets. We build backtesting infrastructure that is designed to give you accurate results — not flattering ones.
The Gap Between a Backtest and Reality
Most off-the-shelf backtesting tools make implicit assumptions that quietly invalidate their results. Orders fill instantly at the signal price. Spreads are ignored or fixed. Slippage does not exist. The historical data is clean and continuous. Position sizing is simplified. These assumptions produce equity curves that bear no relationship to what a live execution of the same strategy would actually generate.
A properly engineered backtesting platform closes this gap by modelling the market as it actually behaves:
Realistic order execution modelling. Market orders do not fill at the last price. We model fill prices based on bid-ask spread at the time of execution, available liquidity at each price level, and configurable slippage assumptions based on order size relative to typical volume.
Tick-level and OHLCV granularity. Bar-based backtesting on OHLCV data introduces ordering ambiguity — within a single bar, did the high or the low occur first? For strategies sensitive to intra-bar order, we support tick-level data processing where the historical data source allows it.
Commission and funding cost modelling. Every trade has a cost. Maker/taker fee structures, overnight funding rates on leveraged positions, and withdrawal fees on realised profits all affect net performance. We model them accurately so your net returns reflect what you would actually keep.
Latency simulation. For strategies where execution latency matters — particularly high-frequency or arbitrage approaches — we model signal-to-execution delay explicitly rather than assuming instantaneous order placement.
Position and risk constraint enforcement. Real trading accounts have margin requirements, position limits, and risk controls. A backtest that ignores these will show trades that could never have been executed in practice. We enforce the same constraints in backtesting that apply in live execution.
What We Build
Custom backtesting engines Where off-the-shelf platforms like MetaTrader's Strategy Tester, Backtrader, or vectorbt do not fit your requirements — because your strategy runs across multiple assets simultaneously, requires custom data sources, operates on non-standard timeframes, or needs to model specific execution conditions — we build a purpose-built backtesting engine tailored to your strategy architecture.
MetaTrader Strategy Tester integration (MQL4 / MQL5) For strategies deployed on MetaTrader, we build and optimise directly within the native Strategy Tester environment in MQL4 or MQL5. This includes multi-currency and multi-timeframe testing, optimisation across parameter ranges using the built-in genetic algorithm optimiser, and forward testing validation to distinguish genuine edge from overfitted curve fitting.
Python-based backtesting frameworks For quantitative strategies requiring statistical analysis, machine learning signal generation, or complex portfolio-level logic, we build backtesting systems in Python using established frameworks as a foundation — extended with custom execution modelling, data pipeline integration, and performance analytics tailored to your specific requirements.
Historical data pipelines A backtesting platform is only as good as its data. We build historical data pipelines that source, clean, and store tick data, OHLCV data, and alternative data feeds from exchanges, data vendors, and onchain sources. Data quality validation — gap detection, outlier removal, corporate action adjustment — is built into the pipeline rather than left to manual inspection.
Performance analytics and reporting Raw equity curve results are not enough. We build comprehensive performance analytics modules covering:
- Risk-adjusted return metrics — Sharpe ratio, Sortino ratio, Calmar ratio
- Drawdown analysis — maximum drawdown, average drawdown, drawdown duration
- Trade statistics — win rate, profit factor, average win/loss ratio, expectancy
- Monte Carlo simulation for robustness testing across randomised trade sequencing
- Walk-forward analysis to validate parameter stability across time periods
- Benchmark comparison against relevant market indices or passive strategies
Parameter optimisation with overfitting protection Optimisation across strategy parameters is a standard part of strategy development — but naive optimisation produces strategies that are perfectly fitted to past data and useless going forward. We implement optimisation frameworks with built-in overfitting protection including out-of-sample validation splits, walk-forward optimisation, and robustness metrics that penalise parameter sensitivity.
Live trading bridge A backtesting platform that cannot be connected to a live execution system has limited utility. We build backtesting platforms with a defined interface to the live trading layer — ensuring that the same strategy logic that runs in the backtest is the logic that executes in production, eliminating translation errors between research and deployment.
Supported Trading Environments
Our backtesting platforms cover the full range of trading environments we work with:
Forex and CFD — tick and OHLCV data, MetaTrader-compatible strategy logic, broker-specific spread and commission modelling
Crypto spot and derivatives — exchange-specific fee structures, funding rate modelling for perpetual futures, order book depth simulation for larger position sizes
Onchain DEX strategies — gas cost modelling, block time constraints, MEV exposure simulation, liquidity depth at specific price levels using historical pool state data
Equities and futures — where clients operate in traditional markets, we integrate with relevant data providers and model exchange-specific execution rules
Technologies Used
- Rust — high-performance backtesting engines requiring tick-level data processing at speed
- C# — backtesting systems integrated with MetaTrader bridges and desktop strategy tools
- MQL4 / MQL5 — native Strategy Tester development and optimisation for MetaTrader-deployed strategies
- Python — quantitative backtesting frameworks, statistical analysis, machine learning signal research
- SQL — historical data storage, trade log persistence, performance analytics queries
- Parquet / columnar storage — efficient storage and retrieval of large tick data sets
- WebSocket / REST — live data feed integration for forward testing and paper trading modes
When to Come to Us
You need a custom backtesting platform when the standard tools are not adequate for your strategy. If your strategy operates across multiple instruments simultaneously with interdependent position logic, requires custom data sources that off-the-shelf platforms do not support, needs to model specific execution conditions accurately, or if you have outgrown MetaTrader's Strategy Tester and need more control over the testing environment — we build what you need.
We also work with traders who have existing backtesting setups that produce results they do not fully trust. If your backtest results do not translate to live performance, we can audit the existing system, identify where the assumptions break down, and rebuild the components that are producing misleading results.
Build a Backtest You Can Actually Trust
The purpose of backtesting is not to produce a good-looking equity curve. It is to generate actionable information about whether a strategy has genuine edge and how it is likely to behave in live conditions. We build the infrastructure that makes that possible.