Backtesting software development
You have a strategy idea. We build the backtesting engine that tells you the truth about it, with realistic costs, no look-ahead, and proper out-of-sample testing. The goal is not a pretty equity curve. It is a result you can actually trust.
What we build
A backtesting engine is only useful if it is harder on your strategy than the live market will be. We build to your specification, with realism built in:
- Honest cost modelling. Transaction costs, slippage, spreads and partial fills, so the result reflects what would have happened, not an idealised fill.
- No look-ahead. The engine only ever uses information a trade could have had at the time, in both the data and the indicators.
- Clean data. Corporate actions, contract rollover and survivorship handled, because dirty data is one of the quiet ways a backtest lies.
- Portfolio level. Many symbols at once, with sizing, exposure and risk measured across the whole book, not one instrument in isolation.
- The metrics that matter. Drawdown, exposure, cost-adjusted results and the risk measures your strategy actually needs, reported clearly and reproducibly.
Why an honest backtest is hard to build
The gap between a good-looking backtest and live trading is where projects go wrong. Closing it takes real engineering:
- Look-ahead hides everywhere. A single future-peeking line can turn a losing idea into a winning chart. Finding and removing every one is careful work.
- Overfitting is seductive. Any strategy can be tuned to fit the past perfectly. Proving it works on data it never saw is what separates a real effect from a lucky one.
- Speed matters. Testing many parameters across many symbols and years has to run fast enough to actually use, without cutting the corners that make it honest.
This is the same discipline behind our research work. See how it plays out in our case study on backtesting a linear-regression support strategy, and in our research and validation service.
You bring the strategy. We build the engine.
We are engineers, not advisers. We do not supply strategies to backtest, we do not give trading calls, and we make no claim about what any strategy will earn. We build the engine and report what the data honestly shows, including the uncomfortable findings. The strategy and every decision stay with you. Most of this work sits inside our custom trading software service.
Common questions
- What makes a real backtesting engine different from a simple loop?
- A loop over history that assumes perfect fills will always look good and will always be wrong. A real engine models costs, slippage and partial fills, refuses to use information a trade could not have known at the time, handles corporate actions and rollovers, and tests out-of-sample. That gap is most of the work, and it is the whole point.
- How do you stop a backtest from fooling us?
- By building the checks in: no look-ahead in the data or the indicators, realistic transaction costs, and validation that only trusts a result if it holds up on data the tuning never touched. We would rather show you an honest, less flattering result than a beautiful one that falls apart live.
- Do you supply the strategy to backtest?
- No. You bring the strategy or the hypothesis. We build the engine that tests it, and we report what the data actually shows. We do not give trading calls and we make no claim about returns.