How we research, build and run our systems
Research
A strategy begins with a hypothesis about how a market behaves — a pattern in how prices move around a session, a reaction that repeats, a dislocation that does not clear immediately. We then test whether the effect holds outside the data that suggested it.
Most hypotheses fail here, and that is the purpose of the step. What we look for is not a good backtest but a reason the effect should exist, and evidence that it survives when the assumptions change.
From test to live
A strategy that survives research moves to paper trading on live market data. It runs in the same environment as a live system — same signals, same execution logic, same monitoring — and we reconcile its results against the backtest to confirm that what we tested is what actually trades.
Only then does it trade real capital, at small size, on our own accounts.
Infrastructure
We built our own stack: data collection, backtesting, signal generation, execution and monitoring during market hours.
Building it ourselves is not a matter of preference. When a system behaves unexpectedly we need to know which component produced the behaviour, and that is not possible when the path between a signal and an order runs through software we cannot inspect.
When things go wrong
Systems fail. Orders do not fill, a position stays open past its exit, a data feed drops mid-session. What matters is whether you find out, and how fast.
Automated checks run against the live state of the account and flag anything that does not match what the strategy should be doing. We reconcile executed trades against expected ones on a regular basis. A strategy that behaves outside its expected range is stopped and goes back to research.
The team
We are a team of five. Research, engineering and execution sit with the same people, which keeps the distance between an idea and a live system short.