Instruments,
not signals.

Most trading systems don't fail because of a bad idea — they fail for lack of validation. A backtest on coarse data with optimistic fills almost always produces a nice, smooth curve, and that is exactly the one that falls apart in a real account. We invert this: before we consider a strategy good, we try everything to break it. What survives is allowed to go live.

It starts with the data

Not with the strategy. We work with real tick / M1 data straight from the broker and check where spread and quotes come from — because a fill that never realistically existed makes every downstream statistic worthless. The logic on top stays deliberately lean: clear rules for entry, exit, stop and position size, instead of a stack of optimised indicators that has merely memorised the past.

The real test: walk-forward

Parameters are optimised on one window and tested on the next, unseen one — rolling across years. A strategy that only shines in the optimisation window fails here. On top come Monte-Carlo simulations over trade order, parameter sensitivity and a look across market regimes: a setup has to hold through crashes, inflation bear markets and grinding sideways phases — not just the one good year.

Conservative into live

Only once that holds does capital come in — with fixed risk per trade in R, a defined maximum drawdown, and no martingale or grid that keeps a losing system artificially alive. In operation the systems run on a VPS, with clean magic-number separation, session force-close and a continuous reconciliation of live results against the backtest. If the two diverge, we investigate — we don't keep trading.

And we say where the limits are

Every one of our research studies states its limitations openly: sample size, single-source data, what wasn't modelled. We promise no returns. We show what the data supports — and what it doesn't. That sells worse than '+300% in backtest', but it is the difference between a system you understand and a black box you can only hope in.