Two-stage backtest
Stage 1 selects parameters on one period. Stage 2 tests the chosen parameters, unchanged, on data from 2024 onward that was not used for selection. An idea that only works on the data it was tuned on does not pass.
> // QUALITY_GATES
How we test an idea
Most trading ideas look good on a chart and fail on real data. Our job is to find that out before you pay for code.
Quality process
Real costs
MetaTrader 5 tests use real tick data with spread, commission and swap. Crypto tests use exchange historical data with trading fees and funding. Results that only exist without costs are not results.
Neighbouring parameters
We test the parameters around the chosen ones. If a small change destroys the result, the idea is fitted to noise and does not pass.
Quality gates
An idea passes only if it clears all gates: results that hold in stage 2, costs included, stable across neighbouring parameters, and a drawdown within the limit you set.
14-day live acceptance
After the build, the bot runs on a live account for 14 days to confirm it follows the approved rules. Acceptance checks the behaviour, not the profit.
Why most ideas do not pass
Most ideas are fitted to the past: they react to patterns that happened once, ignore costs, or depend on one exact parameter value. Stage 2 and the neighbouring-parameter check remove most of them.
Why a "did not pass" report is worth paying for
It costs far less than building, running and losing money with a strategy that never worked. The report shows where the idea breaks, so you can drop it or change it with evidence.