Backtest methodology

An AI model reads your strategy and translates it only into rules the engine can actually run. We show you the translation before testing; if we misunderstood, you correct it.

We never silently fill in values: if we chose a stop, target or indicator period for you, we say so. If something is ambiguous we ask instead of guessing; if a condition isn't supported we don't run the test and tell you why.

Conditions are evaluated at the candle close; the trade is entered at the next candle's open. There is no entering at the close of the very candle that produced the signal.

Stops and targets are checked within each candle. If both are touched in the same candle, we assume the stop hit first (a conservative assumption).

Exit rules are also evaluated at the close and the trade is closed at the next candle's open.

If the next candle opens beyond the stop (a price gap), the trade is not taken; we never book a fake loss at a stop price that could not have been reached.

By default a trade stays open for at most 50 candles, then it is marked to market and closed. If your strategy has a time limit, that is used instead.

Each candle only uses that candle and earlier data. 250 candles are reserved before the test so indicators can warm up; they are excluded from the results.

Higher-timeframe conditions (e.g. trade on 15m, trend on 4h) only use completed higher-timeframe candles.

Structures such as swings, order blocks and FVGs become usable only once they are confirmed.

Every trade pays costs on entry and on exit separately (per side). Default values:

Each trade risks a set share of the account (1% by default); position size follows from the distance between entry and stop.

We also show results in R: 1R is the amount risked on that trade. +2R means twice the risk was won; -1R means the stop was hit in full.

Crypto data comes from OKX perpetual swaps; forex and precious-metal data from Twelve Data.

Testable history depends on the market and timeframe: one test can fetch up to about 30,000 candles for crypto and 5,000 for forex and metals. Strategies that only use the built-in ICT templates have a shorter limit (about 1,500 candles). The free plan tests roughly the last 6 months; the Trader plan tests deeper history within those limits. 1-minute candles exist for crypto only.

If the full range you chose could not be fetched, the result page says so.

Win rate, total R, profit factor and maximum drawdown are computed from all closed trades.

Sharpe and Sortino ratios are computed per trade and scaled by trades per year, without a risk-free rate. With few trades they are very noisy.

A small number of trades (e.g. under 30) proves nothing on its own; we show a warning then. With at least 10 closed trades we also offer a robustness check that resamples your trades with replacement thousands of times (bootstrap) to show how sensitive the result is to luck.

Trying many variations on the same data and keeping the best one (overfitting) produces strategies that look great in the past and fail in the future. Test a strategy on other symbols, timeframes and periods too, then follow it on live prices with virtual money.