Seven steps between your data and your databank. Each one is there to reject strategies that only fit the past.
You export your own 1-minute history from NinjaTrader 8 (futures), MetaTrader 5 or cTrader (CFDs). Every bar is converted to Chicago time, so a strategy means the same thing on every computer. On futures, every simulated trade pays a commission and one tick of slippage each way; on CFDs, it pays a cost based on your broker's spread.
Sessions: US 08:30 to 15:00, EU 02:00 to 10:30, Asia 18:00 to 00:30 (Chicago time).
The engine starts from one of 17 market behaviours described in research, such as the opening range breakout or early-session momentum, and only tunes the entry window, the trigger, the stop, the target and a few optional filters.
One strategy = one line of text, replayed bar by bar by the trading algo in NinjaTrader 8, MetaTrader 5 or cTrader.
A 10-point stop is huge on one market and tiny on another, and even on the same market its meaning changes over the years. So every price distance is a multiple of recent volatility, not a fixed number of points: the same strategy keeps its meaning across markets and years.
Distances are multiples of the 1-minute ATR(14). On MNQ, that ATR went from about 3 points in 2019 to about 16 in 2026.
Before any tuning, the engine checks whether a strategy family behaves the same way on two halves of your history. If it does not, that family is left out for this market.
1,000 random versions per family, compared on alternating blocks of the training data.
The history is cut into training periods and test periods. The search only sees the training periods, and prefers strategies whose neighbours work too: a result that only holds at one exact setting is a coincidence.
Walk-forward blocks (15% training, 5% test) by default; ranking by the performance of 16 slightly changed versions.
Only strategies profitable both in training and in the periods they never saw reach your databank. Near-copies are removed, so ten strategies are ten different bets.
The Deflated Sharpe Ratio and the Probability of Backtest Overfitting are shown for every run.
Each strategy and each portfolio is replayed a thousand times with its days reshuffled, to show the drawdown to plan for, not just the one the backtest happened to show.
The risk figures quote the 95th percentile drawdown; a prop-firm challenge can be simulated on hundreds of past start dates.
The Studio and the trading algo (NinjaTrader 8, MetaTrader 5 or cTrader) run the same strategy code, bar by bar.
Confirm any result yourself in the NinjaTrader Strategy Analyzer, the MetaTrader 5 Strategy Tester or the cTrader backtest, then trade it on a simulation account before going live. Backtests describe the past: markets change, and the decision to trade, and how much, stays yours. See the risk disclosure.