How to run a realistic backtest in NinjaTrader 8, step by step

Most backtests I see don't fail because of the strategy: they fail because of how they were measured. I wrote about the most extreme case here: a strategy showing a "90% win rate" on the chart that was really 40% measured tick by tick. This guide is the setup I use so that the number NinjaTrader gives you resembles the number your account will give you.

Step 1 — Use the Strategy Analyzer, not the chart

Loading a strategy on a chart and looking at the drawn trades is not a backtest: it's an approximation over closed bars. Backtesting happens in the Strategy Analyzer (New → Strategy Analyzer), where you control data, fills, commissions and period.

Step 2 — High order-fill resolution with 1-tick data

This is the setting that separates a serious backtest from a drawing: in the analysis parameters, set "Order fill resolution → High" and select 1 tick as the secondary series. That way every stop and target is evaluated against the real path of price inside each bar, in its real order. Without it, stops that "never trigger" will inflate your win rate.

If your intrabar logic is complex (entries that depend on tick sequence), also consider Tick Replay so indicators compute tick by tick as they do live.

Step 3 — Quality historical data

A tick-by-tick backtest needs real, deep tick data (I use Kinetick). Check how many days of tick data you actually have downloaded: if you only have 20 days of ticks, your "2-year backtest" is only tick-accurate for those 20 days.

Step 4 — Commissions and slippage

Configure your broker's commission template and add slippage in the parameters. On futures like MNQ, 1 tick of slippage per side is a reasonable minimum during liquid hours; more if you trade news or thin liquidity. A strategy that dies when you add commissions and 1 tick of slippage was never a strategy: it was a statistical illusion.

Step 5 — Validate out of sample

If you optimised parameters on 2023-2024, the 2023-2024 result proves nothing: the system "already knew the exam". Reserve a period the optimisation never saw (say 2025-2026) and evaluate there. If the curve only works on the optimised period, that's curve-fitting, not an edge. More on this in how to evaluate an automated strategy.

Step 6 — Simulation first, then small live

An approved backtest doesn't go straight to live: first weeks on simulation or Market Replay, checking that live entries match the backtest logic (if they don't, here are the usual causes). Then a live account with micros and a weekly backtest-vs-live comparison. Between tick-by-tick and live, still expect roughly a ±10% gap — if it's much more, something is mismeasured or misexecuted.

None of this guarantees the strategy wins — nobody honest can guarantee that. What it guarantees is something more valuable: that the number you are looking at means something. From there you can decide with real data, not drawings.