How to Manually Backtest a Futures Strategy Without Fooling Yourself

The biggest risk in manual backtesting is that you already know what happened. Every chart you scroll through has a visible future, and your brain uses it. Done carefully, manual testing teaches you what a setup looks like in different conditions. Done carelessly, it produces a flattering spreadsheet that means nothing.

When manual testing makes sense

Automated testing (NinjaTrader Strategy Analyzer, scripted tests) suits rules you can encode exactly. Manual testing suits setups that rely on judgment: pattern shapes, several conditions that are hard to code together, or context that changes how good the same setup is. The cost is time and a much higher risk of bias, so rules and logging discipline matter more.

The process

1. Write the rules before you open a chart. Vague: enter when price looks oversold near support. Testable: enter long when price trades into a 5-minute bullish fair value gap formed on a wide-range candle, inside 9:30 to 11:00 AM ET, after a 1-minute bullish candle closes; stop below the gap low; target the last swing high or 10 points, whichever is closer. If you cannot write it without gaps, you will apply different standards to different days.
2. Fix the test period in advance. Pick the dates, then split them. Use one block to develop and refine the rules and set a second block aside untouched for a final check. Choose periods that include trending days, rotation days, scheduled-news days and quiet days. Be aware that the contract matters: test ES and NQ on the front-month contract, and if you use a continuous series, check how it handles rollovers, because a bad merge produces gaps and fake setups. Roll dates and session times are not universal, so consult CME's schedule.
3. Replay without looking ahead. In NinjaTrader 8, Market Replay plays historical data forward bar by bar once you have downloaded the historical data (the NT8 help guide covers the Playback data steps). Connecting to the Playback connection and choosing a SIM account are two separate steps, so check both. The official documentation at ninjatrader.com/support/helpguides/nt8 is the authority on the exact menus. If you scroll charts instead, advance one bar at a time and cover the right edge.
4. Log every candidate setup. Record date and time, direction, entry, stop, target, outcome, and your grade for the setup (define A, B, C beforehand). Also log setups that met the rules but you would have skipped, and write down why. If the reasons are not in your written rules, you have found a hidden filter, and either the rules need updating or the filter is hindsight.
5. Include costs. Apply commissions and fees from your own statement, and assume slippage on stops. A strategy tested without costs is a different strategy from the one you will trade.
6. Compute the numbers. Win rate, average winner and loser, profit factor (gross profit divided by gross loss), longest losing streak, and performance by time of day, setup grade and market condition. The Trade Performance window in NinjaTrader is a good way to review SIM trades once you have them.

Reading small samples honestly

Sixty trades looks like a lot after a month of scrolling and is small statistically. A rough guide for the uncertainty in an observed win rate near 50 percent, assuming independent trades and the normal approximation:

So an observed 55 percent on 60 trades is consistent with a true rate below 50. I would not set fixed pass thresholds such as a particular win rate or profit factor, because any threshold depends on your reward-to-risk and costs. Use the edge calculator to see what your numbers imply, and treat a result as a reason to keep testing, not as proof of an edge.

Common errors

Hindsight. Logging only the setups you remember working. The fix is to log everything the rules allow, in order, before you look at the outcome.

Rule drift. Changing the rules mid-test after a few ugly losses. If the rules change, restart the count, and ideally test the changed rules on the held-back block.

Cherry-picking the period. Testing only the months where you know the idea worked. Include a stretch where you expect it to struggle, and report that part too.

Testing the fill, not the trade. Replay assumes you get your price. In live markets your limit order has a queue position and a stop can slip.

Backtest, then forward test

A manual backtest tells you how the written rules behaved on past data. Forward testing on a SIM account tells you whether you can execute them live. It will expose entries you hesitate on and exits you move. It will not reproduce live fills, queue position or the feeling of real money at risk, and a SIM streak is not live proof. Use both, and treat the first live contracts as the next test, not as confirmation.