The Algomyth engine backtest prints $87,359 net on a mid-size account. The reason that number is worth anything is that you can produce it yourself: same platform, same instrument, same period, same costs, same report. Here is how, click by click, in the NinjaTrader 8 Strategy Analyzer.
A backtest is a simulation of a strategy over historical data, explained on its own terms in what is a backtest. Done carefully it is the most honest thing in trading, because it is reproducible. Done carelessly it is a machine for lying to yourself. The gap between the two lives entirely in the settings, so we will go slowly.
A backtest is only as honest as its settings. Defaults flatter results: no commissions, optimistic fills, too little data. Change five settings and the same strategy can look twice as good or half as good. Fix them before you trust a single number.
The Strategy Analyzer is the tool inside NinjaTrader 8 that runs a strategy against historical bars and prints a performance report: net profit, win rate, profit factor, maximum drawdown, the full list. It is separate from a live chart. You point it at an instrument and a date range, it replays history bar by bar, and it tells you what the strategy would have done. If the platform itself is new to you, start with NinjaTrader 8 explained and NinjaTrader 8 for beginners, then come back here.
Before the first run, a short checklist:
Five settings move the number more than anything else. Get these wrong and the report is fiction.
| Setting | What it changes | What to use |
|---|---|---|
| Order fill resolution | How precisely fills are simulated inside each bar | High, or use tick replay, so fills are not assumed at impossible prices |
| Tick replay | Replays every tick inside the bar instead of just the open, high, low and close | On for a realistic test; it is slower and needs tick data downloaded |
| Commissions | Broker and exchange fees per contract | Your real rate, about $1.90 per contract round turn, never zero |
| Slippage | The gap between the price you wanted and the price you got | A conservative allowance, never zero for market orders |
| Data amount | How many days or years are actually tested | As much as you have; a handful of days proves nothing |
This is where a published figure is either honest or decorative. The $87,359 on the performance page is what the report prints once these fields are filled in: $1.90 per contract round turn in commission and one tick of slippage, which is $0.50 on MNQ and $1.25 on MES. Leave them at zero and you will get a bigger number that means nothing. The full breakdown of what comes off the top is in slippage and commissions.
Hypothetical performance. The $87,359 net figure is the result of a backtest of the Algomyth engine over historical data, not a live account. Simulated results are prepared with hindsight, carry no financial risk, and cannot fully reflect real execution, slippage or liquidity. Past performance, real or simulated, does not guarantee future results.
The full report is published with its period and settings stated in the open, and the live signal log lets you compare the present against it, line by line.
Three failures turn a good-looking report into a worthless one.
Overfitting. If you keep adjusting parameters until the report looks great on this exact history, you have not found an edge, you have memorized the past. The result will not survive contact with new data. This is the single most common way people fool themselves, and it has its own article: overfitting and curve fitting.
Too little data. A strategy that looks brilliant over three weeks has told you almost nothing. Markets go through quiet stretches and violent ones, and a real test has to include the bad conditions, not just a lucky window. That is why the published test runs from January 2022 to September 2026, including plenty of periods the strategy did not enjoy.
Look-ahead bias. If the strategy accidentally uses information it could not have known at the time, for example the close of a bar that has not finished yet, the backtest shows profits that are impossible to capture live. Careful bar handling and tick replay guard against it. Any result that looks too clean deserves suspicion, not celebration.
The honest way to test is to design the rules on one slice of history and then check them on a different slice the rules never saw. If it holds up on data it was not tuned on, you have something. If it only works on the slice you built it on, you have an overfit wearing a costume. A single all-history run that looks perfect is a flag, not a trophy. This is exactly why reproducibility matters more than size: a number you can regenerate on your own machine, over data you choose, is far harder to fake than a screenshot on a sales page.
The Strategy Analyzer is part of NinjaTrader 8, and you can run backtests on historical data without funding a live account. What you need is the platform installed on Windows and enough historical data downloaded for the period you want to test. Live trading and real-time data are separate arrangements, but backtesting the strategy on past data does not require them.
A quick test over a few days finishes in seconds. A full multi-year run with tick replay switched on can take several minutes or more, because the engine is replaying every tick across nearly four years of data. This is where the 8 GB of RAM matters, and where a faster machine is worth having. If it is unbearably slow, you are usually short on memory.
Small differences almost always trace back to settings: a different commission rate, a different slippage assumption, a shorter data range, or a lower fill resolution. That is a feature, not a bug, because it lets you substitute your own broker's costs and see the result on your terms. If the number is wildly different, check that you loaded the full period and the correct MNQ contract on a 1-minute series.
For a realistic test of an intraday strategy, yes, because it simulates fills using every tick inside the bar rather than assuming a price the market may never have offered. It is slower and it requires tick data to be downloaded, but it is the difference between a fill you could actually get and a fill the report invented. For a rough first look you can leave it off, then confirm with it on before you trust anything.
No, and any source that says otherwise is selling something. A backtest is a hypothetical simulation prepared with hindsight, and it cannot fully reproduce live slippage, liquidity, outages or your own behavior under a drawdown. Past performance, real or simulated, does not guarantee future results. A clean, honest backtest raises your confidence that the rules have an edge; it does not promise the edge repeats.
Everything we write comes back to one idea: verify, then decide. The live terminal exists for exactly that.
Unlock Algomyth Live