How to evaluate an automated strategy for NinjaTrader 8 without falling for inflated results
You've been looking at automated strategies, some look too good to be true, and you're not sure how to tell a serious strategy from one with a polished backtest. This guide is for exactly that.
I'm not going to give you a magic formula. I'm going to give you the criteria I use myself after 8 years studying markets and over a year developing and trading algorithms on NinjaTrader 8.
The first thing to understand: a good backtest is a necessary condition, but not sufficient. You can have a perfect backtest with a strategy that doesn't work in real markets. And you can have a modest backtest with a solid strategy. The backtest is the starting point, not the conclusion.
The metrics that actually matter
| Metric | What it measures | Reference |
|---|---|---|
| Profit Factor | Gross profit divided by gross loss | Good: +1.5 / Very good: +2 |
| Max Drawdown | Largest drop from a peak to the next trough | Over 30% of capital is hard to tolerate |
| Number of trades | Statistical sample size | Minimum 100-200 trades for significance |
| Win rate | Percentage of winning trades | Don't evaluate alone — combine with win/loss ratio |
| Period analysed | How long the backtest covers | Minimum 1 year. Better if it covers different market regimes |
| Instrument traded | Which contract and size was used | MNQ ≠ NQ. NQ results can look inflated vs real capital |
What the backtest doesn't tell you
Optimisation bias (overfitting)
If someone tests 500 parameter combinations and shows you the best-performing one, that result isn't representative. The strategy is fitted to the past, not tested on it. The more parameters a strategy has, the easier it is to overfit.
Slippage and commissions
Many backtests don't include the real cost of commissions or slippage (the difference between the price you see and the price you actually fill at). In high-frequency strategies this can completely change the result.
Look-ahead bias
Some algorithms unknowingly use future information in their calculations. In backtest it works perfectly. In real time, that data doesn't exist yet. The results collapse.
Markets change
A backtest from 2019-2021 may show brilliant results due to the volatility of that period. That guarantees nothing about 2025 or 2026. Always look for results across different market periods.
Immediate red flag: if you're shown a 1-2 week backtest on the NQ (full contract) instead of MNQ (micro), with spectacular results — be suspicious. It's the easiest way to inflate numbers without technically lying. A 1% move in NQ is $2,000. In MNQ it's $200. Results are ten times smaller in real terms.
The difference between backtest and forward testing
The backtest applies the strategy to historical data that already existed when it was programmed. The result always has some bias because the developer, even unconsciously, has seen that data.
The forward test applies the strategy to new data — market that occurred after the strategy was finalised. It's much more reliable precisely because that data didn't exist when the strategy was designed.
Drawdown: the number people ignore that matters most
Drawdown is the largest drop you would have had to endure at some point. And "endure" is exactly the right word, because drawdown isn't just a number — it's an emotional experience.
A strategy with a -$10,000 drawdown in backtest means at some point you would have watched your account fall $10,000 from its peak. The real question isn't whether the backtest survives it — it's whether you survive it without touching anything.
Practical rule: the maximum drawdown should represent a fraction of your capital that you can sleep peacefully with. If the maximum drawdown is more than you're willing to lose without panicking, that strategy isn't for you — regardless of the results.
Questions you should ask before buying any strategy
- How many trades does the backtest have? Fewer than 100 is not statistically significant.
- What period does it cover? Minimum one year, covering different market types.
- Does it include commissions and slippage? If not, real results will be worse.
- What instrument? MNQ or NQ? The difference is enormous in terms of capital required.
- Is there forward testing or only backtest?
- Can you see their YouTube channel or some continuous public proof? Anyone can manufacture a backtest. A history of live streams is much harder to fake.
- What happens if the strategy stops working? Is there support? Updates? Or is it buy-and-forget?
Frequently asked questions
Want to see how we apply these criteria in our strategies?
At Inver Mind we publish complete backtest data — period, instrument, profit factor, drawdown, number of trades — without hiding anything. And you can watch them live on YouTube.
View available strategies