Backtesting
Test your strategies against historical data to evaluate performance.
Backtesting
Backtesting lets you test a strategy against historical market data to see how it would have performed in the past. It's the most important step before deploying any strategy.
Running a Backtest
- Navigate to your portfolio dashboard.
- Click the "Backtest" button.
- Select a date range (e.g., January 2020 to today).
- Click "Run Backtest" and wait for results.
You can also ask Aurora: "Backtest my portfolio from 2020 to today."
Understanding Backtest Results
Every backtest report includes these key metrics:
| Metric |
What It Means |
Good Value |
| Total Return (%) |
How much the portfolio gained or lost |
Higher is better |
| Sharpe Ratio |
Return per unit of risk (volatility) |
> 1.0 is good, > 2.0 is excellent |
| Sortino Ratio |
Like Sharpe but only penalizes downside volatility |
> 1.5 is good |
| Max Drawdown (%) |
Largest peak-to-trough decline |
Lower is better (< 20% is conservative) |
| Benchmark Comparison |
Your strategy vs. SPY (S&P 500) |
Outperforming SPY is the goal |
OHLC vs. Intraday Backtesting
NexusTrade supports two backtesting modes:
- OHLC (Daily): Uses daily open/high/low/close data. Faster and available on all plans. Good for strategies that trade on daily signals.
- Intraday: Uses minute-by-minute data. More accurate for strategies that need precise entry timing. Available on paid plans.
Token Cost
Backtests draw from your daily research token budget — the same pool used by Aurora chat and optimization runs.
Pricing by backtest type:
| Type |
Cost |
| Daily (non-options) |
Free |
| Daily (options) |
0.01 tokens per calendar day |
| Intraday (non-options) |
0.125 tokens per calendar day |
| Intraday (options) |
0.25 tokens per calendar day |
Examples:
- A 1-year daily non-options backtest: free.
- A 1-year daily options backtest: 365 × 0.01 = ~4 tokens.
- A 1-year intraday non-options backtest: 365 × 0.125 = ~46 tokens.
- A 1-year intraday options backtest: 365 × 0.25 = ~91 tokens.
Intraday quicktests (under 90 days) are free for all paid users, regardless of data type.
Event generation (enabling the "generate events" flag on an intraday backtest) multiplies the cost by 5×, since the engine produces a full trade event log.
Daily research token grants: Starter 1,000/day, Financial Genius 3,000/day, Wall Street Elite 6,000/day.
Important Caveats
Past performance does not guarantee future results. A strategy that backtests well may perform differently in live markets.
Overfitting
The biggest risk in backtesting is overfitting — creating a strategy that is so perfectly tuned to historical data that it fails on new data. Signs of overfitting:
- Extremely high returns that seem too good to be true
- The strategy only works on a very specific date range
- Adding more conditions keeps improving the backtest but the strategy is overly complex
How to Avoid Overfitting
- Use simple strategies: Strategies with fewer conditions tend to generalize better.
- Test on multiple date ranges: If a strategy only works from 2020-2023 but fails on 2018-2020, it may be overfit.
- Use train/validation/test splits: Optimize on one period, validate on another, and do a final test on a third.
- Paper trade before going live: Real-time paper trading is the ultimate validation.
Related
After you run a backtest, pin the equity curve in a portfolio notepad so the thesis and the chart stay together. You can overlay up to four runs on one graph.