Introducing the Stock Market AI Research Tool (SMART)¶
The Stock Market AI Research Tool (SMART) is a family of AI models built into the Empirical Markets platform. Each model produces daily, actionable trading signals, and every signal comes with a transparent backtest so you know exactly what you're getting into before you trade.
Most AI tools give you a prediction with no context. SMART gives you the prediction and the historical track record attached to it: win rate, average return, profit factor, and a full return distribution — all computed on held-out test data. These metrics are calculated using well-defined trading strategies, so you don't have to guess how to use the predictions.
In this post, we'll walk through how to navigate the SMART interface, what each model predicts, and how to interpret the performance metrics.
SMART Models¶
SMART currently includes four models, each targeting a different market phenomenon:
| Model | What It Predicts |
|---|---|
| Polaris | Whether a stock's return over the next N days will be strongly negative, mildly negative, neutral, mildly positive, or strongly positive |
| Kepler | Same as Polaris but using volatility-clock time instead of calendar time |
| Proxima | Whether the risk-reward setup over the coming horizon favors a long or short position |
| Cygnus | Whether there will be a breakout to the upside or downside |
Over time, more models will be added, and others may be deprecated.
Actionable Trading Signals¶
First, after logging in to Empirical Markets, navigate to the SMART model interface and select a model.
Then, select a model variant. Model variants may have one or more trading strategies associated with them. Each of these strategies have associated historical return metrics attached to them, along with details about the backtest. Inspecting these metrics gives you an understanding of the risk-reward profile for the strategy.
Each signal is directional. Variants are labeled ↑ Long or ↓ Short, indicating which direction the model expects the edge to favor.
Understanding the Backtest Metrics¶
Every SMART strategy card shows the following metrics, all computed on hold-out test data (meaning data not seen during training):
Entry & Exit Strategy¶
Before reading any number, understand how the trade was simulated:
- Entry Strategy: Signals are entered at the next day's open after a signal fires — this is a realistic, implementable assumption. There's no look-ahead.
- Exit Strategy: The default is First Profitable Exit — if there is a profitable close within the horizon, exit there; otherwise hold to the end of the horizon.
- Horizon: The number of trading days the strategy looks forward (e.g., 10 days).
For each model, new strategies will be added over time, alongside their performance stats.
Performance Statistics¶
| Metric | What It Means |
|---|---|
| Hit Rate | Percentage of signals that resulted in a profitable trade. A hit rate above 50% means more wins than losses |
| Mean | Average return across all signals. Positive means the strategy has been profitable on average |
| Median | Median return. This is less sensitive to outliers than the mean. If the median is positive, most trades ended in the green |
| Avg Win | Average return on winning trades only |
| Avg Loss | Average return on losing trades only |
| Profit Factor | Gross profit divided by gross loss. A profit factor > 1.0 means the strategy made more than it lost in aggregate. |
Return Distribution Chart¶
Below the headline stats, a bar chart shows the strategy return distribution by percentile — the forward return at each percentile of all signals. This tells you the full shape of outcomes:
- Green bars (above zero): percentiles where signals produced gains
- Red bars (below zero): percentiles where signals produced losses
- The wider the green portion, the more consistently the strategy generated positive returns
A strategy where even the 25th percentile bar is positive is meaningfully different from one where only the top half is green.
Get Started¶
SMART is available now on Empirical Markets. Sign up for a free account to access the Markets Intelligence Dashboard, or upgrade to a paid plan for full access to all SMART models, ticker-level signals, and the full historical data API.