Confidence Score
The Confidence Score (0–100) measures whether there is enough consistent evidence to trust your strategy's results. A strategy might look profitable, but with too little evidence behind it, you can't yet tell a real edge from a lucky run.
Warning
Minimum trade requirement: AlgoChef needs a minimum number of trades before it scores a strategy. Below that, the Confidence Score — and the Profitability, Risk, and CSI scores — aren't calculated at all; they show Not enough data (in gray). This is the most common reason a strategy has no CSI yet.
How It Works
The Confidence Score is a proprietary score that answers one plain question: is there enough consistent evidence? A handful of trades proves nothing, no matter how good they look — and a strategy whose results only showed up during one lucky stretch is less trustworthy than one whose behaviour is steady across its whole history.
The exact measures, weights, and thresholds inside the score are proprietary — like a credit score, you see the result and what drives it directionally, not the internal formula.
Info
Want to check for luck directly? The Confidence Score doesn't run a chance test. For that, use Overfit Likelihood, which checks whether the edge can be told apart from noise, and Monte Carlo simulation, which shows how much your results could vary.
What Makes a Good Confidence Score
| Score Range | What It Means |
|---|---|
| 80–100 | Highly reliable — strong, consistent evidence behind the results |
| 60–79 | Good confidence — results are likely meaningful with minor uncertainty |
| 40–59 | Moderate confidence — more evidence or more consistency needed |
| 1–39 | Low confidence — too little consistent evidence to trust the results yet |
How to Improve
- Too little history? This is the #1 cause. Extend your backtest period, or test on more markets. There is no shortcut — statistical confidence requires evidence, and evidence means trades.
- Erratic results? If your strategy's edge appears and disappears across its history, confidence drops. Look at Rolling Returns in the Quality Report to see whether the edge is steady or concentrated in one period.
- A few huge winners carrying everything? That's worth knowing, though it's a separate diagnostic rather than an input to this score. Check outlier dependency in the Advanced metrics.
- Want to quantify the uncertainty? Run a Monte Carlo simulation — it directly tests how much your results could vary with the sample you have.
For context on why thin evidence is one of the most common reasons strategies fail live, see Why Strategies Fail — Insufficient Sample Size.
Tip
A strategy with no CSI yet usually just needs more trades — and that's easy to misread. It doesn't mean your strategy is bad; it means there isn't enough evidence yet to know. More trades fix it.
Tip
Ready to check your strategy's statistical confidence? Start your 14-day free trial.