What the Five Scores Mean
Every strategy in AlgoChef gets five scores, each from 0 to 100. Before you import your own data, learn to read them — they're the center of every screen in the app.
The five, in one breath
| Score | The question it answers |
|---|---|
| Profitability | Does it make money, and does it make it consistently? |
| Risk | How well does it protect your capital on the way? |
| Confidence | Is there enough evidence to trust these results statistically? |
| CSI | The bottom line — a risk-first composite of the three above |
| Health | Is the strategy still behaving the way it used to? |
Think of the first four as a job interview (is this strategy worth hiring?) and Health as the performance review (is it still doing its job?).
Read them in this order
- CSI first. It's the single number that balances profitability, risk, and reliability. A strategy can't buy a good CSI with returns alone — weak risk control or thin evidence drags it down by design.
- Health second. A great CSI on a strategy that's actively degrading is a trap. Health compares recent behavior against the historical baseline.
- Then the three components. If CSI is lower than you expected, Profitability, Risk, and Confidence tell you which dimension is the weak leg.
The tiers
All scores share the same color language:
- Excellent (80–100) — top-tier; hard to reach on purpose
- Good (60–79) — solid, with room to improve
- Caution (40–59) — needs attention before real capital
- Failed (0–39) — does not meet the quality bar
Health adds a fifth tier, Critical, for strategies showing active degradation.
Info
A 55 is not an insult. Most honest backtests land in Caution somewhere. The tiers exist to direct your attention, not to pass final judgment — a Caution Confidence score often just means "not enough trades yet."
Circuit breakers: when a score is capped
Sometimes you'll see a score that seems "stuck" low. That's usually a circuit breaker — a hard safety rule that caps the score when something disqualifying shows up, such as an excessive drawdown, no measurable edge, or an unrealistically perfect win rate. The Quality Report flags exactly which breaker fired, so you always know why a score is what it is.
Component gates work similarly: CSI can't exceed its weakest failing component. A strategy brilliant at two dimensions and broken at the third is still broken.
Tip
See this live: open Sample_Gate_Blocked_EMD_Daily_Long from your samples and look at the Strategy Quality section — it exists specifically to show you what a gated score looks like.
Common beginner questions
Why is my score low when my metrics look fine? Because single metrics lie by omission. Great net profit with 40 trades has a Confidence problem. Great Sharpe with a 45% drawdown has a Risk problem. The CSI deep dive covers this in detail.
Can two strategies have the same CSI but be very different? Yes — that's what the three components are for. Same composite, different weak legs, different fixes.
Do the scores tell me how much to trade? No. Scores describe validation quality; position sizing is a capital question — that's what Monte Carlo Capital Sizing and Portfolio Studio's Robustness tab are for.
Go deeper (optional)
Each score has its own reference page: CSI · Profitability · Risk · Confidence · Health
Next lesson
Time to get your own data in: Importing Your First Strategy →