Casey Score Index (CSI)

The CSI is AlgoChef's master composite score. It combines Profitability (P), Risk (R), and Confidence (C) into a single 0–100 number using a proprietary weighted formula, with safety mechanisms to flag dangerous strategies.

The formula ensures that a zero in any component pulls the entire score down — you can't compensate for terrible risk with great profitability.

Safety Mechanisms

Beyond the composite formula, the CSI includes multiple layers of protection to ensure the final score reflects real-world tradability. These include circuit breakers that cap the score when fundamental problems are detected, priority gates that let a critically weak sub-score override the composite result, and quality warnings that flag concerns for manual review.

Examples of conditions the CSI guards against:

• Insufficient trades — a strategy without enough history can't be scored reliably
• Excessive drawdown — severe capital loss disqualifies a strategy regardless of other metrics
• No statistical edge — if returns show no real profitability, the score is forced into the Failed tier

If any sub-score is critically low, it can override the composite formula so that a single strong dimension never masks a dangerous weakness.

CSI tiers and status

The CSI maps to four tiers. Status labels describe validation and monitoring posture — not how much capital to deploy (that is your decision outside AlgoChef).

Why is my score low when my metrics look fine?

CSI weighs different metrics with different weights — and it isolates a composite score for each dimension. For example, in a metric like Return/Drawdown, the return component feeds Profitability but not Risk, while the drawdown component feeds Risk but not Profitability. So a single metric you see on the dashboard can be split across dimensions, weighted differently in each.

That means a strategy where every visible metric looks acceptable can still score low if the dimensions CSI weighs most heavily — especially risk — are mediocre rather than strong. Acceptable across the board doesn't mean strong across the board.

How is the score calculated?

CSI evaluates three dimensions — profitability, risk management, and statistical confidence — using multiple metrics per dimension. Risk carries the heaviest weight because that's what blows up accounts. Each dimension has a gate: if any one fails, the entire score fails. There are also circuit breakers that catch extreme conditions like excessive drawdown or no statistical edge.

Think of it like a credit score: you know the categories FICO evaluates (payment history, credit utilization, length of credit history), and you know the rating scale. But you don't know their exact weights or internal formula — and you don't need to. What matters is that the methodology is sound, consistent, and that you understand what drives your score up or down.

Same philosophy here. The tiers (Excellent, Good, Caution, Fail) and circuit breakers are fully transparent. The internal weights are not — and that's intentional.

How Robust Is Your CSI?

The Monte Carlo Simulation tests whether your CSI stays stable across 25,000 randomized scenarios. If the tier holds up (e.g., stays "Good" in 90%+ of simulations), your edge is robust. If it drops to "Caution" or "Failed" under randomization, the backtest may be misleading.