How to Read Parameter Sensitivity & Stability

1. Sensitivity Indicator

Each parameter is evaluated for how strongly performance changes as the parameter value changes across optimization trials.

Sensitivity is expressed as a stability coefficient derived from cross-trial analysis. Higher values indicate stronger dependency between parameter selection and outcome.

Parameters are categorized into qualitative stability bands:

• Stable — performance remains largely consistent across tested values
• Reliable — moderate influence on performance
• Needs Review — outcome meaningfully depends on parameter choice
• Fragile — small parameter shifts materially alter results

Fragile parameters increase structural risk, particularly when multiple such parameters interact.

2. Parameter Risk Score

The Parameter Risk Score summarizes overall structural stability.

It combines:

• The highest observed sensitivity level
• The number of parameters requiring review
• The presence of fragile dependencies

The score is scaled from 0 to 100, where higher values indicate stronger robustness and lower overfitting risk.

Penalization increases as sensitive or fragile parameters accumulate. Even if average sensitivity appears moderate, concentrated fragility may materially reduce the final score.

3. Risk Classification

Based on the aggregated score, structural risk is categorized into:

• LOW — parameters exhibit broad stability
• MODERATE — manageable sensitivity; monitor carefully
• HIGH — elevated dependence on tuning
• CRITICAL — structural instability likely driven by optimization bias

This classification reflects deployment risk, not profitability.

4. Deployment Logic

[Parameter Stability] interacts with other structural safeguards before deployment approval.

Deployment may be restricted if:

• Data integrity conditions are not satisfied
• Forward performance shows significant decay
• Parameter stability falls below acceptable structural thresholds

When forward degradation metrics are unavailable due to insufficient validation periods, deployment decisions may be conditionally restricted depending on system governance settings.

The overall deployment verdict considers parameter stability alongside [validation], [risk profile], and execution realism.

5. Governance Metrics (Advisory)

Additional diagnostics may be displayed to assist advanced review, including:

• Signal attenuation patterns
• Risk-adjusted return retention
• Performance drift across regimes
• Tail-risk behavior under stress
• Multi-parameter interaction effects

These indicators provide qualitative insight and do not directly modify the Parameter Risk Score. They are intended for expert evaluation rather than automated gating.

How to Interpret the Block

Prioritize:

1. Presence of fragile parameters
2. Concentration of sensitivity in key inputs
3. Overall structural classification

A strategy with strong returns but high parameter fragility may be less reliable than one with moderate returns and stable parameter behavior.

Robust systems tolerate reasonable parameter variation without structural collapse.