# Sensitivity Analysis

Sensitivity analysis tests how sensitive your optimal decisions are to changes in input assumptions. The core principle is ensuring your solution remains optimal across a range of plausible parameter values - both for projection data and solver settings.

Rather than relying on a single "best" solution, it runs multiple simulations with controlled noise injected into projections, xMins and settings, then measures which recommendations remain robust across different scenarios.

This helps answer the key question: "How confident should I be in this recommendation if my projections are slightly off?"

Warning

This is a computationally heavy process. Not recommended on mobile devices, though technically possible.

# How It Works

Available in: Both Transfer Solver and Linear Optimiser

Process: Runs multiple solver simulations with slightly varied inputs, then evaluates which moves appear most consistently across runs.

# Settings

# Simulations

Number of simulated solves to run with varied inputs.

Default: 20 Recommended: 20-100

# Sim Noise (Global)

Overall noise level applied across all categories when set.

Options: 0-5 (None, Low-, Low, Medium, High, High+)
Default: 3

# Focused Noise Controls

When Sim Noise is not set, you can control specific areas:

Rating Noise: Varies team & player ratings xMin Noise: Varies expected minutes while maintaining coherency (nailed players more stable, uncertain players more variable)
Settings Noise: Varies solver parameters (FT Value, Time Decay, Bank Value, etc.) between simulations to ensure solutions remain optimal across different strategic approaches

# Sim Evaluations

After simulations complete, re-evaluates the top N root moves using default (noiseless) scoring.

# Scoring System

Root moves are scored based on simulation performance:

Top root move in a sim: 3 points
Second best root move: 2 points
Third best root move: 1 point

Note

For the Linear Optimiser, scoring adapts to the number of Solve Lines: 4 lines = 4, 3, 2, 1 points; 3 lines = 3, 2, 1 points, etc.

# Output

Normal solves: Ordered table of root moves ranked by total score WC/FH solves: Weighted proportion showing how often each player appeared in top plans

# Best Use Cases

Risk assessment: Identify which recommendations are robust vs fragile
Idea exploration: xMin noise particularly highlights players with high upside potential if your xMin beliefs are optimistic
Confidence building: See which players consistently appear despite input variations

Note

If a move appears in most simulations, it's likely robust. If it only appears in the regular noiseless solve, it may be marginal and sensitive to small changes.