SPE Simulation and Comparison

Simulation is where SPE converts structure into decision evidence. The engine samples variable states, propagates impacts, and produces outcome distributions that can be compared across scenarios.

How propagation works

1. Each variable starts with an initial probability over NN, N, O, P, and PP.
2. A strategy scenario applies direct impacts to Drivers.
3. The simulation samples variable states.
4. The Cross-Impact Matrix uses patterns to propagate effects from source variables to target variables.
5. Processes mediate effects and Outcomes collect final results.
6. Repeated Monte Carlo runs produce probability distributions for each Outcome.

Initial probability

Initial probability is the model's starting belief about each variable before scenario impacts and downstream propagation. It can be uniform, polarized, or user-defined. It should reflect the team's current knowledge, not a desired result.

When a variable has no strong evidence, the default uniform distribution is acceptable for a draft model. When evidence exists, adjust the distribution and document the rationale.

How a scenario is calculated

A scenario belongs to a strategy and contains the direct Driver impacts for that strategy state. During calculation, those impacts become the entry condition for the model. The simulation then propagates the impact through the selected patterns and produces distributions for the Outcomes.

Simulations are started from the Simulation tab. Sensitivity and Strategy Results use saved simulation output; they are review and interpretation areas, not separate places to spend compute.

Four simulation processes

SPE provides four simulation processes. They are not interchangeable. Each process asks a different question about the same model.

Generic simulation

Use Generic simulation first. It runs the model as currently defined: initial probabilities, cross-impact patterns, active strategies, and feasible scenarios.

Generic simulation answers:

• Given the current model, what state distributions are most likely?
• What do Outcomes look like before applying stronger scenario assumptions?
• Is the model structurally complete enough to support comparison?

This is the main baseline run. When the model has active strategies and feasible scenarios, Generic simulation can also provide the first source for Strategy Results.

Base Modifiers simulation

Base Modifiers simulation applies strategy scenarios through the configured base modifier matrix. This treats scenario impacts as tendency adjustments rather than hard overrides.

Use it when a strategy should influence a Driver but should not completely force the Driver into a single state. This is useful for softer interventions, uncertain implementation effects, or strategies whose impact depends on context.

Compare Base Modifiers against Generic results to see whether the recommendation remains stable when strategy impacts are interpreted more gradually.

Direct Impact simulation

Direct Impact simulation applies the scenario's selected Driver state directly before cross-impact propagation. It is a stronger treatment than Base Modifiers.

Use it for stress tests, strong interventions, policy mandates, operational commitments, or scenarios where the team intentionally wants to force a Driver toward NN, N, O, P, or PP.

Compare Direct Impact against Base Modifiers to understand whether the conclusion depends on a hard scenario assumption. If the recommendation changes sharply, document that dependency.

Sensitivity simulation

Sensitivity simulation forces each Driver and Process through NN, N, O, P, and PP, then measures how much Outcomes move. It is not intended to select a winning strategy.

Use sensitivity to identify:

• high-leverage Drivers and Processes
• fragile assumptions
• relationships that need better evidence
• variables that can change the strategy comparison

Sensitivity is usually the heaviest simulation process because it repeats runs across variable-state combinations. Start with fewer iterations while the model is still changing, then increase only when the structure is stable.

Iteration guidance

Use small runs for setup checks and larger runs for final comparison. Avoid rerunning high-iteration simulations after every small edit.

Albarena caps SPE simulations at 100,000 iterations per run to prevent accidental or abusive compute use. If results vary materially between runs, review the model before increasing iterations: initial probabilities, cross-impact patterns, strategy scenarios, and Outcome scales are usually more important than simply adding more samples.

Comparison over prediction

The most useful output is often not "the model predicts 72." It is "Scenario 1 is 9 points higher than Scenario 2 on Outcome A, but 4 points worse on Outcome B, and the gap remains under sensitivity testing."

Use these comparison forms:

• expected value difference
• probability of reaching a desirable state
• probability of falling into an undesirable state
• frequency of scenario matches in Monte Carlo rows
• robustness under External Agent assumptions
• sensitivity of the comparison to key Drivers or Processes

Interpretation discipline

Treat SPE outputs as decision-support evidence, not as deterministic forecasts. A model with weak assumptions should be used to structure discussion. A model with reviewed variables, justified impacts, and sensitivity checks can support stronger recommendations.