@RISK FAQ: Iterations, Simulations, and RISKOptimizer Trials

Applies to: @RISK 6.x and Up
“Trials” applies to: @RISK Industrial

 

Iterations vs Simulations

Iteration: The smallest unit of a simulation. At each iteration, @RISK draws random values for all distributions, recalculates your model, and stores output values.
Simulation: A collection of iterations. At the end of a simulation, @RISK generates reports and histograms summarizing all iterations.

Example: 5,000 iterations and 3 simulations → 3 histograms for each output, each summarizing 5,000 iterations.

Guidelines: Typically, use N iterations and 1 simulation. One simulation with 15,000 iterations is usually more stable than three simulations of 5,000 iterations each.

When to use multiple simulations:

• Testing different controlled variables (e.g., prices or raw materials) using RiskSimtable functions.

• Embedding multiple scenarios or assumptions in RiskSimtable for separate simulations.

• Testing model stability by running multiple simulations with the same assumptions.

If simulation results vary widely across runs, your model may be unstable or you may need more iterations.

RISKOptimizer Trials

Trial: One run of the simulation using a set of values for adjustable cells in the optimization model.
Valid Trial: A trial that satisfies all hard constraints. Invalid trials are discarded.

Key points:

• RISKOptimizer uses results from previous trials to choose values for subsequent trials.

• The number of valid trials can be less than the total number of trials if some violate constraints.

• A low proportion of valid trials suggests restructuring the model for faster optimization.

 

Related Articles:

Random Number Generation, Seed Values, and Reproducibility

For Faster Optimizations