Fixed vs. Random Seed

 

Problem
You want to understand the difference between a fixed seed and a random seed in @RISK, and how seed choices affect reproducibility and simulation results.

 

Random Number Generation
@RISK uses a portable subtractive random number generator with a very long cycle and effectively infinite period. Random variables are generated using inverse transform methods. By default, the starting seed is clock dependent.

 

Seed Values
In the Simulation Settings dialog, you can choose the random number seed:

• Fixed Seed: Enter an integer between 1 and 2,147,483,647. Using a fixed seed ensures that the simulation produces the same results every time it is run (assuming the model is unchanged and no external random factors are added).

• Random Seed (Choose Randomly): The seed is chosen based on the computer’s clock, producing different results each simulation.

 

Reasons to use a fixed seed

1. Model development: See clearly how changes to the model affect results.

2. Reproducibility: Share the model so others get the same results when running the simulation (on the same @RISK version and identical model).

RiskSeed() function
You can assign a RiskSeed() to an individual input distribution, giving it its own independent random number sequence. RiskSeed() is ignored for correlated distributions.

 

Multiple Simulation Runs

• If Multiple Simulations Use Different Seed Values is checked:

  • Choose Randomly: each simulation uses a new seed.

  • Fixed: each simulation uses a different seed, but the sequence is the same each time the run is executed.

• If the box is not checked:

  • Choose Randomly: all simulations use the same seed for a run, different seed for the next run.

  • Fixed: same seed is used both within and between runs.

 

Monte Carlo vs. Latin Hypercube
The sampling method (Monte Carlo or Latin Hypercube) is set in Simulation Settings. Changing the sampling method or the number of iterations can change the results, even with a fixed seed, except for distributions with RiskSeed() assigned.

 

Reproducibility Tips

• Avoid Excel functions like =RAND(), NOW(), TODAY() or non-reproducible macros.

• Keep the workbook structure consistent (avoid moving sheets or adding workbooks).

• Multiple CPU: Reproducibility can be affected if your model references other iterations. Disable multiple CPUs if using inter-iteration references.

• Different @RISK versions typically produce results within normal statistical variation.

 

References

Donald E. Knuth, Seminumerical Algorithms, 3rd Ed., 1998.

William H. Press et al., Numerical Recipes, 1986.

 

Related Articles:

Latin Hypercube Versus Monte Carlo Sampling

replacing RAND() functions with RiskBernoulli() or RiskUniform()

Different Results with Same Fixed Seed

Different Results with Multiple Workbook Copies

What Was My Random Number Seed?

Random Number Generators