DCT Lexicon · Simulation
Monte Carlo simulation
Where distributions cannot be computed in closed form, they are estimated by simulation: generate many sample paths and read probabilities and averages as frequencies and sample means.
Vol. I, Section 8.9, verbatim (shortened to its first sentence's substance)
In depth
In the book
Where distributions cannot be computed in closed form—which is almost everywhere beyond the linear-Gaussian case—they are estimated by simulation: generate many sample paths and read probabilities and averages as frequencies and sample means. The Monte Carlo estimator of , for a quantity computed from a path, is the sample mean over independent simulated paths. Two classical results make it honest. The law of large numbers guarantees that , and the central limit theorem supplies the error bar: is approximately normal around with standard error , where is the standard deviation of , estimated from the same sample. For a probability, .
Vol. I, Section 8.9 (Monte Carlo Simulation of Enterprise Dynamics), verbatim
Executive example
The square-root law prices precision: one more decimal digit costs a hundredfold more paths. Asked of the model of Worked Example 8.1, the question “what is the chance of breaching the floor?” returns, in three independent runs, 25.0 percent ± 4.3 with 100 paths, 25.2 ± 1.4 with 1,000, and 24.5 ± 0.4 with 10,000 (one standard error, in percentage points; the Worked Example’s 100,000 paths give 24.6 ± 0.14).
Vol. I, Section 8.9, verbatim
Common misreading
A common mistake is to report a design’s chance on its own sample.
Vol. II, LOS 9.5, verbatim
How AXIOM uses it
AXIOM simulates the declared ranges and channels with a stated seed. A preview runs on a few thousand paths and is labelled preview; a certified result runs on 100,000 fresh paths, so a plan chosen on design paths is judged on paths it has never seen. Every estimate is shown with its standard error.
Drafted for the author's approval, from AXIOM's method passports "Stochastic business case" and "Simulator"
In the book: Vol. I, Ch. 8: Stochastic Enterprise Dynamics · Vol. II, Ch. 9: Stochastic Enterprise Optimization
Related: Standard error · Common random numbers · Breach probability · Percentiles (P10, P50, P90) and bands
V1.ME.8.MonteCarlo · V1.Tab.8.4 · V1.WE.8.1 · V2.Thm.9.3


