Sampling, seeds and run control

Methods and settings

What the method changes

Monte Carlo draws random input quantiles. Latin Hypercube spreads draws across the probability range of each input, which can improve marginal coverage for a given run count. Neither method supplies a correlation model, fixes incorrect distributions or guarantees accurate tail probabilities from a small sample. The connected financial calculation is the same for each sampled set of inputs.

Record the model as well as the seed

Repeatability requires the same seed, configured inputs, sampling settings and model revision. Changing assumptions while keeping the seed does not reproduce the old economic result. Run details show the context and completion information. Export the model as a separate backup when the assumptions need to be retained; configuration persistence does not preserve the completed sample arrays.

Read progress and cancellation honestly

The simulation runs in the browser with progress, completed runs, elapsed performance and an estimated remaining time. Runtime depends on model complexity, number of inputs, run count and device. Cancel stops further work; completed values may be shown as an explicitly incomplete simulation. Do not compare that subset as though it were the requested full sample, particularly when invalid or incomplete runs are material.