Core concepts
Key concepts
| Concept | Description |
|---|---|
| Assumption | An uncertain input cell. Each simulation iteration, this cell is overwritten with a value sampled from its distribution. |
| Output | A result cell or range. Its value is recorded after each iteration. |
| Iteration (rep) | One complete recalculation of the workbook with sampled assumption values. |
| Config | A saved set of assumptions and outputs for a workbook, so you don't have to re-configure each session. |
Workbooks and configs
| Concept | Description |
|---|---|
| Workbook | An .xlsx file you upload as the simulation model. Pocketstats renders it in IronCalc — a read-only in-browser spreadsheet engine. |
| Assumption cell | A cell whose value is replaced by a sampled distribution on each simulation iteration. Highlighted in blue. |
| Output cell | A cell whose value is recorded after each iteration. Highlighted in green. |
| Config | A saved set of assumptions and outputs for a workbook. Configs persist across sessions so you don't need to re-configure on each visit. |
Assumptions and distributions
Assumption — a single cell you've tagged as uncertain. Instead of a fixed value, the cell receives a random sample on each iteration drawn from the distribution you choose.
Distribution — a mathematical model of uncertainty. You pick one that matches what you know: use normal when values scatter symmetrically around a mean, triangular when you have a low/most-likely/high estimate, uniform when any value in a range is equally plausible, and so on.
Sampling — on each iteration the simulation engine draws one value per assumption and writes it to the cell before recalculating the workbook. After all iterations finish, the spread of output values reflects the combined uncertainty of all your assumptions.
Replications
Replication (rep) — one complete pass through the model. The engine draws one random value per assumption, recalculates the workbook, and records each output. A single rep produces one data point per output.
Iteration — another name for a replication. The two terms are used interchangeably.
The more reps you run, the more stable your statistics become. A small number of reps (100–1,000) gives you a rough picture. Larger numbers (5,000–10,000) produce smoother histograms and more reliable percentiles.
Reading results
In a Monte Carlo simulation, each output cell produces many sampled values — one per replication. The histogram displays the frequency distribution of these values, showing which outcomes are more or less likely.
- Histogram shape: A narrow, tall peak means low uncertainty (most iterations produced similar values). A wide, flat shape means high uncertainty (values varied widely).
- Distribution center: The median (P50) and mean tell you the central tendency of the output. If they differ significantly, the distribution is skewed.
- Spread: Standard deviation and the range between P5 and P95 quantify how much the output can vary.
- Pr(x < X): The probability that the output falls below a threshold you define — useful for answering questions like "What is the chance we lose money?"