Enter three-point estimates for your tasks and run thousands of simulated projects to see the range of finish times and the chance of meeting a deadline.
A Monte Carlo simulation runs your project thousands of times, drawing each task's duration at random between its optimistic and pessimistic estimates. For three tasks the PERT mean is 16.3 days, but the simulation shows P50 16.1, P80 18.4 and P90 19.6 days: the plan's real finish range.
Enter each task's optimistic, most likely and pessimistic duration, as in a PERT estimate. The simulator treats the tasks as running one after the other, draws a duration for each task at random, adds them up, and repeats thousands of times. The results describe the whole spread of possible finish times.
The random seed makes a run repeatable: the same tasks, distribution, number of runs and seed give the same numbers, and the seed is saved in the link you copy. "Re-run with a new seed" shows how stable the percentiles are. This is the sampling companion to the PERT calculator, which gives a single mean and standard deviation by formula.
beta-PERT: X = O + (P − O) × Beta( 1 + 4(M − O) ÷ (P − O), 1 + 4(P − M) ÷ (P − O) )
triangular: X between O and P with its peak at M total = X1 + X2 + … + Xn
Pk = the value below which k% of the sorted totals fall
Beta-PERT concentrates draws near the most likely value, with a mean of (O + 4M + P) ÷ 6, matching the PERT formula. Triangular gives wider tails and a mean of (O + M + P) ÷ 3. Random numbers come from a seeded generator (mulberry32), so results are reproducible. Large runs execute in a Web Worker so the page stays responsive.
Each example uses numbers you can load into the calculator above; the results match the tool exactly.
Design 2 / 4 / 8 days, build 5 / 8 / 17 days, test 2 / 3 / 4 days. The PERT formula gives an expected 16.33 days with a standard deviation of 2.26.
These are the simulator's default inputs. The median is slightly below the mean because the build task has a long tail.
Switching to triangular keeps the same tasks but gives the tails more weight.
The deadline looks much riskier, which shows how much the choice of distribution matters.
P50 is the median: you finish before it half the time. Committing to it means being late half the time, so teams that want confidence quote P80 or P90. The distance between P50 and P90 is the risk buffer the estimates imply: 3.5 days in example 1. If it is larger than your schedule contingency, the plan is optimistic.
The simulator treats tasks as sequential and independent. Real tasks overlap and share causes of delay, which makes the true spread wider than the independent-sum result. Use it for a defensible range, and use the critical path calculator when tasks depend on each other in a network.
A method that runs the project many times with random task durations drawn from your estimates, then reports the spread of results, such as the 80th percentile finish time.
The durations that 50%, 80% and 90% of simulated runs finished within. P80 means that 80% of simulated runs were done by that time.
PERT calculates a mean and standard deviation by formula and assumes a normal curve. The simulation samples the distributions directly and shows the real shape, including skew from long pessimistic tails.
Both use your three estimates. Beta-PERT clusters outcomes near the most likely value; triangular spreads them more evenly, so its tails and percentiles are wider.
A new random seed gives a different random sample. With 10,000 runs the percentiles move by a small amount, which shows how stable they are.
10,000 is a good default. 5,000 is fine for a quick look and 20,000 gives smoother percentiles.
Last updated 19 October 2026. Last reviewed 19 October 2026. Reviewed by Umesh Manral. Formulas and tests: methodology.