Monte Carlo trial counts by tier

A reasonable question people ask: why does Haven only run 1,000 simulations by default? Computers are fast. Why not 100,000? The honest answer is that for the number you actually make decisions on, 1,000 is enough.

Why 1,000 is the default

Each Monte Carlo run produces a Chance of Success number — the percentage of simulated futures in which your portfolio survives the horizon. At 1,000 trials, that number is stable to within about ±1 percentage point from one rerun to the next. So a plan that comes back at 87% on one run is going to come back at 86% to 88% on a fresh run with a different random seed.

That's tight enough for the decision the number drives. You don't change your retirement date based on the difference between 87% and 88%. You change it based on the difference between 87% and 65%. The default trial count is sized for the answer you actually need.

What 10,000 trials buys you

On Pro, you can raise the trial count to 10,000. Two things get better:

1. The projection chart's confidence bands smooth out. With more trials, the shaded percentile bands stop having visible roughness at the edges. The median line and the 25th-to-75th and 10th-to-90th bands all become cleaner.
2. The success number tightens by about a factor of three. Standard error of a proportion scales with the square root of the sample size, so 10x the trials gets you about 3x the precision — roughly ±0.3pp instead of ±1pp.

What 10,000 trials does not do is move the success number meaningfully. If 1,000 trials says 87%, 10,000 trials says 87% with more decimal places of confidence. The decision doesn't change.

When to raise it

Bump the trial count up when:

• You're presenting your plan to someone (a partner, an advisor) and you want the chart to look as smooth as possible.
• You're investigating tail behavior — the worst 5% of outcomes — and want a less jittery picture of the bottom band.
• You've made a small change to the plan and want to be sure the success-rate movement isn't sampling noise.

For day-to-day use, leave it at the default. The extra runtime isn't free, and the answer doesn't change.

Related

For what the success number actually means and how to act on it, see how Monte Carlo works in Haven.