How Monte Carlo works in Haven

A single projected line on a chart hides the most important question: what happens when the market doesn't cooperate? Monte Carlo answers that question by running thousands of versions of your future, each with a different sequence of returns, and reporting how many of them succeed.

What "success" means

A run is a success if your portfolio is still positive at the end of your retirement horizon, given the spending you specified. A failure means you ran out before the horizon ended. The Chance of Success number you see on the dashboard is the percentage of runs that succeeded.

So 92 % doesn't mean your portfolio grows 92 % of the time. It means in 92 out of 100 simulated futures, the money lasts as long as you planned for. The other 8 are runs where bad sequence-of-returns risk, especially early in retirement, draws down the portfolio faster than it can recover.

Where the returns come from

Haven uses Shiller's historical S&P 500 dataset (the canonical academic source for long-run US equity returns) plus historical bond returns and inflation. Each simulated year draws a return from the historical distribution, with volatility that matches what markets have actually done — not a smooth average.

For Canadian portfolios, returns are blended with Canadian equity and bond data so a TFSA holding XIC isn't modeled as if it were the S&P 500.

We don't draw from a normal distribution. Real markets have fatter tails than a bell curve — big up years and big down years happen more often than a normal distribution would predict. Using historical data preserves that reality.

How many runs

Haven runs 1,000 simulations by default. That's enough to stabilise the success number to within about ±1 percentage point. More runs would make the number smoother but wouldn't change the answer in any meaningful way.

If you have a Pro or Premium subscription you can increase the run count for tighter confidence bands on the projection chart, but the success number you make decisions on stays roughly the same.

What the projection bands mean

The shaded violet area on your projection chart isn't "where your money will be" — it's the spread of where your money could be. The thick line in the middle is the median run. The outer band is roughly the 10th to 90th percentile. The inner band is roughly the 25th to 75th.

Read it as: "in the middle 50 % of futures, my portfolio is somewhere in this inner band at this date." Half of futures fall outside it. The point is to make the uncertainty visible, not to pick a single number.

What number to aim for

Rough guidance, not a rule:

• 90 % and up — robust. Your plan survives normal market volatility.
• 70 % to 90 % — workable. You'll likely succeed but a bad sequence of early-retirement returns could force adjustments.
• Below 70 % — fragile. Worth either pushing the FIRE date, lowering target spending, or saving more.

Don't chase 100 %. To get there you usually have to over-save by a meaningful margin, which is its own cost (time you don't get back, lifestyle you don't enjoy now). For most people, the right answer is "high enough that I sleep at night, low enough that I'm not delaying my life unnecessarily."

What to actually do with a low number

If your Chance of Success is lower than you want, the levers in rough order of impact:

1. Push the FIRE date out. Each extra year of contributions plus market growth has compound impact. A two-year delay often does more than a 10 % savings-rate increase.
2. Lower target spending. Going from $80k/year to $72k/year retirement spending is a roughly 10 % cut in the target portfolio you need.
3. Save more. Helps but takes the longest to compound.
4. Adjust the asset mix. Higher equity allocation usually raises expected returns and median outcomes. It also raises the chance of bad early years. Don't reach for return when you should be reaching for time.

If your Chance of Success is comfortably above 90 % and you keep adjusting trying to get closer to 100 %, that's the system telling you that the plan is already fine. The next thing to optimize is probably not the math.