Show HN: I made a retirement simulator: 125 years of data, 25 countries

nnhootan about 4 hours ago 32 commentsRead Article on foresightplanner.com

FR version is available. Content is displayed in original English for accuracy.

⚡ Community Insights

Discussion Sentiment

77% Positive

Analyzed from 1484 words in the discussion.

Discussion (32 Comments)Read Original on HackerNews

bryanlarsen•about 3 hours ago

It's really nice to see a simulator that removes the home country bias. So many of these types of simulations use American data. But that's pretty extreme cherry-picking: America has had a pretty exceptional run these last 100+ years. Any retirement simulator cannot assume that America's exceptional growth will continue. Maybe it will, but you cannot plan on a maybe.

nhootan•about 3 hours ago

Thanks, that's most of why I built it. A US only backtest is one draw where you already know who won. This is 25 countries over 125 years, block bootstrapped with the domestic country resampled, so something like Japan after 1990 shows up.

bryanlarsen•about 2 hours ago

Do you have Germany 1914 in your dataset? There's no safe withdrawal rate for that particular data point...

nhootan•about 2 hours ago

The Germany's data is in since 1901, but there is a gap between 1922-1923, cause the underlying dataset, DMS, has no bonds data for that period.

rachelb1331•about 3 hours ago

I didn't want to put in my email address in order to access the detailed calculator... The quick check doesn't have nearly enough information to give anyone an idea of what actually to expect...

nhootan•about 3 hours ago

Fair, and it's the part I'm least settled on. The account exists because the full plan is a server side run that gets stored, not to build a mailing list.

Being straight about it: the long-run panel is the Dimson-Marsh-Staunton dataset, which is a paid commercial licence, plus the compute. So this won't be free forever. The Quick Check will stay free and anonymous.

The detailed side is where the real inputs are: mortality table life expectancy, CPP/OAS or Social Security start ages, pensions, tax, flexible spending. Four inputs can't tell you much, agreed.

bluGill•about 4 hours ago

This doesn't know the important part. Will I live to 62 as the earliest of my ancestors died, or 98 as the latest? My family history is about average for lifespan but that is a big difference. If I live to 62 I should retire at 50 - or perhaps I shouldn't have bothered saving at all since the growth isn't there. Of medical advances make me love to 150 I should save until I'm 80 which will be a nice income to really enjoy the rest of my life.

I want more defined benefit accounts, but they are hard to find

karmakurtisaani•about 3 hours ago

So the thing is that no one knows that part.

bluGill•about 3 hours ago

For me no, but for a few thousand people it isn't to hard. Which is why I look for defined benefit plans that are not scams, and in frustrated that I can't fibd them

Eddy_Viscosity2•about 3 hours ago

Can't we just get AI to tell us? I'm sure it will confidently give an answer.

nhootan•about 3 hours ago

That's in the full planner, just not the Quick Check. Horizons are sampled per person from actuarial mortality tables, so you get a distribution of lifespans rather than one guessed end age. You can also pin a fixed age and run your 62 and your 98 separately, which given that spread is probably worth doing.

betaby•about 4 hours ago

Better be safe I guess:

Foresight Planner is not yet available in your region. We're launching in Canada (excluding Quebec) first. We don't yet support Quebec, because we don't yet offer the product in French. If you'd like us to let you know when we expand, email hi@foresightplanner.com.

Foresight Planner n'est pas encore disponible dans votre région. Nous lançons d'abord au Canada (hors Québec). Pour être informé·e de l'élargissement de notre service, écrivez à hi@foresightplanner.com.

Region: CA-QC

nhootan•about 3 hours ago

It's Bill 96 and Law 25. No French UI yet, so blocking beats being non compliant. It's near the top of the list.

fusslo•about 3 hours ago

huh, yesterday I vibe coded a similar tool. It lets me set my current savings, investment, retirement. Expected yearly savings, expected number of years of work left. After retirement, only withdrawals are allowed (investments, savings still get returns tho)

Codex suggested a bunch of improvements: monte carlo, investment volatility, expense inflation, capital gains tax etc. One thing codex didn't ask about was social security. So that's another factor that could be improved.

I wanted to know 'do I really need to work 30 more years?'

What would be really interesting is discovering an attainable formula where my investments outpace my expenses. Or, what variables let me die with basically $0 in the bank

At the end, I realized I had no idea what the calculator was REALLY doing. So it could be right... it could be really wrong too. Having historical data might help solve those issues

nhootan•about 2 hours ago

Your last paragraph is why I did using MC simulations based on credible data. Once you pick a return and a volatility, the output just restates your inputs and you can't tell a wrong answer from a right one. Historical data doesn't fix that, but it moves the assumption somewhere you can argue about, and shocks arrive as correlated bundles that actually happened.

Dying with roughly $0 is a real mode in the app: floor and ceiling, or Yale style smoothing, where spending flexes instead of staying fixed.

drop_star•about 4 hours ago

Why do I need to sign into everything these days? The moment I see that I close the site.

nhootan•about 2 hours ago

The sign in is required for detailed plans to save the plans. Those plans also incure real comupte on the server. QC results are based on a cache in the browser and don't need sign in.

albumen•about 4 hours ago

You don’t need to for its quick check, which shows the core feature.

bluGill•about 4 hours ago

You need it to dig deeper. I have a pension which I can get at 65.

I have options to collect social security. What if I draw my 401k first and wait until 72 to start as? Or I can get take as at 62 and the 401k grows more...

Edit point is you need an account to start asking those questions.

nhootan•about 3 hours ago

The timing half is there: CPP/OAS or Social Security start ages, and pension start and end ages, so claiming at 62 versus deferring is something you can run both ways today.

Draw order across accounts isn't exposed yet. Right now the engine allocates by equalizing marginal tax rates across accounts each year rather than draining one at a time, so it's solving that ordering instead of letting you set it. The manual override is a fair ask and it's coming, the engine already takes the order as a parameter and it just isn't wired to the UI.

rconti•about 4 hours ago

yeah, this is quite literally the first thing i saw when i opened the page.

BretonForearm•about 4 hours ago

Why not let people try it from abroad? Not in the USA? No luck. Why does the withdrawal rate need to be a multiple of 0.5%? It's curious that the highest chance was associated with the highest (100%) stock ratio.

pelagic_sky•about 3 hours ago

I'm in NZ and it worked. It gave me US figures and numbers though. And I find there are existing FIRE calculators already out there that are tried and true.

nhootan•about 2 hours ago

Glad it ran, but the numbers aren't US. Once you login, "Other, rest of world" gives you a gross projection with tax and benefits off. The returns come from the same 25 country bootstrap everyone gets, with the domestic country resampled rather than pinned to the US.

What's genuinely missing for you is NZ tax and NZ Super. So it worked, but you got the generic version.

On the existing calculators, they're good and I use them too. Most are tried and true against US history, which is the one sample where equities did unusually well. That's the gap I was after.

nhootan•about 2 hours ago

Sorry, that's the EU account gate rather than a geoblock. You can browse and run the Quick Check from anywhere except Quebec, but account creation is blocked in the EU/EEA/UK/CH, so the full builder is out of reach.

The reason is that GDPR has extraterritorial reach and real penalties, and doing it properly is weeks of work I haven't done for a market I'm not launching in yet. I'd rather block the signup than pretend I'm compliant. Not a satisfying answer if you're the one blocked, I know.

The 0.5% step is because the Quick Check is a 66 cell precomputed grid shipped to your browser, so it runs with no account and no server. The builder takes dollar amounts. On residence, there's an "other, rest of world" mode that runs an untaxed gross projection, so it isn't US only.

The 100% stocks result holds up. In the international panel, bonds get wrecked by inflation more often than equities do, so more stocks wins on "did the money last" while being much rougher along the way. Cederburg and co-authors find the same. That's why the full detailed results show the full distribution and the drawdowns, not one number.

memoryleakgame•about 4 hours ago

So its a monte carlo simulation but I can't enter growth rate country time horizons or anything like that?

Brother I have a better claude code vibe app than this in an hour

You need to really talk to a financial planner and learn what the proper inputs are

jawns•about 4 hours ago

I got 82.4% regardless of whether I started with a portfolio value of $1M or $3M.

Something is wrong with the calculations/simulations.

If I can't trust the Quick Check, why would I trust anything else on the site?

Eric714•about 3 hours ago

That's because the withdrawals are 4%. If you have $3M, maybe lower the withdrawals to 1.25%. That would be the same as 4% of $1M.

jawns•about 3 hours ago

I don't think that's accurate.

If I withdraw 4% of a $1M portfolio, I only have $960K left to compound.

If I withdraw 4% of a $3M portfolio, I should have $2.88M left to compound.

If the question is "Will your money last?" I would not expect the simulation results to be identical.

Edit: I failed here. Downvotes are right. I simulated it myself and saw the results. Leaving this comment here so others can learn from my mistake.

Eric714•about 4 hours ago

Same here

nhootan•about 3 hours ago

That result is right. The Quick Check takes a withdrawal rate, not a dollar amount, and runs with taxes off, so it's scale invariant: $1M at 4% and $3M at 4% are the same simulation times three.

That's what lets it be instant and anonymous, since the whole answer space fits in a small grid I ship to your browser. It stops being true once taxes are on, which is what the full builder does. The page should say so. I'll add a line.

nhootan•about 4 hours ago

I got tired of retirement calculators that either assume one fixed return forever, or run Monte Carlo by pulling each year independently out of a bell curve. Real markets don't behave like that. Bad years cluster, and a crash in your first three years of retirement is a completely different problem from the same crash at 85.

So this one samples multi-year blocks out of actual history instead: 25 countries, 1900 to 2025, reassembled into thousands of possible futures. It's a stationary block bootstrap, which is the standard approach in the lifecycle-investing literature. When a plan fails it tells you which way it failed, whether that was weak growth, bad sequencing, or inflation at the wrong moment.

Canada and the US have full tax and benefit modelling (RRSP/TFSA and CPP/OAS on one side, 401(k)/IRA and Social Security on the other). Everywhere else you get a pre-tax projection. If you're in the EU or UK you can run the Quick Check, but you can't sign up yet, because I haven't finished the compliance work.

There's no account linking, and the Quick Check needs no signup at all. The methodology is written up at /methodology. It's an educational tool, not financial advice.

It's free right now. Paid plans come later, and there's a founding rate if you want to lock one in early: $50/year for as long as you stay subscribed, first 50 people. https://buy.stripe.com/14A7sN6Gycny54rdXv5Rm01 Nothing is gated behind it today.