Portfolio backtester (UK / GBP investors)

Use real historical data to simulate how different retirement portfolios would have performed — as if you had retired in 1872, then 1873, then 1874, and so on through history. Everything is shown in today's money (inflation-adjusted).

How to use it: set your starting assets and withdrawal rate, then move the cash, gold and bond sliders to build a mix. The chart shows how that portfolio fared starting in each year of history; the failure rate is how often it ran out. The optimiser suggests the lowest-failure mix for your settings.

Withdrawal & horizon spending

£40k in the first year, then grows with inflation each year.

Equity universe which market, and adjustments

Global is a spread of developed-world stock markets — a more realistic basis, as it includes the European and Japanese crashes the US avoided. US only is the S&P 500, the best-performing major market of the 20th century and so a rose-tinted choice.
Leave unticked to hold overseas assets in their own currencies (most common). Tick if you currency-hedge — the pound fell against most currencies over this period, so hedging generally made things worse.
Shares today are historically expensive, and starting from a high valuation has tended to mean weaker returns over the following decade. When ticked, the early years' equity returns are reduced to reflect that — automatically using a larger reduction for the US market (dearer) than for Global. See "How it works" for the figures.

Withdrawal strategy which pot funds spending

Equity-last + replenish: in up years, sell equity for spending and sweep some of the year’s equity gain into the buffers (cash, then bonds, then gold) to refill toward the inflation-adjusted target. In down years, draw the buffers in order (cash, then bonds, then gold) so you never sell equity into a drop.
Failure rate
9.1%
start-years where the pot ran dry
vs optimum
5.0% worse
lowest found: 4.1%
vs all-equity
5.0% worse
all-equity: 4.1%
Median end wealth
£923k
real GBP
10th-%ile
£56k
bad-case (not ruin)
Terminal real wealth by start year — strategy ┄ all-equity · ┄ start · — lowest-ruin opt · 1872–2022 · ruin
£100k£1.00m£10.00mstart £1.00m187218871902191719321947196219771992
Material failure risk (9.1%). At 4.0% the withdrawal is too high for this portfolio to reliably last 30 years. Here the buffers slightly raise failures versus all equity — the safe assets dragged on growth. The main fixes are a lower withdrawal rate or turning on spending flexibility (cutting spending in bad years).

Allocation buffers + equity

Earns the real historical UK cash rate each year. Positive most decades; deeply negative in the 1970s.
Ladder uses real UK 0–5yr gilt total returns (Stocker 2024, from the Heriot-Watt/IFoA/ESCoE database, 1870+). A slightly conservative proxy for a held-to-maturity ladder, which would feel marginally safer in rate-spike years. Over the long run it lands within a whisker of cash.
Earns the actual real GBP gold return. Flat under the gold standard; powerful in inflationary crashes (1973–74: +70%), inert in deflationary ones.
Long duration: savaged in inflationary regimes (−27% real in 1974, −29% in 2022), but the one buffer that rallies in deflationary crashes (+16% in 1930, +12% in 2008) when central banks cut hard.
Your mix
60%
20%
20%
Equity 60% Gold 0% Cash 20% Bond buffer 20%
The default — 60% equity, 20% short gilts, 20% long bonds — approximates a "60/40" lifestyle or target-date fund (e.g. Vanguard LifeStrategy 60), whose 40% bond sleeve spreads across short, medium and long maturities. Set gold and the buffers to zero for a pure-equity portfolio.
Optimiser the best risk/reward mixes for these settings · updates automatically
Safest mix (lowest failure rate): cash 0% (savings) / gold 0% / bonds 0% → fails 4.1% of the time, median £2.51m
CashTypeGoldBondsBond typeEquityRuinMedian10th-%ile
0%—0%0%—100%4.1%£2.51m£0.43m
0%—5%0%—95%5.0%£2.53m£0.33m
Each row is undominated. Click one to apply it (cash and bond types are both tried). Reflects the current withdrawal rate, strategy, index, horizon and valuation.

Things to try experiments worth running

Move the cash / gilt-ladder, gold and long-bond sliders one at a time and watch how each shifts the curve in a different way.
The starting mix already models a "60/40" lifestyle / target-date fund. Drag all three buffers to zero for 100% equity and compare — heavy bond allocations, the norm as these funds de-risk near retirement, generally perform poorly in this backtest.
Test the famous "4% rule". Set 4% with US-only, unhedged, valuation adjustment off: it looks safe. Now switch to Global, or tick hedging, or allow for today's high prices — the failure rate climbs. The 4% rule was largely derived from the exceptional US experience; it travels less well to a global, sterling, expensive-market starting point.
Watch the "vs optimum" figure as you slide. It shows how far your mix is from the lowest-failure-rate allocation the optimiser found — often surprisingly little, which is the point: withdrawal rate and flexibility matter far more than fiddling with the mix.
Turn on "cut spending in a crash" at a high withdrawal rate. Spending flexibility usually rescues a plan more than any change to the asset mix does.

How it works method & data sources

For each possible retirement start year the tool runs your portfolio through the full horizon against what actually happened next, taking out your inflation-adjusted withdrawal each year via the chosen strategy. Failure means the pot hit zero before the end; those start years show a red band on the chart. Everything is in today's money: returns are converted to sterling and adjusted for UK inflation, so the withdrawal rate is a real, inflation-proofed income.
Currency: "unhedged" holds overseas assets in their own currencies and applies the historical sterling exchange-rate path; hedging removes that currency movement.
Valuation adjustment: based on the CAPE ratio — share price divided by 10-year average inflation-adjusted earnings — which is a rough guide to how dear a market is. A high starting CAPE has historically been followed by weaker 10-year returns; fitting that relationship on US data since 1881 gives roughly: expected 10-year real return ≈ 0.3% + 0.90 × (1 ÷ CAPE). When the adjustment is on, the tool works out the gap between that CAPE-implied figure and the long-run average (~6.7%/yr) and subtracts it from the first 12 years' equity returns (largest in year one, fading to zero by year 12). It auto-selects the starting CAPE for the chosen market — about 38 for the US (expensive) and about 29 for Global (developed-markets, less so) as of early 2026 — so US gets a larger reduction. It lowers the average early return only; it does not raise the odds of a crash, which dear markets also tend to do, so if anything it understates the risk of starting expensive.
Data — equity: US is S&P real total returns (Shiller, 1872+). Global is GDP-weighted developed markets (Jordà-Schularick-Taylor Macrohistory, 16 countries, 1872–1969) joined to MSCI World (1970 on).
Data — other: cash, inflation and exchange rates from the Bank of England "Millennium of Macroeconomic Data"; short gilts (0–5yr) from Stocker (2024); long gilts (10yr) and global government bonds (GDP-weighted, GBP-hedged) from Jordà-Schularick-Taylor; gold from historical year-end USD price converted to real GBP (price history). Use the links at the foot of the page to inspect every series and run the test suite.
▸ Run test suite ▸ View / copy raw data
Caveats — why not to rely on this alone: it tests the handful of overlapping sequences that actually happened, which is a thin and non-independent sample, so the failure rates are illustrative rather than precise — the future can be worse than anything on record. The valuation adjustment captures lower average returns but not the higher crash risk that dear markets carry. It ignores tax, dealing costs and which account (ISA/SIPP/GIA) you draw from. The Global market data carries a mild upward survivorship bias (markets that vanished aren't in it). A teaching instrument, not financial advice.
Privacy: this page collects no data, uses no analytics or trackers, and sends nothing to any server — all calculations happen on your device. The only thing it can store is your own settings, and only if you tick "Keep my settings" above; that is saved locally in your browser (never transmitted) and is deleted the moment you untick the box. There is no account and no sign-up.