SIPP Retirement Optimiser

Why most online pension calculators are misleading

The standard UK pension projection — the kind every provider shows you in your annual statement — works like this: take your current pot, add your assumed annual contribution, compound at an assumed rate (typically 2%, 5% and 8% real, per the FCA's COBS 13 Annex 2 standardised assumptions), and report what you'd have at retirement. It's a deterministic single-point projection that hides the only thing that really matters: how reliably your plan works.

Two retirees with identical "average" returns can end up with wildly different outcomes depending on when the bad years happen. A 30% drawdown in your first three years of retirement is catastrophic; the same drawdown in your last three years is barely noticeable. This is called sequence-of-returns risk, and it's the single biggest reason why financial-planning point-estimate calculators give a false sense of security.

Academic research on retirement sustainability — from Bengen (1994)'s original 4% rule, through the Trinity Study, to the modern Pfau / Kitces literature — all converges on the same point: you cannot model retirement with averages. You have to model the distribution, and you have to model it with returns that have realistic clustering and fat tails.

What this calculator does differently

This tool runs 10,000 simulated retirement paths through three stages:

1. Accumulation phase. From your current age to your chosen retirement age, your pot grows by a randomly drawn historical return each month, and your annual contribution is added at the start of each year.
2. Decumulation phase. From retirement age to age 95, you withdraw your target real income each year (less any State Pension once it kicks in at age 67). The pot continues to be invested and earns or loses money each month.
3. Outcome scoring. Each path either reaches 95 with money still in the pot ("success") or hits zero before then ("ruin"). The fraction of successful paths is the headline probability of meeting your income target.

Crucially, the returns are not drawn from a Gaussian distribution or a log-normal — they're drawn directly from history, in blocks. We'll explain why that matters in the next section.

The block-bootstrap method

Real market returns aren't independent. A bad year tends to be followed by a not-great year; a great year is more likely to be followed by another good one. Volatility clusters. Inflation persists. None of these features survive if you draw each year's return independently from a normal distribution — yet that's exactly what most "Monte Carlo" pension calculators do.

A block bootstrap preserves these properties. Instead of drawing single years, we draw contiguous 12-year blocks from the historical record. So a 30-year simulated path might be "1929–1940" followed by "1973–1984" followed by "2008–2019". Each block is a real, lived sequence of returns including its autocorrelation, its inflation regime, and its volatility clustering.

We use a 12-year block by default. Shorter blocks underweight long regimes (the 1970s stagflation, the 1980s–90s disinflation, the 2000s lost decade). Longer blocks reduce the number of effectively independent draws and increase Monte Carlo noise. Twelve years is roughly one business cycle and matches the practice in academic retirement-finance papers using UK data (Pfau, Tomlinson; Dimson, Marsh, Staunton).

The dataset itself — UK equity total returns, UK gilt total returns, and UK CPI from 1900 to 2024 — is constructed from the Bank of England's Millennium of Macroeconomic Data, the Jordà-Schularick-Taylor Macrohistory Database, and the Barclays Equity Gilt Study. Summary statistics are cross-checked against the Dimson-Marsh-Staunton numbers published in the annual UBS Global Investment Returns Yearbook. The data file is shipped in the page bundle — /src/data/historical-returns.json — and the source provenance is documented in /src/data/README.md.

Why we don't model "expected returns" as a single number

Two pension calculators can both assume "a 5% real return" and produce wildly different results. The reason is that the same expected arithmetic return implies very different geometric (compound) returns depending on volatility — the well-known volatility drag:

Geometric return ≈ Arithmetic return − ½ × Variance

For UK equities with ~9% arithmetic mean and ~20% standard deviation, the geometric mean is about 7% — a 2-percentage-point haircut every year, compounding for decades. A calculator that quotes "5% real" without saying whether it's arithmetic or geometric is, mathematically, ambiguous to the tune of hundreds of thousands of pounds over a 30-year horizon.

Our simulator sidesteps this entirely because we never use a single-number expected return — we sample real history. The output you see in the dataset card on the right shows the geometric mean of the chosen asset mix after the fee you specified, which is the right number to compare against other tools.

What about the State Pension, tax, and the Annual Allowance?

The State Pension is included as an optional flat real annuity from age 67 (the current State Pension Age for anyone born after April 1960; check your own date at gov.uk/state-pension-age). We use the full new State Pension rate of £230.25/week as projected for 2026/27, equivalent to roughly £11,973/year (DWP rates, April 2026). This assumes 35 qualifying NI years; check your forecast at gov.uk/check-state-pension.

Tax on drawdown is not modelled — your target income is treated as gross. In practice, 25% of your pot can typically be taken as a tax-free lump sum (the Lump Sum Allowance, £268,275 from April 2024, HMRC PTM063300), with the remainder taxed at your marginal rate. Modelling this properly requires the calculator also to know your other income, which we keep out of scope here — use the Take-Home calculator once it's live, or pair this with your accountant's spreadsheet.

The Annual Allowance for 2026/27 is £60,000 (HMRC PTM053100) — the maximum you can contribute across all pensions in a year and still get full tax relief. The input slider caps you at this. Two wrinkles the warnings panel will flag:

• Carry-forward of unused allowance from the three previous tax years can let you contribute more than £60k in a single year — useful if you're catching up.
• The tapered Annual Allowance reduces the £60k by £1 for every £2 of adjusted income above £260,000, down to a floor of £10,000 (HMRC PTM057100). High earners need to compute their adjusted income carefully before contributing.

What to do if your probability of ruin is high

If the headline number is below ~80%, the plan is fragile. The levers that move it, in roughly decreasing order of effectiveness:

1. Reduce the target income. Sustainable withdrawal rates have a steep cliff above ~4% real of the pot. Going from £30k/year to £25k/year on a £600k pot is the difference between fragile and robust.
2. Increase contributions. Each extra £1,000/year for 30 years is roughly £75,000–£100,000 of additional real pot in the median case.
3. Delay retirement by a few years. This is the single most powerful lever, because it both (a) gives the pot more years to grow and (b) reduces the number of years over which you have to draw from it. Two years of delay can shift the probability of success by 10–15 percentage points.
4. Reduce fees. A 1% annual fee compounded over 30 years eats roughly 20–25% of the final pot. The cheapest global tracker is 0.07–0.25%; the difference between that and a 1.5% adviser-overlay is, plausibly, several extra years of retirement income.
5. Adjust equity weight. Higher equity weight raises the median outcome but widens the range — it doesn't unambiguously improve success probability for income targets near the boundary. The optimum equity weight depends on your tolerance for the bottom-decile outcome, not the median.

What this calculator does not do (and why)

• It doesn't model variable withdrawal strategies (Guyton-Klinger guardrails, the Bogleheads variable-percentage-withdrawal rule, dynamic glide-paths). These can materially improve success rates and we plan to add them in a future revision.
• It doesn't model annuitisation. Buying an annuity at retirement is a different problem — it removes longevity risk at the cost of giving up upside.
• It doesn't model the bond yield environment at the start of the simulation. Block-bootstrap inherits the average historical bond experience; in a low-real-yield starting environment the early years of any gilt-heavy plan will likely underperform the historical average. This is a known limitation of all bootstrap methods and a reason why some practitioners prefer regime-conditioned simulation. The trade-off is added model complexity and assumption sensitivity.
• It doesn't model your specific tax position. Pension tax is highly individual. Treat the calculator as a planning tool, not a tax advice engine.

Sources & further reading

• HMRC Pensions Tax Manual — definitive UK pension tax rules.
• gov.uk State Pension forecast
• FCA COBS 13 Annex 2 — the FCA's standardised projection assumptions, useful if you want to compare this to your provider's statement.
• Bank of England Millennium of Macroeconomic Data — UK CPI and inflation series back to 1209.
• Jordà-Schularick-Taylor Macrohistory Database — UK equity and gilt total returns 1870–present.
• UBS Global Investment Returns Yearbook — annual update of the Dimson-Marsh-Staunton long-run returns dataset.

This calculator runs entirely in your browser. We do not store, log or transmit your inputs to any server. See the disclaimer for the usual warnings about why a calculator is not a substitute for regulated financial advice.