Dollar-cost averaging, backtested

The backtest is real — but the question underneath it is right to ask.

DCA works in volatile assets for a mechanical reason: you buy more units when price is low and fewer when it's high, so your average cost per unit is always below the arithmetic average of prices over the period. This is Jensen's inequality applied to purchasing, not an opinion. The math is not in dispute.

The survivorship bias critique is more serious. Bitcoin is the asset that survived and appreciated dramatically. A DCA backtester built on an asset that went to zero would show 100% loss regardless of cadence. The tool doesn't tell you Bitcoin will keep appreciating — it tells you what disciplined buying into Bitcoin's actual history would have produced.

What the backtester is and isn't:

• It is a precise accounting of historical outcomes under a specific buying rule
• It is not a forecast, and past CAGR numbers are not expected future returns
• The comparison to lump-sum is genuinely informative: in strongly trending assets, lump-sum often wins; in volatile ones, DCA reduces timing risk at the cost of some upside

The most honest use of this tool is stress-testing your assumptions — picking a start date that includes a major drawdown (2021, 2022) and seeing what the numbers look like, not just the favorable periods.

On average, in a strongly trending asset, lump-sum wins because more capital is exposed to the trend earlier. Vanguard's classic study found lump-sum beat DCA roughly two-thirds of the time across equities, bonds and a 60/40 mix.

Bitcoin complicates the answer with the size of its drawdowns. Lump-sum at a cycle top — late 2017 or late 2021 — produced multi-year underwater periods that DCA softened materially. The toggle in the tool lets you check both on the exact window you care about.

Rule of thumb the data supports:

• If your conviction is high and your start date is anywhere except a euphoric top, lump-sum tends to outperform
• If you can't sleep through a 70% paper loss in month one, DCA buys behavioural durability at a small expected-return cost

In practice the difference is much smaller than people expect. Over multi-year horizons, weekly, biweekly and monthly DCA into Bitcoin converge to within ~1–3% of each other on average cost — the dominant variable is how long you bought for, not how finely you sliced it.

Weekly buys smooth the noisiest weeks slightly better and avoid the worst single-month entry. Monthly is operationally easier and pairs naturally with payroll. Pick the cadence you'll actually execute for years, not the one that backtests 0.4% better.

They change the numbers, not the conclusion. The backtester uses weekly closes with no fees or taxes — a clean upper bound on realised return.

Reality drag, rough sizing:

• Exchange fees: 0.1%–1.5% per trade. A 0.5% fee on weekly $100 buys costs roughly the same as one month of buys per year
• Spread on retail venues: typically 0.2%–1% on Bitcoin, larger on smaller venues
• Taxes on eventual sale: jurisdiction-dependent, often the largest single drag

None of these flip a multi-year DCA from a 5× to a loss, but they do compress headline returns. The tool's outputs are best read as the ceiling — your real outcome will be a few percent below.