The Lab

How we tested

• Prices: Binance spot daily closing prices, from 2017 to September 2026. Nothing is ever sold, and there is no leverage.
• The same money, four ways, over 1, 2 or 4 years:
  • DCA: equal buys every day, every week or about every month;
  • all at once: everything on day one;
  • half now: half on day one, the other half kept as cash. This is the fair check: on average, DCA has about half its money in the coin. If DCA's calmer ride is only that, half now gets it too;
  • cash: the money kept, worth what was put in.
• Costs: a 0.1% fee on every buy. Cash waiting to be spent earns nothing.
• Many start dates: a new start every Monday, so the result is a spread across hundreds of starts, not one lucky or unlucky path. Starts overlap, so they are not independent tries.
• The coins:
  • Bitcoin, Ether and Solana, over their whole history. These are coins that lasted, so they flatter any plan that buys and holds (survivorship bias, a long-known trap in fund studies).
  • The coins picked at the time: the 20 most traded on the first Monday of 2018, 2020 and 2022, whether they later did well or died. Each was started only in that year.
  • Every coin is reported on its own.
• The bar, set before the run:
  • DCA beats all-at-once if it ends ahead in most starts and its middle result is higher.
  • DCA reduces regret if it loses money less often and its typical worst fall is smaller.
  • DCA is more than keeping cash back if it ends ahead of half now in most starts, and loses no more often.

The maths, step by step

1. Why DCA wins when the price falls. Put in 1,000. The price is 100 on day one and 50 on day two (no fees, to keep it simple).

• All at once: 1,000 ÷ 100 = 10 coins, worth 10 × 50 = 500.
• DCA, half each day: 500 ÷ 100 + 500 ÷ 50 = 5 + 10 = 15 coins, worth 750.
• Half now: 500 ÷ 100 = 5 coins, worth 250, plus 500 in cash = 750.

DCA and half now end the same. DCA's edge here is the cash it hadn't spent yet when the price fell.

2. Why it loses when the price rises. The price is 100, then 200.

• All at once: 10 coins, worth 2,000.
• DCA: 500 ÷ 100 + 500 ÷ 200 = 5 + 2.5 = 7.5 coins, worth 1,500.

3. The "averaging" in the name. Buying the same amount each time buys more coins when the price is low. So DCA's average buying price is never above the average price over the period. But all-at-once bought everything at the first price, and in a rising market that's lower still.

What we found

1. On Bitcoin, Ether and Solana, all at once usually ended higher.

• Weekly DCA's middle result was below all-at-once in 7 of the 9 tests (three coins, three lengths). It was ahead in only 24% to 43% of starts.
• Bitcoin, weekly DCA against all at once, as a multiple of the money put in:

• Buying daily, weekly or monthly made almost no difference.

2. DCA did make the ride calmer.

• Bitcoin over 1 year: the typical worst fall was 29% with DCA, against 52% all at once. The worst of all: 61% against 83%.
• Ether over 1 year: DCA lost money in 39% of starts, against 47%.
• Not always: Solana over 2 years lost money in 44% of DCA starts, against 36% all at once.

3. But that calm is the cash it holds back.

• DCA kept on average 51% to 60% of its money in the coin. Half now kept 52% to 64%.
• DCA beat half now in only 31% to 46% of starts. The one exception was Solana over 4 years (63%).
• So most of DCA's calmer ride is simply having less in the coin.

4. Who wins depends on what the price did. Bitcoin, weekly DCA over 1 year, share of starts where it beat all at once:

DCA wins when the price falls during the buying. All at once wins when it rises.

5. The coins picked at the time.

• 2018: only Bitcoin and Ether were in the list (few coins had been listed long enough).
• 2020, before a bull run: all at once ended higher on all 20 coins over 1 year. Over 2 years, DCA's middle result lost money on 12 of the 20, all at once's on 3: DCA kept buying through the 2021 top.
• 2022, before a bear market: DCA beat all at once on 18 of the 20 over 1 year. Over 4 years, most lost money either way: DCA's middle result was below what was put in on 15 of 20 coins, all at once's on 14.
• On seven coins that kept falling, DCA ended with 18% to 29% of the money, all at once 3% to 9%.
• LUNA, which collapsed in May 2022: all at once lost everything. DCA ended with 81% to 95%, because the buying stopped at the crash and the rest was still cash.
• DCA lost less on a dying coin. It didn't save it.

What it means

• DCA is not an edge. On coins that rose, putting everything in at once ended higher, most of the time. DCA gave up return for a smaller fall, and keeping half the money in cash made almost the same trade.
• Holding earned what the coin earned. Bitcoin rose, so holding it did well. Most of 2022's top-20 coins lost money however they were bought. The result depends on the coin, not the method.
• What DCA really does: it spreads the timing, so no single bad day decides the result. That's worth something to a nervous buyer. But it isn't extra profit.
• No leverage still stands: our earlier tests found leverage made plans worse, not better.

Credits

• Benjamin Graham, The Intelligent Investor, Harper & Brothers, 1949: dollar-cost averaging, described for ordinary investors (the practice is older).
• George M. Constantinides, "A Note on the Suboptimality of Dollar-Cost Averaging as an Investment Policy", Journal of Financial and Quantitative Analysis, 1979.
• Meir Statman, "A Behavioral Framework for Dollar-Cost Averaging", Journal of Portfolio Management, 1995: why people use it anyway.
• Stephen J. Brown, William Goetzmann, Roger G. Ibbotson and Stephen A. Ross, "Survivorship Bias in Performance Studies", Review of Financial Studies, 1992.
• Our result, that lump sum usually wins in a rising market, agrees with that earlier work. The half-now control and the start-picked coin lists are our own design; the lists follow our study "Survivorship: how the coins were picked, and what it does".
• Data: Binance's public spot daily prices.
• PerpGremlin and Claude, Buying steadily, or all at once?, TerminalGremlin, 2026: the tests, the code and the words.