Bitcoin DCA backtest: lump sum vs weekly and monthly buying, every start month

Should money already in hand go into bitcoin all at once, or in installments over a year? We tested both for every start month from January 2012 to November 2025, putting $1,200 to work in one of three ways and valuing the result 12 months later at Bitstamp's BTC/USD daily close, with the last measurement on November 1, 2026. The crypto hub covers the wider risks of holding coins, our guide to dollar-cost averaging in crypto explains recurring purchase plans, and the bitcoin price page shows where the price stands today.

What did the bitcoin DCA backtest find?

A lump sum beat monthly buying in 68.3% of the 167 start months, while dollar-cost averaging trimmed the worst 12-month loss from 72.1% to 55.0%. Both results come from the same trade-off: money invested on day one spends the whole year exposed to bitcoin's price, while installments keep part of it in cash, which helped only when the price fell.

• Lump sum won most often. It beat 12 monthly purchases in 68.3% of start months after 12 months and in 68.5% after 36 months, and it finished first of the three approaches in 67.7% of start months.
• Medians favored the lump sum by a wide margin. The median 12-month return was 79.0% for the lump sum, 41.9% for monthly buying and 40.6% for weekly buying.
• Installments cut the worst case, not the loss rate. The worst 12-month result was −72.1% for the lump sum against −55.0% for monthly and −52.4% for weekly buying, yet each approach lost money in roughly a quarter of start months, from 25.7% to 28.7%.
• Timing near a peak decided most DCA wins. Monthly buying beat the lump sum in every start month of 2014 and in 11 of the 12 start months of 2018.
• The two installment schedules ended close together. Weekly buying finished first among the three approaches in 24.0% of start months and monthly buying in 8.4%.

How was the backtest built?

Each test invests $1,200 starting on the first day of a month, in three ways, and values every portfolio at bitcoin's daily close 12 months after the start.

Fees, trading spreads and taxes are left out, and every purchase fills at one exchange's daily close, so the results describe the pattern rather than any real buyer's fills.

The rule that decides each start month follows from the same arithmetic. Both portfolios are valued at the same closing price, so whichever holds more bitcoin wins. The lump sum holds more whenever the start-day price was below the average cost per coin that the installments ended up paying. A fixed dollar amount buys a larger quantity when the price is down, as the SEC's Investor.gov glossary notes in its definition, so that average cost sits below the simple average of the purchase prices.

How often did a lump sum beat dollar-cost averaging?

Over 12 months, the lump sum finished first among the three approaches in 67.7% of start months and beat monthly buying head to head in 68.3%.

The two figures differ because the summary ranks all three approaches at once. When spreading purchases paid off, weekly buying usually edged out monthly buying, so weekly took 24.0% of first places and monthly only 8.4%. Head to head, monthly buying beat the lump sum in the remaining 31.7% of start months.

Medians show the usual size of the gap: 37.1 percentage points between the lump sum's 79.0% and monthly buying's 41.9%. December 2012, the best start month for all three, shows it at its widest.

The headline share is close to what Vanguard found for stocks. In its 2023 paper, Cost averaging: Invest now or temporarily hold your cash?, investing at once beat a three-month plan 68% of the time over one-year periods in the MSCI World index from 1976 to 2022, with idle cash earning nothing, and longer spreading periods tilted the odds further toward the lump sum. Our guide to lump sum vs dollar-cost averaging covers the stock-market side.

Did dollar-cost averaging reduce the worst outcomes?

Yes, mainly at the extremes: the worst 12-month loss shrank from 72.1% for the lump sum to 55.0% for monthly and 52.4% for weekly buying.

How often money was lost barely changed. The lump sum ended a 12-month window below $1,200 in 28.7% of start months, monthly buying in 26.3% and weekly buying in 25.7%. Installments made bad years less bad far more than they made them rarer.

The worst windows came at different times. The lump sum's worst start was February 2014, two months after bitcoin's record close of December 4, 2013. The installments' worst began in July 2021: they kept buying through the climb to the November 2021 record, and for that start the lump sum lost less (42.7%) than monthly (55.0%) or weekly buying (52.4%). Spreading purchases helps when the price falls early in the buying year and hurts when it rises first and falls later.

Which start years favored dollar-cost averaging?

Monthly buying won mostly in start years close to a market top: all 12 start months of 2014, 11 of 12 in 2018 and 8 of 11 in 2025.

Each of those three years sits next to one of bitcoin's large declines. Our bitcoin drawdowns study dates them: record closes on December 4, 2013, December 16, 2017 and October 6, 2025 were each followed by a fall of more than half. Buying in installments through the following months meant paying lower prices on the way down, while a lump sum had locked in the high price on day one.

The mixed years are more revealing. The lump sum won 8 of 12 start months in 2019, 4 in 2021 and 5 in 2022, years when the price swung both ways and the winner depended on whether the big move came before or after most installments. In 2021 both approaches had negative medians, −37.5% for the lump sum and −33.5% for monthly buying: installments softened a falling market without escaping it.

At the other end, the lump sum won every start month of 2012, 2016, 2020, 2023 and 2024. Median lump-sum returns in those years ran from 67.0% for 2024 starts to 1,207.3% for 2012 starts, and money invested on day one had the longest time to ride the rise.

What happened to start months since 2018?

Since 2018 the lead swung between the two approaches in long runs, and the chart shows how much wider the lump sum's range of outcomes was.

Read from left to right, the chart breaks into five stretches:

• Monthly buying led in every start month from January to November 2018, the year after the December 2017 record.
• The lump sum led in every start month from December 2019 to February 2021, including April 2020, when it returned 782.9% against 361.2% for monthly buying.
• Monthly buying led in every start month from August 2021 to June 2022, as the November 2021 record gave way to a long decline.
• The lump sum then led for 29 straight start months, from September 2022 to January 2025.
• Monthly buying led in every start month from May to November 2025. For an October 2025 start it gained 10.0% while the lump sum lost 28.5%.

Within these 95 start months, the lump sum's result ran from −71.6% (January 2018) to 782.9% (April 2020), and monthly buying's from −55.0% (July 2021) to 361.2% (April 2020). The narrower band is the case for installments; the higher line most of the time is the case for the lump sum.

Does holding for 36 months change the winner?

No: once the last purchase is made, both portfolios hold only bitcoin and move by the same percentage, so the leader at 12 months stays ahead.

That is why the lump sum's head-to-head win rate barely moved, from 68.3% of 167 start months after 12 months to 68.5% of 143 after 36 months. The small difference comes from the shorter list of start months: in every one of the 143, the approach that led at 12 months still led at 36, a property Vanguard's paper also notes.

What does change is the size of the numbers. Median 36-month returns were 381.4% for the lump sum, 281.9% for monthly and 292.0% for weekly buying. Losses became rare: 0.7% of start months for the lump sum and for monthly buying, and 2.8% for weekly buying. The worst 36-month results were −21.3% for a lump sum started in December 2013, and −5.8% for monthly and −6.6% for weekly buying started in February 2021.

Weekly buying's higher loss share comes from four of the five start months from October 2020 to February 2021, when its purchases ran into the 2021 peak. A lump sum started in October 2020 was up 163.5% after 36 months, while weekly buying from the same date was down 4.5%.

What did each start month return after 12 months?

The full table lists all 167 start months, so a reader can look up any month and compare the three approaches side by side.

Neighboring months can differ enormously, which is the clearest sign of how much a single start date matters. A lump sum started on November 1, 2013 gained 60.6% over the next year, while one started a month later, on December 1, 2013, lost 60.6%. Monthly buying from the same two dates lost 34.8% and 30.2%, a far smaller swing between them. That steadiness, rather than a higher average, is what spreading purchases bought.

What are the limits of this bitcoin backtest?

The test uses one exchange's daily closes, a fixed $1,200 budget and a schedule that always starts on the 1st, and it leaves out fees, taxes and interest.

• Overlapping windows. Neighboring start months share 11 of their 12 months of prices, so the 167 results are not 167 independent experiments. The whole sample spans only a handful of bitcoin booms and busts, and the 68.3% figure rests on them.
• Idle cash earns nothing. Money waiting for its turn could have earned interest in the places covered by our guide to where to park cash short term. In stocks, Vanguard found that crediting Treasury bill interest to waiting cash cut the lump sum's win rate from 68% to 65%.
• Costs are left out. A flat fee per purchase weighs most on frequent small buys, while a percentage fee takes the same share from every approach and leaves the ranking unchanged.
• Taxes are left out. Each purchase is its own tax lot with its own cost basis. The IRS's virtual currency FAQ treats a gain as long-term only after more than one year, counting from the day after purchase, so a US sale exactly 12 months after the start would be short-term for every lot. Rules differ by country.
• One price source and time. Bitstamp's close at midnight UTC stands in for every purchase; other exchanges and times of day would give slightly different numbers.
• One coin that survived. Bitcoin is the oldest and largest cryptocurrency, so these results describe its history, not crypto in general; a coin that later collapsed would look very different.

What does the history mean for bitcoin buyers?

It shows a trade-off rather than a rule: a lump sum usually ended higher, and installments usually narrowed the damage when the timing was poor.

Past performance does not predict future results. These numbers come from one asset over one stretch of history, and bitcoin's path from here could look nothing like its path from 2012 to 2026.

Some points hold regardless of what comes next:

• The question only arises for money already in hand. Buying with part of each paycheck is dollar-cost averaging by necessity, and the comparison here does not apply to it.
• The schedule mattered less than the start date. Weekly and monthly buying had median 12-month returns 1.3 points apart (40.6% and 41.9%), while moving the start by a single month could swing the lump sum's result by more than 100 points.
• Installments did not make losses rare. About a quarter of 12-month windows lost money whichever approach was used, so spreading purchases does not remove bitcoin's price risk. Our guide to types of investment risk covers the wider picture.
• Waiting on the sidelines has a cost too. Vanguard's paper also compared cost averaging with simply holding cash for a year, and in its stock data cost averaging came out ahead 69% of the time.

How can you reuse and cite this backtest?

The dataset is free to reuse under the CC BY 4.0 license, including commercially, as long as you credit the source and link back to this page.

The CSV file holds the 167 start months from the full table, one row per month, with returns in percent. A suggested credit line is “Source: Sonic Wallet analysis of Bitstamp BTC/USD daily closes, data through November 1, 2026.” If you trim, chart or recalculate the data, say so, because the license terms ask users to indicate changes. Our editorial policy explains how studies are built, checked and corrected, and when the data is refreshed the dates at the top of this page change with it. For a companion study on whether the weekday of a purchase matters, see crypto returns by day of the week.