Try it: your start date, your daily amount
Set it to 2017 and it's the test on this page. Move the date or the amount and watch the five ways of buying play out from there.
The short answer
Buying the same amount every day turned $32,170 into $169,704. Four rules did better without ever selling, and every one meant sitting on cash for years.
One rule that sells did far better: buy before each halving, sell after it. It also leans hardest on history repeating.
The rest of this page is the working. Every number is in a table, and the method is at the bottom so you can check it.
Why start at the 2017 top
Because the early years make every plan look like genius. Start the clock in 2011 and you're not testing a strategy, you're testing whether you were early.
Look at the scale. Between 2011 and the 2017 top the price went up about 1,760 times.
$10 a day from September 2011 is $55,150 put in. Today it would be worth about $79.3 million.
Every rule on this page looks brilliant from there. That tells you nothing about which rule to use now.
So I started on 16 December 2017, the highest daily close of that run, at $19,188. It's the worst day to begin, so it's the fairest test.
If a plan only works when you start at the bottom, it isn't a plan. It's luck with a spreadsheet.
It also gives a long, rough ride. The test runs almost nine years, through three big falls:
- −83% from the 2017 top to December 2018.
- −77% from November 2021 to November 2022.
- −53% from October 2025 to June 2026.
And there's one more reason. Some rules need years of past prices before they work, like the 4-year average, so by 2017 every rule has its full history to use.
The rules of the test
Every strategy gets exactly the same money on exactly the same days. The only thing that changes is when it spends it.
- $10 a day, every day from 16 Dec 2017 to 6 Oct 2026. That's 3,217 days and $32,170 each.
- Money waiting for its signal sits in cash. At the end it counts at face value. I tested 4% interest on that cash too, and the order didn't change.
- No hindsight. Every signal uses only prices up to that day: the highest price so far, or an average of the past.
- Counted on 6 October 2026, at $85,322 a bitcoin.
- No fees and no tax. Real apps charge more for lots of small buys, which favours weekly or monthly over daily.
All 23, ranked
Most of the field finished within 2% of plain daily buying. Four rules pulled clear, and the ones that time the halving cycle came last.
None of these ever sells. The halving play that does is further down.
Here's the full table. Two columns matter more than they look: how much actually got into Bitcoin, and how long the money sat waiting.
| How it buys | Worth today | Against every day | Put into Bitcoin | Still in cash | Longest wait |
|---|---|---|---|---|---|
| A year after the last high | $237,988 | +40.2% | $32,170 (100%) | $0 | 946 days |
| Only below the 4-year average | $213,185 | +25.6% | $31,200 (97%) | $970 | 983 days |
| Only below the 200-week average | $210,424 | +24.0% | $31,660 (98%) | $510 | 963 days |
| While 70%+ below the high | $205,386 | +21.0% | $19,110 (59%) | $13,060 | 1,306 days, still waiting |
| Only on a fresh 50% fall | $184,496 | +8.7% | $30,940 (96%) | $1,230 | 1,303 days |
| While 50%+ below the high | $172,790 | +1.8% | $31,320 (97%) | $850 | 955 days |
| Half daily, half while 50%+ below | $171,247 | +0.9% | $31,745 (99%) | $425 | 955 days* |
| While 20%+ below the high | $169,748 | level | $32,170 (100%) | $0 | 208 days |
| Half daily, half while 20%+ below | $169,726 | level | $32,170 (100%) | $0 | 208 days* |
| Every week, $70 | $169,718 | level | $32,130 (100%) | $40 | 6 days |
| Every day, $10 | $169,704 | level | $32,170 (100%) | $0 | never |
| Lean-in (more when cheap) | $169,633 | level | $32,135 (100%) | $35 | 13 days |
| Every month | $168,416 | −0.8% | $32,120 (100%) | $50 | 30 days |
| While 30%+ below the high | $168,377 | −0.8% | $32,170 (100%) | $0 | 649 days |
| Fresh dips: 10% off the 30-day high | $164,540 | −3.0% | $31,200 (97%) | $970 | 138 days |
| Only below the 200-day average | $161,859 | −4.6% | $31,680 (98%) | $490 | 385 days |
| Half daily, half for a fresh 20% dip | $160,062 | −5.7% | $31,890 (99%) | $280 | 267 days* |
| Fresh dips: 20% off the 90-day high | $150,421 | −11.4% | $31,610 (98%) | $560 | 267 days |
| Only on a fresh 30% fall | $142,532 | −16.0% | $29,740 (92%) | $2,430 | 1,106 days |
| Only on a fresh 20% fall | $140,242 | −17.4% | $30,920 (96%) | $1,250 | 314 days |
| Bear years only, pile spread out | $130,702 | −23.0% | $28,348 (88%) | $3,822 | 1,096 days |
| Bear years only, pile on 1 Jan | $92,829 | −45.3% | $32,170 (100%) | $0 | 1,096 days |
| 550 to 915 days after a halving | $85,825 | −49.4% | $32,170 (100%) | $0 | 1,075 days |
"Longest wait" is the longest run of days with no buy. *For the half-and-half rows it's the saved half, because the daily half never waits.
"A year after the last high" shows $0 in cash for a reason you'd never guess. The last high was 6 October 2025, so its year was up on 6 October 2026, the day I counted, and it spent its whole $9,470 pile that morning.
It's a level, not a fresh fall. The rule buys on any day the price sits at least 20% under its highest close so far.
Bitcoin has been more than 20% under its October 2025 high every day since 13 November 2025, so that rule has bought every day since, and holds no cash. The "fresh fall" and "fresh dips" rows are the ones that wait for a new fall.
All $32,170 on 16 December 2017, the day this test starts and the 2017 top at $19,188, would be worth $143,049 today. That's less than buying every day, because every penny went in at the peak.
It's not a fair fight, though. Nobody has nine years of savings on day one.
What it looked like along the way
The final number hides how different these felt to live through. The chart at the top shows five of them week by week, from the 2017 top by default.
This is what the waiting looks like. Each sawtooth is a pile of cash growing until the signal fires.
The green line is still climbing. The 70% drop rule last bought in March 2023, and it's sitting on $13,060 in cash, 41% of everything it was given.
So far, this cycle's worst fall is 53%. Its signal hasn't come.
Dynamic DCA: $5 a day, $5 saved for a dip
It finished at $169,726, against $169,704 for $10 every day. That's a tie.
The reason surprised me. Since the 2017 top, Bitcoin has been at least 20% below its high on three days out of four (2,407 of 3,217).
So the saved $5 hardly ever waits. Its longest wait was 208 days, and it was buying on the day I counted.
Measure the dip from the last 90 days instead and it does worse. It finished −5.7%, because it's half of the fresh-dip strategy, and that one finished near the bottom.
Why buying the dips loses
Because in a rising market the dip often doesn't take the price back to where you started saving. You wait, the price climbs, and the dip lands higher than the day you started.
Waiting for 20% off the 90-day high finished −11.4%. Waiting for 10% off the 30-day high came out behind daily buying on every one of the 118 start dates I tried.
So I tested the purest version too: save, and buy only on the day the price has fallen 20% from its high since you last bought. Then the high resets and you wait for the next fall.
It finished −17.4%, from 43 separate buys. A fresh 30% fall did −16.0%, and only a fresh 50% fall came out ahead, at +8.7%.
A short dip is common, so the cash never builds into much. It still misses the climb before every dip.
The crash-waiters that won, and the catch
Buying only below the 4-year average finished $213,185, or +25.6%. The 200-week average and the 70% drop rule were close behind, and the best rule of all gets its own section below.
All of them wait for a real crash. That's the good news, and it's also the catch.
The 4-year average rule bought nothing for the first 817 days. Then almost its whole result came from three days:
| Day | Cash it had saved | Price that day |
|---|---|---|
| 12 Mar 2020, the COVID crash | $8,180 | $4,842 |
| 16 Jun 2022 | $8,220 | $20,383 |
| 25 Jun 2026 | $9,840 | $59,696 |
Think about what that asks of a real person. Two years of buying nothing, then spending eight thousand dollars on the scariest day of the decade.
And it's fragile. Move the line a little and the winner becomes a loser:
Wait until the price is 10% under the 4-year average and it finishes −41.7%. The COVID crash only went 8% under the line and 2026 only 4%, so it misses both.
It buys once, in June 2022, and ends with $13,050 still in cash.
With hindsight, anyone can draw the line in the right place. The rule only worked because the crashes happened to reach it.
Does it hold if you start another month?
Mostly, yes. I re-ran every strategy from the 1st of every month, January 2015 to October 2024, all counted on the same day.
Watch the worst-start column.
| How it buys | Beat every day | Typical result | Worst start | Best start |
|---|---|---|---|---|
| Only below the 4-year average | 77% | +16.2% | −71.3% | +37.0% |
| Only below the 200-week average | 77% | +13.8% | −69.2% | +36.5% |
| A year after the last high | 76% | +11.4% | −7.1% | +44.3% |
| While 50%+ below the high | 72% | +1.2% | −54.3% | +34.9% |
| Half daily, half while 50%+ below | 72% | +0.6% | −27.1% | +17.5% |
| While 20%+ below the high | 41% | 0.0% | −5.6% | +4.4% |
| Half daily, half while 20%+ below | 41% | 0.0% | −2.8% | +2.2% |
| Every week, $70 | 7% | −0.3% | −1.3% | +0.3% |
| Lean-in (more when cheap) | 2% | −0.5% | −2.0% | +0.1% |
| Every month | 7% | −2.0% | −6.3% | +0.8% |
| While 30%+ below the high | 7% | −2.0% | −12.6% | +6.0% |
| Half daily, half for a fresh 20% dip | 10% | −3.3% | −9.3% | +2.5% |
| Fresh dips: 10% off the 30-day high | 0% | −3.6% | −15.3% | −1.8% |
| While 70%+ below the high | 42% | −4.1% | −67.1% | +22.0% |
| Only on a fresh 50% fall | 40% | −4.2% | −86.3% | +32.9% |
| Fresh dips: 20% off the 90-day high | 10% | −6.6% | −18.7% | +5.1% |
| Only below the 200-day average | 14% | −7.2% | −75.3% | +4.4% |
| Only on a fresh 20% fall | 10% | −11.7% | −25.5% | +1.8% |
| Only on a fresh 30% fall | 2% | −20.0% | −46.6% | +2.8% |
| Bear years only, pile spread out | 15% | −25.2% | −87.3% | +12.3% |
| Bear years only, pile on 1 Jan | 2% | −45.2% | −92.3% | +4.3% |
| 550 to 915 days after a halving | 0% | −50.4% | −91.1% | −5.1% |
The 4-year average still won more often than not. But its worst start, September 2015, finished −71%, because the signal barely fired and the cash missed the whole 2016 to 2017 run.
Weekly, monthly and the half-and-half rules hardly moved, wherever they started. That's the point of them.
Timing the four-year cycle
Bitcoin's history has a rhythm. The supply cut called the halving comes every four years, and so far a peak has followed it, then a crash.
So people try to time it. I tested the four ways I hear most, using only what you'd know on the day.
| The rule | From the 2017 top | What happened |
|---|---|---|
| Buy once a year has passed with no new high | +40.2% | The best result on this page. |
| Buy only in the bear years (2018, 2022, 2026), saving the rest of the time | −23.0% to −45.3% | Spread over the year, or spent on 1 January. |
| Buy 550 to 915 days after a halving, the year after past peaks | −49.4% | The worst result on this page. |
| Buy 500 days before a halving, sell 500 days after | +255% | The best result on this page, and it sells everything twice. |
A year after the high: the best rule, with a big if
It saves while the price keeps setting new highs. Once a full year passes without one, it spends the pile and keeps buying until the next high.
It fired on 16 Dec 2018 at $3,194 and on 8 Nov 2022 at $18,550, within weeks of both bottoms. And it isn't one lucky number:
Every wait from about 180 to 495 days beat buying every day. On other start dates it won 76% of the time, and its worst start was only −7.1%.
It works because the last three lows came 406, 364 and 378 days after their highs. That's three data points.
So far, this cycle's low came after 267 days, in June 2026, and the rule missed it. It only started buying on the day I counted, so if the price falls further from here it will be buying the whole way down.
Buy before the halving, sell after it: the big winner, with the biggest if
This is the cycle play most people mean. Buy in the 500 days before each halving, then sell everything 500 days after it and wait in cash for the next one.
It turned the same $32,170 into $603,011, which is +255% against buying every day. Even after UK capital gains tax on both sales it's +141.6%.
| $10 a day from the 2017 top | Worth today | Against every day |
|---|---|---|
| Buy every day | $169,704 | 0.0% |
| Buy only in the 500 days before a halving, never sell | $218,475 | +28.7% |
| Buy 500 days before, sell everything 500 days after | $603,011 | +255.3% |
| The same, after UK capital gains tax on each sale | $410,024 | +141.6% |
| Buy 500 before, sell 500 after | Day | Amount | Price |
|---|---|---|---|
| Bought | 28 Dec 2018 | $3,780 | $3,885 |
| Sold | 23 Sep 2021 | $77,703 | $44,904 |
| Bought | 7 Dec 2022 | $87,103 | $16,836 |
| Sold | 2 Sep 2025 | $594,021 | $111,238 |
And it isn't one lucky setting. 118 of 143 combinations of buy and sell days beat buying every day, and it won from 92% of the other start dates.
The money comes from the selling. Buying only before the halving and never selling did +28.7%, and only 7 of 13 versions of that won.
It rests on two finished cycles in this test, and the rule was drawn from that same history. The peaks came 525 to 549 days after each halving.
It means selling all your Bitcoin twice, and it has been in cash since September 2025. Its next buy window opens around 4 December 2026, 500 days before the next halving's expected date, so its lead only holds if the next cycle repeats.
Bear years and buying after the halving: both lose
Buying only in the bear years loses because it spends too early. In 2018 and 2022 the lows came in December and November, and it also sits out cheap recovery years like 2019 and 2023.
Buying only in a window after the halving is worse. Only 8 of 57 window settings beat buying every day, and the typical one finished −26.5%.
The popular version starts buying around day 550, which landed right on the 2021 and 2025 peaks. The saved pile went in at about $64,000 and $108,000.
Has buying every month ever lost over four years?
Not once. I tested £100 on the 1st of every month, held four years, for every start month since 2011: 134 of 134 ended up.
Over one year it's a different story. 43 of 170 one-year windows lost money, about one in four.
In the short run it loses often. The four-year record is what has happened so far, and the next four years haven't happened yet.
Got a lump sum? All at once, or over six months
It's close to a coin flip. Since 2014, putting it all in on the day beat six monthly buys 59% of the time, a year later.
From the 2017 top it was 54%.
Stocks are different. Vanguard found a lump sum won 68% of the time there, but Bitcoin moves so much in six months that spreading out is worth more.
Your timing does matter a little. After a six-month rise the lump sum won 65% of the time, and after a six-month fall it was a coin flip at 49%.
Near the high it's flipped lately. From the 2017 top, a lump sum within 10% of the all-time high won only 36% of the time.
| $32,170 on | Price | All in that day | Spread over 6 months |
|---|---|---|---|
| 2017 peak, 16 Dec 2017 | $19,188 | $143,049 | $269,792 |
| 2018 bottom, 15 Dec 2018 | $3,180 | $863,269 | $659,844 |
| 2021 peak, 8 Nov 2021 | $67,559 | $40,628 | $59,952 |
| 2022 bottom, 21 Nov 2022 | $15,766 | $174,096 | $127,636 |
| 2025 peak, 6 Oct 2025 | $124,728 | $22,006 | $30,691 |
| 2026 low so far, 30 Jun 2026 | $58,526 | $46,899 | not finished yet |
Spreading out wins at every peak and loses at every bottom. Nobody knows which one they're standing on.
A lump sum usually comes out a little ahead. Spreading it over six months costs a few percent on a typical day and protects you from the day you'd regret most.
Three months is a middle way. It was more than 20% behind on only 6.5% of start days.
What the numbers say
Here's what the numbers did, from the worst place to start. None of it is advice.
| Question | What the test found |
|---|---|
| Best on paper | Buy 500 days before each halving, sell 500 days after. +255% (+141.6% after tax), but it sells everything twice and rests on two cycles. |
| Best without selling | Buy once a year has passed with no new high. +40.2%, and only −7.1% at its worst start, but it rests on three past lows. |
| Next best | Buy only below the 4-year or 200-week average. About a quarter ahead, after years of waiting, and fragile. |
| Easiest to stick to | Every day, week or month. All within about 1% of each other, and never stuck in cash. |
| Worst | The halving clock and bear-year buying, then the short dips. All behind from the top and on most other starts. |
| Dynamic DCA, 20% | A tie with daily, because Bitcoin is 20% under its high most of the time. |
| Got a lump sum | All at once wins a little more often than not. Six months of buys protects you from landing on a top. |
| A middle way | Half every day, half saved for a 50% drop. +0.9% from the top, worst start −27%. |
The plan that wins in real life is the one you're still doing in year three. For most people that's the boring one.
Then make sure it ends up with you
Four years of buying is a big pile, and most people leave it on the exchange they bought it on. Here's how to hold it yourself without the tech headache.
Go deeper
- The study everyone quotes: Vanguard, Cost averaging: invest now or temporarily hold your cash? (2023). A lump sum beat spreading it out 68% of the time. Vanguard also says that's a different question from investing from each pay cheque, which is the one most of us face.
- Watch it explained: The Investing Strategy So Simple It Feels Like Cheating.
- For the next crash: If This Found You, You'll End Up Rich, and Is Bitcoin dead? I've thought so four times now.
- Every four-year hold since 2017: the four-year window, priced lump sum by lump sum.
Method and sources
- Prices: Bitstamp's daily close for BTC/USD, from its first day (18 Aug 2011), via its public API. Pounds use the European Central Bank's reference rate. Cross-checked against Yahoo Finance's BTC-GBP series.
- Signals: "Below the all-time high" uses the highest daily close up to that day. The averages are simple averages of past daily closes: 200 days, 1,400 days (200 weeks) and 1,461 days (4 years).
- Lean-in: spends a share of its cash each day that grows as the price falls below its 200-day average, and nothing when the price is more than 2.4 times that average.
- Cycle rules: halvings on 28 Nov 2012, 9 Jul 2016, 11 May 2020 and 20 Apr 2024. A bear year is the second calendar year after a halving. "A year after the high" counts days since the highest daily close so far.
- Other start dates: the 1st of every month from Jan 2015 to Oct 2024, 118 starts, all counted on 6 Oct 2026.
- Four-year test: £100 on the 1st of each month, 48 buys, valued on the day the 49th would have been. The same test on every buy day from the 1st to the 28th: 3,730 windows, none lost.
- Not included: fees, tax, and the spread an app charges. Results are counted on one day and will move with the price.
FAQ
What is Bitcoin DCA?
Dollar-cost averaging. You buy the same amount on a schedule, every day, week or month, whatever the price is doing. In the UK it's sometimes called pound-cost averaging. The point is that you never have to guess the right moment, so a crash just means the same money buys more.
Is it better to DCA Bitcoin or buy the dip?
In this test, buying short dips did worse than buying on a schedule, from the 2017 top and from almost every other start. Waiting for deep crashes did better from 2017, but only after years in cash, and small changes to the rule turned it into a loss.
Is daily, weekly or monthly DCA best?
From the 2017 top they finished within about 1% of each other, and the same was true from almost every other start. So the schedule barely matters. Fees usually decide it: most apps charge more on lots of small buys, so weekly or monthly tends to cost less than daily.
What is dynamic DCA?
Buying a set amount on a schedule and keeping some back to spend when the price drops. With half saved for a 20% drop from the high, it tied with plain daily buying, because Bitcoin spends most of its time at least 20% under its high.
Should I put a lump sum into Bitcoin all at once?
Since 2014, all at once beat six monthly buys 59% of the time, measured a year later, which is close to a coin flip. It was more than 20% behind on about one start day in six, and 49% behind if you went in on 6 January 2018. Spreading over three to six months gives up a little on average and takes the worst days off the table.
Does timing the halving cycle work?
It depends which way round. Buying 500 days before each halving and selling 500 days after did the best of everything here, +255%, because it sold near both peaks. Buying only in a window after the halving did the worst, and both rest on very few cycles.
Why start the test at the 2017 peak?
To leave out the early years, when the price rose about 1,760 times and every plan looks brilliant. $10 a day from 2011 would be worth about $79.3 million. Starting at a top is the hardest test, and by 2017 every rule had years of past prices to work from.