Key takeaways
- Rebalanced monthly from January 2015 to July 2026, a 1% bitcoin sleeve lifted a 60/40's annual return from 8.97% to 9.61%; a 5% sleeve reached 12.17%.
- Maximum drawdown barely moved: -20.05% for the 60/40 against -22.10% with 5% bitcoin, because a 5% position cannot outweigh a 25% equity fall.
- Start-date sensitivity dominates everything: the 5% sleeve added 3.19 points a year from 2015, 0.69 points from 2024, and subtracted 1.68 points from 2025.
- Rebalancing rules matter almost as much; left untouched from 2014, a 1% sleeve grew to 38.8% of the portfolio and lifted volatility to 22.6%.
- BlackRock calls 1-2% reasonable while conceding bitcoin's return potential sits before widespread adoption, which is the sample this backtest measures.
Would you have held a 5% bitcoin allocation in 2015?
Two friends, same money, same 60/40 portfolio. In January 2015 one of them puts 5% into bitcoin and promises to rebalance every month, no matter what happens. The other doesn't bother.
Eleven years on, the first one looks like a genius. Their portfolio compounded at 12.17% a year against 8.97% — three extra points a year, from a sleeve that never took up more than a twentieth of the money.
So that settles it, doesn't it?
Not even slightly. Run the same experiment and change one thing: start in January 2025 rather than 2015. Same asset. Same monthly rebalancing. Same end date. Now the 5% sleeve doesn't add 3.2 points a year — it takes 1.7 away.
The sleeve didn't change. The decade did. And nobody gets to pick which decade they retire into, which is the actual subject of this piece.
The version of this you'll see quoted everywhere
Start with the flattering table, because it's the one that gets screenshotted. Here is what each sleeve size did to a 60/40 between January 2015 and July 2026, rebalanced every month.
| Bitcoin sleeve | Return a year | Volatility | Max drawdown | Sharpe |
|---|---|---|---|---|
| None (60/40) | 8.97% | 9.85% | -20.05% | 0.72 |
| 1% | 9.61% | 10.01% | -20.46% | 0.77 |
| 2% | 10.26% | 10.23% | -20.87% | 0.81 |
| 3% | 10.89% | 10.48% | -21.28% | 0.85 |
| 5% | 12.17% | 11.09% | -22.10% | 0.91 |
Read across the rows and the trade looks lopsided in bitcoin's favour. The 5% sleeve added 3.2 percentage points of annual return. It cost you 1.2 points of annualised volatility, and roughly 2 points of maximum drawdown. Sharpe moved from 0.72 to 0.91.
The drawdown column is the one that catches people out. Ask a friend what a 5% bitcoin holding would do to their worst year, and most will guess it wrecks it. It doesn't come close.
Bitcoin fell 75.6% on month-end prices between December 2017 and January 2019, then another 73.0% between October 2021 and December 2022. Neither collapse did much to a rebalanced sleeve. The worst 60/40 fall in this window bottomed in September 2022 at -20.05%. With 5% bitcoin sitting alongside it, the fall reached -22.10%.
That isn't luck. It's arithmetic you can do on the back of an envelope. A 5% position that halves costs you 2.5% of the portfolio. Equities falling 25% cost you 15%. At these weights bitcoin can't compete with your shares for control of the drawdown, which is a useful reminder that drawdown and standard deviation measure different kinds of risk and don't have to move together.
Matt Hougan and David Lawant reached the same conclusion in a 2021 CFA Institute Research Foundation brief. Their wording: average maximum drawdown "remains largely flat for allocations to bitcoin between 0% and 4% because at this size allocation, bitcoin never competes with the equity allocation to drive the portfolio's maximum drawdown". Above 4%, they found each extra percentage point of bitcoin added about one point of drawdown.
Where these numbers come from
If you want to check any of this yourself, here's exactly what it's built from.
Everything is computed from month-end total-return series: the SPDR S&P 500 ETF for shares, the iShares Core US Aggregate Bond ETF for bonds, and the BTC-USD reference price, all pulled from Yahoo Finance. The window runs from 31 December 2014 to 31 July 2026, which is 139 monthly returns.
The bitcoin sleeve is funded pro-rata from both sides, so a 5% sleeve sits beside 57% shares and 38% bonds. Sharpe ratios use the three-month Treasury bill rate from the St Louis Fed. Unless I say otherwise, the portfolio goes back to target every month, which is the strictest possible reading of "a 1% allocation". The same pro-rata construction is used to test a cash sleeve against a fully invested 60/40 across 92 years of US returns.
Move the start date, and the backtest sensitivity swallows the case
Same 5% sleeve. Same monthly rebalancing. Same July 2026 end date. The only thing that changes is the year you happened to begin.
| Start | 60/40 a year | With 5% bitcoin | Difference |
|---|---|---|---|
| Jan 2015 | 8.97% | 12.17% | +3.19pp |
| Jan 2018 | 9.42% | 10.88% | +1.46pp |
| Jan 2021 | 8.61% | 9.66% | +1.05pp |
| Jan 2024 | 13.37% | 14.06% | +0.69pp |
| Jan 2025 | 12.36% | 10.68% | -1.68pp |
So which of those rows is yours? Whichever year you first had money to invest is doing more work in your answer than any view you hold about bitcoin.
The benefit shrinks by a factor of more than four between a 2015 start and a 2024 start, then turns negative. Nothing about the asset changed. The later windows simply contain less of bitcoin's early repricing and more of its recent one.
One calendar year carries an outsized share of the whole result. Bitcoin returned 1,369% in 2017. Strip 2017's twelve monthly returns out of the series entirely and the full-window gap from a 5% sleeve falls from 6.88 points a year to 2.53 under annual rebalancing. Imagine standing on the first morning of 2017, trying to work out whether that year was coming. Nobody could, on either side of it.
The recent record is bleaker than most allocation notes admit. Bitcoin lost 6.3% in 2025 and a further 28.2% in the first seven months of 2026. At the end of July 2026 it sat at $62,814, which is 45.7% below its month-end peak of $115,758 in July 2025. Across calendar 2025 a monthly rebalanced 5% sleeve turned a 13.56% year for the 60/40 into 12.79%. If the first table sold you on the idea, that is the year you'd have lived through first.
Two people can both claim a 1-5% bitcoin sleeve and mean different things
Picture two more investors. Both buy a 1% sleeve at the end of 2014. One trims it back to 1% on schedule; the other means to get round to it. What happens after the first month is the entire difference between them.
Rebalanced monthly from January 2015, a 1% sleeve added 0.64 percentage points a year. Rebalanced once a year each December, the same 1% sleeve added 1.54 points, because it was allowed to run inside the calendar year and 2017 was a calendar year. Never rebalanced at all, the 1% sleeve added 4.68 points a year over an unrebalanced 60/40 baseline. Running the comparison across four rebalancing rules rather than one shows how far a bitcoin sleeve travels between trades.
That last figure comes with a catch, and it's a big one. A 1% sleeve bought at the end of 2014 and left alone would have been 38.8% of the portfolio by July 2026. A 5% sleeve would have been 76.8%. Whatever the second investor now owns, it isn't a 60/40 with a small satellite.
Would you have sat still for that? The 5% version ran at 41.2% annualised volatility with a worst drawdown of -63.3%. That's a bitcoin fund wearing a 60/40 costume, and it shows why a 60/40's weights drift materially inside a single year even without a bitcoin sleeve in the mix. That headline figure is contested: Fulkerson, Jordan, Riley and Yan, writing in the Financial Analysts Journal in 2026, rebuilt the calculation on the same sample and put the cost of poor timing at 0.10% a year rather than 1.2%.
So anyone reporting bitcoin-sleeve results without naming the rebalancing rule has left out roughly half the answer. The choice between annual rebalancing and 5% threshold bands is normally a second-order decision. With an asset this volatile, it's first-order.
What the people who publish on this actually say
The BlackRock Investment Institute set out its position in December 2024, in a note by Paul Henderson, Vivek Paul, Samara Cohen and Robert Mitchnick. Their conclusion: "a 1-2% allocation to bitcoin is a reasonable range for a multi-asset portfolio if investors believe it will become more widely adopted and can bear the risk of potentially rapid price plunges".
Notice the reasoning is risk budgeting, not return forecasting. A 1-2% weight, they wrote, "contributes to overall portfolio risk at levels comparable to a single 'Magnificent 7' stock in a 60/40 portfolio". Beyond 2%, they argue, portfolio risk rises disproportionately. The same disproportion is what makes employer stock concentration hard to size, with the added problem that the salary keys off the same company.
My own numbers land close to that. On monthly returns over this window, a 1% sleeve accounts for 2.9% of portfolio variance, a 2% sleeve for 6.4%, and a 5% sleeve for 19.6%. Your sleeve's share of risk grows about four times faster than its share of your capital, which is the sort of thing that doesn't show up on a statement.
Academic work points somewhere very different. Yukun Liu and Aleh Tsyvinski, in a 2018 NBER working paper later published in the Review of Financial Studies, ran a Black-Litterman exercise on bitcoin data from January 2011 to May 2018. An investor who believed bitcoin would keep performing as it had should have held 6.1%. Even at half its historical performance, the model said 3.1%. Their wider finding matters more: cryptocurrencies "have no exposure to most common stock market and macroeconomic factors", and their measured correlation with stocks over that sample was 0.16.
So which is it, 1-2% or 6.1%? The gap isn't really a disagreement about bitcoin. It's a disagreement about whether you budget by risk or by expected return, and the two disciplines answer different questions.
The 2021 CFA Institute brief found that a quarterly rebalanced 2.5% sleeve, from January 2014 to September 2020, lifted a 60/40's cumulative return by 23.9 percentage points while volatility barely moved, at 10.5% against 10.3%. Sharpe went from 0.54 to 0.75. Worth knowing who wrote it, though: Hougan is chief investment officer of Bitwise Asset Management and Lawant was a researcher there. The brief is careful and its caveats are honest, but a bitcoin asset manager produced it.
The case against reading any bitcoin allocation backtest forward
This is the strongest objection on the page, and it doesn't come from sceptics. It comes from BlackRock.
The same note that lands on 1-2% says the greatest return potential "lies in the period before widespread adoption, when expectations and narratives may drive repricing". It then adds that widespread adoption "would dull bitcoin's key driver for further sizable price rises" and would make the case for a permanent holding less clear.
Read those two sentences together and the backtest above stops being evidence of a risk premium. Here is the same point as a single holding, which is as concrete as this gets. Say you had bought one coin at the end of 2014, when the price was $320, and then done nothing at all with it. No trading, no rebalancing, no second purchase. In July 2026 that coin is worth $62,814. You have 196 times your money, at a 57.7% annual rate, from something that never paid you a penny along the way.
So where did that money come from? Not from earnings, and not from coupons. It came from whoever bought next. That is what a one-time repricing looks like when an asset moves from fringe to exchange-traded product. A repricing pays once. Bond coupons and equity earnings pay repeatedly, which is why their historical records carry information about the future in a way this one may not.
Vanguard's stated reason for declining to launch its own crypto funds points at the same gap. The firm says it focuses on "products that generate cash flow in a transparent way, such as interest payments and dividends". Bitcoin has no coupon, no earnings and no dividend, so every pound of your return has to come from what the next buyer pays.
The Hougan and Lawant brief concedes the point in its own way: the 2014-2020 result "is notable, but it is also unsurprising: it captures a period during which bitcoin's price appreciated substantially". They also record that bitcoin has had six bear markets of more than 70%, and that with volatility that large "the choice of the starting point can have a dramatic impact".
There's a second problem, and it's the sample. It's short. There are 139 monthly observations of an asset running at about 72% annualised volatility, which puts the standard error on its average annual return at roughly 21 percentage points. The arithmetic average return itself is 69.9% a year. Think about what that pairing means. The uncertainty band is close to a third the size of the estimate, and it swamps whatever weight an optimiser hands back to you. The gold literature carries the same argument, which is why the 5-10% gold case struggles with its own 45-year record despite a far longer sample.
The diversification argument has been weakening
Bitcoin's case in your portfolio rests partly on low correlation with shares. That correlation has been drifting up.
On monthly returns against the S&P 500 ETF, bitcoin's correlation was 0.27 from 2015 through 2020. From 2021 through July 2026 it was 0.48. Over the whole window it's 0.35, well above the 0.16 that Liu and Tsyvinski measured on 2011-2018 data.
BlackRock flagged exactly this risk, describing bitcoin's correlations as unstable and warning that investors "may not be able to rely on it as reliable cushion against risk-off sentiment hitting other parts of the portfolio". A rising correlation and a maturing investor base are the same phenomenon seen from two angles. As bitcoin became something institutions hold in a risk budget, it started behaving like the other things in that risk budget.
The Bank for International Settlements looked at who was actually doing the holding. In a February 2023 bulletin, Giulio Cornelli, Sebastian Doerr, Jon Frost and Leonardo Gambacorta found that almost three-quarters of users downloaded a crypto exchange app when bitcoin was above $20,000, and concluded that retail investors "have chased prices, and most have lost money". Your realised return and the asset's return are not the same number, and reported crypto fund flow data measures something narrower than it appears to.
UK participation is falling as prices fall. The Financial Conduct Authority's 2025 consumer research found that the share of UK adults holding cryptoassets dropped from 12% in 2024 to 8% in 2025, on a nationally representative sample of 2,353 people. Bitcoin was still the most commonly held, at 57% of users.
What would change the conclusion
Three things would move this materially, and you can watch all three yourself.
The first is another repricing on the scale of 2017 or 2020. If bitcoin triples again, every number in the first table shifts up and the start-date sensitivity gets worse, not better. That isn't a forecast. It's a description of how the arithmetic behaves.
The second is correlation. If the monthly correlation with equities settles above roughly 0.6 for a multi-year stretch, the diversification argument stops working and your sleeve becomes leveraged equity beta with worse tails. The move from 0.27 to 0.48 is already halfway there. The rolling correlation behind that move, split by volatility regime, is measured week by week elsewhere on the site. Splitting the same question by market state rather than by start date, crypto diversification measured 0.14 in calm markets and 0.33 once the S&P 500 was already 5% below a record.
The third is volatility. Bitcoin ran at about 72% annualised over this window. Suppose widespread adoption compresses that towards 30%: a 1-5% sleeve then becomes almost invisible in your portfolio, and the risk-budgeting case for capping it at 2% loses its force. You'd be looking at a different asset, needing a different analysis. Sizing ether raises the question in a sharper form, because the ethereum vs bitcoin record pairs almost the same compounded return with materially more volatility.
What wouldn't change the conclusion is another year of backtests. The window is the finding here. Anyone quoting a single number for what bitcoin did to a 60/40 is quoting a start date somebody chose, and the critics of these studies are right to press on that before anything else.