Key takeaways
- From 31 December 2015 to 2 September 2026, a 1% bitcoin sleeve that was never rebalanced returned 16.10% a year, against 13.28% for a 5% sleeve rebalanced back to target every month.
- The untouched sleeve stopped being a sleeve. It finished at 36.51% of the portfolio and sat through a 43.95% drawdown, against 21.00% for the plain 60/40.
- Annual rebalancing still let a 5% bitcoin sleeve reach 47.52% of the portfolio, on 18 December 2017, because a calendar rule only looks at the weight on calendar dates.
- Move the start to 31 December 2021 and the effect changes sign: 8.63% a year for the 5% sleeve rebalanced annually, against 8.07% for the untouched one.
- A 60/40 of SPY and AGG left alone returned 11.37% a year against 9.86% rebalanced, so 1.51 points of the drift premium had nothing to do with bitcoin.
A 1% bitcoin sleeve left alone beat a 5% sleeve held at 5%
You want to know what a small bitcoin position would have done to an ordinary portfolio, and whether keeping it small was the decision that mattered. It wasn't. The rebalancing rule was.
Over the period from 31 December 2015 to 2 September 2026, a 60/40 portfolio of the SPDR S&P 500 ETF (SPY) and the iShares Core U.S. Aggregate Bond ETF (AGG), rebalanced once a year, returned 9.86% a year. Add a 1% bitcoin sleeve and never touch it again, and the same portfolio returned 16.10%. Start with five times as much bitcoin, at 5%, but rebalance back to 5% every month, and it returned 13.28%.
So the smaller starting position finished 2.82 percentage points a year ahead of the larger one. It managed that by ceasing to be small. By 2 September 2026 the untouched 1% position was 36.51% of the portfolio, and on the way there it sat through a 43.95% drawdown. The plain 60/40's worst fall over the same stretch was 21.00%.
That is the whole result, and both halves of it are awkward. Sleeve size set the risk you started with. The rebalancing rule decided what you were holding a decade later.
How this test was built, and what it leaves out
Three daily price series, all pulled for this piece. Bitcoin is the Coin Metrics community reference rate, which takes a volume-weighted median of trades within one-minute intervals and then a time-weighted average across a 61-minute observation window, fixed at midnight UTC. Equities are SPY, which tracks the S&P 500 at a gross expense ratio of 0.0945%. Bonds are AGG, which tracks the Bloomberg US Aggregate Bond Index at 0.03%. Both are adjusted closes, so dividends are counted.
The sleeve is funded pro rata from both legs. A 5% bitcoin sleeve is 57% SPY, 38% AGG and 5% bitcoin, so the equity-to-bond ratio inside the rest of the portfolio never changes. Four rules are tested: never rebalanced, annual on the last trading day of December, quarterly, and monthly.
What the test excludes matters as much as what it includes. There are no trading costs, no bid-offer spreads, no custody fees and no tax. Every figure is in dollars, so a sterling investor's outcome over the same period would differ by whatever the exchange rate did. And the arithmetic below silently assumes bitcoin survived the whole sample, which it did, but which nobody could have known in 2015.
The outcome table: three bitcoin sleeve sizes, four rebalancing rules
Every cell runs from 31 December 2015 to 2 September 2026. Returns are annualised. The drawdown column is the worst peak-to-trough fall in portfolio value. The last column is how many times the rule actually traded.
| Sleeve | Rule | Return a year | Worst drawdown | Bitcoin weight at the end | Rebalances |
|---|---|---|---|---|---|
| None | Annual | 9.86% | 21.00% | 0% | 10 |
| None | Never | 11.37% | 23.86% | 0% | 0 |
| 1% | Never | 16.10% | 43.95% | 36.51% | 0 |
| 1% | Annual | 11.57% | 21.35% | 0.82% | 10 |
| 1% | Quarterly | 10.84% | 21.21% | 1.30% | 42 |
| 1% | Monthly | 10.51% | 21.74% | 0.99% | 128 |
| 3% | Never | 22.14% | 59.50% | 63.78% | 0 |
| 3% | Annual | 14.67% | 22.50% | 2.47% | 10 |
| 3% | Quarterly | 12.72% | 22.05% | 3.88% | 42 |
| 3% | Monthly | 11.89% | 22.12% | 2.96% | 128 |
| 5% | Never | 26.20% | 65.19% | 74.98% | 0 |
| 5% | Annual | 17.45% | 24.19% | 4.14% | 10 |
| 5% | Quarterly | 14.57% | 23.41% | 6.43% | 42 |
| 5% | Monthly | 13.28% | 23.26% | 4.93% | 128 |
Read across a row and you learn very little. Read down a column and the pattern is hard to miss. More frequent rebalancing produced a lower return in every sleeve size, and the penalty grew with the sleeve. At 1% the gap between never and monthly was 5.59 points a year. At 5% it was 12.92 points.
The drawdown column is where the rebalanced rules earn their keep. Every rebalanced portfolio, at every sleeve size, held its worst fall between 21.21% and 24.19%. That is within 3.19 points of the plain 60/40. The untouched versions ran from 43.95% to 65.19%. Rebalancing didn't buy return here. It bought a portfolio that still resembled the one you designed.
Annual rebalancing still let a 5% bitcoin sleeve reach 47.52% of the portfolio
The end-weight column flatters the calendar rules, because it measures them shortly after a trade. The peak weight tells you what the rule allowed in between.
On 18 December 2017, a 5% sleeve rebalanced each December had grown to 47.52% of the portfolio. The 3% version reached 34.73% and the 1% version 14.80% on the same day. Quarterly rebalancing capped the 5% sleeve at 18.70%, and monthly at 8.99%. The never-rebalanced 5% sleeve peaked at 84.84%.
This is the part that gets lost when a rebalancing rule is described by its average. A calendar rule controls the weight on the calendar dates and nowhere else. Between those dates the position is whatever the market makes it, and with an asset whose annualised daily volatility over this sample was 65.60%, that can be a long way from target. If you'd told yourself in January 2017 that you were running a 5% crypto allocation, by December you were running something else.
Some of the drift premium had nothing to do with bitcoin
Before crediting bitcoin with the whole gap, it's worth checking the control. A 60/40 of SPY and AGG that was never rebalanced returned 11.37% a year over the sample, against 9.86% for the annually rebalanced version. That is 1.51 percentage points from doing nothing, in a portfolio with no crypto in it at all. Equities beat bonds over the decade, and letting the equity weight drift upward captured more of that.
Compare like with like and the bitcoin effect shrinks but survives. Against the untouched 60/40, an untouched 5% bitcoin sleeve added 14.83 points a year. Against the annually rebalanced 60/40, an annually rebalanced 5% sleeve added 7.59 points. The rebalancing decision still roughly halves the contribution. It just isn't quite as dramatic as the raw table suggests.
Start at the end of 2021 and the rebalancing effect changes sign
Everything above depends on bitcoin rising from 429.68 dollars on 31 December 2015 to 77,188.70 on 2 September 2026, a multiple of roughly 179.6. Change the window and the ranking inverts.
| Portfolio | From 31 December 2021 | From 31 December 2018 |
|---|---|---|
| 60/40, annual | 7.38% | 11.23% |
| 1% sleeve, never | 7.91% | 13.78% |
| 1% sleeve, annual | 7.64% | 11.90% |
| 3% sleeve, never | 7.99% | 15.70% |
| 3% sleeve, annual | 8.14% | 13.21% |
| 5% sleeve, never | 8.07% | 17.43% |
| 5% sleeve, annual | 8.63% | 14.49% |
| 5% sleeve, monthly | 8.16% | 13.80% |
From the end of 2021, rebalancing helped. The 5% sleeve rebalanced annually returned 8.63% a year against 8.07% for the untouched version, and the 3% sleeve shows the same ordering at 8.14% against 7.99%. Only the 1% sleeve keeps the original ranking, at 7.91% untouched against 7.64% annual, and the difference there is 0.27 points.
The reason is straightforward. Bitcoin fell 64.2% in 2022, was down 6.3% in 2025 and was down 11.8% again over the first eight months of 2026. A rule that buys back to target after a fall gets more of the asset at lower prices. Over 2015 to 2026 that rule was selling into the biggest rise in the sample. Over 2021 to 2026 it was buying into a flat and falling one.
The 2018 column shows a third pattern. There, quarterly rebalancing beat annual for both the 3% sleeve, at 13.42% against 13.21%, and the 5% sleeve, at 14.82% against 14.49%. Three start dates, three different answers about which rule won. That is what a table this sensitive to its window looks like, and it is the reason our earlier piece on bitcoin allocation at 1% to 5% put backtest sensitivity at the centre rather than the headline number.
What each rule cost to run: 10 trades against 128
The trade counts in the first table are not a rounding detail. Over the same 10.7 years, the annual rule traded 10 times, the quarterly rule 42 times and the monthly rule 128 times. Each of those is a taxable event outside a wrapper, and each pays a spread on an asset that does not trade like a large-cap equity.
Bands sit between the two extremes. A 5% bitcoin sleeve rebalanced only when its weight left a 2.5% to 7.5% band traded 19 times and returned 13.42% a year. Widening the band to 0% to 10% cut that to 6 trades and returned 13.47%. Both landed close to the monthly outcome of 13.28% on a fraction of the turnover, which is the same result the crypto rebalancing work reached from the trade-count side.
Vanguard's 2022 rebalancing paper reached a compatible conclusion for conventional portfolios. It found annual rebalancing optimal across the allocations it tested, and attributed most of annual's advantage over monthly to market exposure rather than to lower transaction costs. The mechanism it describes is the one visible here, amplified by an asset with a much wider return spread against the rest of the portfolio.
The strongest objection: almost nobody sat through this
The case for leaving a bitcoin sleeve alone rests on a survivor's arithmetic. Bitcoin's own worst peak-to-trough fall inside this sample was 83.15%. To collect the 26.20% a year from the untouched 5% sleeve, you'd have had to hold a position that was 84.84% of your portfolio at its peak, through that fall, without selling. Someone capable of that was not running a 5% sleeve. They were running a bitcoin portfolio with an equity sleeve attached.
The institutional view sizes the position from risk rather than from hindsight. BlackRock's December 2024 note puts it plainly:
We believe 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.
The same note observes that a 1% to 2% allocation contributes portfolio risk comparable to a single Magnificent 7 stock in a 60/40, and that larger allocations would skew portfolio risk excessively. Read against the table above, that is not a disagreement about the arithmetic. It is a statement that the risk budget, not the terminal wealth, is what the weight is for. A rule that lets a 5% position become 47.52% has abandoned the budget whether or not the return was good.
Where the UK rules sit around all of this
This is a description of past price data, not a promotion. Since 8 October 2023 cryptoasset marketing to UK consumers has fallen under the financial promotions regime, on the same footing the FCA uses for other high-risk investments. That requires firms to use "specific risk warnings and positive frictions (such as a 24-hour cooling off period)" in their consumer journeys. From 8 October 2025 retail investors have been able to access crypto exchange traded notes, but only where they trade on a UK Recognised Investment Exchange, and there is no Financial Services Compensation Scheme cover behind them. The FCA's ban on retail access to cryptoasset derivatives stayed in place.
None of that changes the arithmetic. It does change what the arithmetic is allowed to be used for, which is understanding the mechanism rather than acting on a backtest.
What this test cannot tell you
One asset, one sample, one currency. The window contains a single 179.6-fold rise that dominates every cell in the table, and there is no second bitcoin history to check it against. Ten and a half years is short for a claim about rebalancing, which is measured in decades elsewhere. The figures ignore tax and dealing costs, both of which fall hardest on the monthly rule, so the gap between the calendar rules is if anything understated here.
A backtest is also not a distribution. It is one path. The 2021 column is not a different opinion about rebalancing; it is the same rule meeting a different sequence of returns, and there is no reason the next decade resembles either one.
What would change the conclusion
The result that a bitcoin sleeve does better untouched depends entirely on bitcoin continuing to out-return the rest of the portfolio by a wide margin. Narrow that margin and the whole table collapses toward the 2021 window, where 8.63% and 8.07% sit 0.56 points apart and the rule barely registers. The last two calendar years already look like that: down 6.3% in 2025, down 11.8% through 2 September 2026.
So the number worth watching isn't the sleeve size. It's how far the position drifts from where you set it, and how long you let it stay there before you notice. LedgerTouch computes that drift on live holdings, which is the only part of this a portfolio tracker can actually answer for you. The rest is a question about what risk you meant to take, and a table of past prices cannot settle that one.