Put 5% of a portfolio in bitcoin. Give every holding a 5% tolerance band. Then count the trades.
Over the ten years to 4 August 2026, that rule fired 516 times. The same 5% band applied to the equity sleeve in the same portfolio fired five times. Not five hundred. Five.
That gap isn't a quirk of bitcoin's returns. It's arithmetic, and it shows up around any small sleeve of a very volatile asset.
Where the hundred-to-one gap comes from
A tolerance band can mean two different things, and the difference matters enormously here.
A relative band is a percentage of the target weight. Five per cent of a 5% target is a quarter of a percentage point either side. The sleeve may sit between 4.75% and 5.25% of the portfolio. That corridor is half a point wide.
An absolute band is a number of percentage points. Five points either side of a 5% target means anywhere from 0% to 10%.
Same words, and corridors nineteen times different in width. Now apply the relative version to the 95% equity sleeve in that portfolio. Five per cent of 95 is 4.75 points, so equities may run from 90.25% to 99.75%. That corridor is 9.5 points wide.
So one rule, written once, hands the fastest-moving asset the narrowest corridor. The band tightens exactly where the volatility rises.
How fast the sleeve actually moves
To put numbers on it, I took daily S&P 500 closes from the St Louis Fed and daily bitcoin closes from Coinbase, and lined them up on the 2,512 trading days from 5 August 2016 to 4 August 2026.
Over that decade the S&P 500 ran at an annualised volatility of 18.1%. Bitcoin ran at 67.2%, or 3.7 times as much. The correlation between their daily returns was 0.27. Fidelity Digital Assets reached a similar conclusion from a different window, putting bitcoin at three to nearly four times as volatile as equity indices between 2020 and 2024.
Now feed that into the corridor. Start a 95/5 portfolio at its exact target and let one day pass. The median day moves the bitcoin sleeve 0.135 percentage points. The corridor allows 0.25 points. So on a typical day the sleeve uses more than half its allowance, and on 26.4% of days a single session is enough to breach it outright.
The tail is worse. On 12 March 2020 one day moved a freshly rebalanced sleeve 1.56 points, more than six times the corridor's half-width.
The trade log follows from that. Monitoring daily and rebalancing on breach, the 5% relative band produced 516 events in ten years — about 52 a year, with a median gap of three trading days between them. The first breach arrived on day 22. In the busiest calendar year it fired 91 times.
Daily monitoring is doing some of that work, so it's worth removing it. Check the same portfolio once a month instead, and the 5% relative band still fires 89 times in ten years. Vanguard's long-run study of a 60/40, monitored monthly on a 5-point threshold, counted 58 rebalancing events across 92 years. Ten years of a small crypto sleeve out-traded nine decades of a balanced portfolio.
What it costs to run
Frequency alone doesn't settle anything. Costs do. Turnover is the honest measure, and here it's the number that stings.
The tight band traded 20.6% of portfolio value a year. For comparison, Harvey and co-authors report that rebalancing a whole 60/40 portfolio back to target every month generates rebalancing trades of 10.2% a year. A sleeve worth a twentieth of the portfolio was generating double the turnover of an entire monthly-rebalanced balanced fund.
Swap the relative band for a 5-point absolute band and turnover falls to 2.7% a year, from five trades rather than 516.
What does that cost? It depends entirely on how you hold the asset. Harvey's team assume 30 basis points for cash equities, citing a Norges Bank estimate. Apply a 30bp round trip to the crypto leg and the tight band gives up about 6 basis points a year. Apply 120 basis points, which is closer to a retail exchange fee paid twice, and the drag rises to 25 basis points a year. The 5-point band costs 3 basis points at the same 120bp assumption.
Tax is the part people underestimate. HMRC treats selling a cryptoasset — or exchanging it for a different cryptoasset — as a disposal for capital gains purposes. The tight band generated 285 sales in ten years. Against an annual exempt amount of £3,000, a rule that manufactures roughly 29 disposals a year is doing something a rule with 0.5 disposals a year isn't. None of that shows up in a turnover figure.
This is the same trade-off that governs annual rebalancing versus 5% threshold bands in an ordinary stock and bond portfolio. Adding a high-volatility sleeve doesn't change the trade-off. It moves you far along it, and it does so without anyone deciding to.
Three ways out
The first is to widen the band on the volatile sleeve alone. A common two-part formulation does this: rebalance on a 5-point absolute move, or a 25% relative move, whichever is smaller. On a 5% target the 25% relative test gives a 3.75% to 6.25% corridor. In the sample that produced 44 events over ten years and 6.2% annual turnover — a twelfth of the trades of the naive 5% relative rule, and a corridor that still stops the sleeve doubling.
The second is to switch from relative to absolute bands for small sleeves. A 2.5-point band, letting the sleeve run from 2.5% to 7.5%, produced 21 events and 5.7% turnover. Vanguard's long-run work is instructive on how little you give up: on data from 1926 to 2018, a 60/40 monitored monthly with a 5-point threshold rebalanced 58 times in 92 years and returned 8.22% a year after tax, against 8.20% for a rule that rebalanced 1,116 times. Not rebalancing at all returned more, at 8.74%, but ran at 14.0% volatility with an average equity weight of 85%.
The third is to rebalance with cash flow. Under the tight band, 231 of the 516 events were purchases of bitcoin rather than sales, and the average trade was just 0.40% of portfolio value. Contributions and dividends can absorb trades that small without a sale ever happening, which removes the tax event and half the spread. It only works on the buying side, and it stops working once the sleeve is the thing that's grown. Contributions could have absorbed a little under half the trades here, leaving 285 sales that still needed funding from somewhere.
None of the three requires giving up the discipline. Each one accepts a wider corridor around the sleeve that moves fastest, on the grounds that the corridor was never the point. Holding roughly the intended weight was.
The counter-argument, which has real force
Here's the strongest objection to all of this. Rebalancing a volatile, weakly correlated sleeve doesn't just control risk. It harvests a genuine return, and the more faithfully you hold the weight constant, the more of it you get.
Scott Willenbrock formalised this in the Financial Analysts Journal in 2011. The diversification return of an asset is approximately half the difference between its variance and its covariance with the portfolio, weighted by its share. Critically, he argues that the source is the rebalancing itself — selling what has risen in relative value and buying what has fallen — not the variance reduction that people usually credit.
Run his formula on the actual figures. A 5% bitcoin sleeve held at a constant weight, at 67.2% volatility and 0.27 correlation, throws off roughly 0.99% a year of diversification return. That is a large number. It's four times the worst cost estimate above, and it lands on the whole portfolio, not just the sleeve.
So the case against wide bands is straightforward: widen them and you stop holding the weight constant, so you capture less of a premium worth about 100 basis points.
What survives that objection
Two things, and they pull in the same direction.
First, Willenbrock's own result says the diversification return depends only on variances and covariances, and "does not depend on the details of the rebalancing". The premium comes from holding a roughly constant weight over time. It doesn't come from the 516th trade. A 25% relative band keeps the sleeve within 1.25 points of target, which is a roughly constant weight by any reasonable standard.
Second, the premium isn't free money. Granger, Greenig, Harvey, Rattray and Zou showed in 2014 that rebalancing is equivalent to a buy-and-hold portfolio plus a short straddle on the relative value of the assets. That induces negative convexity: the strategy magnifies drawdowns precisely when assets diverge hardest. Harvey's later work put a figure on it. Through the 2007-2009 crisis a monthly-rebalanced 60/40 had a maximum drawdown 1.2 times worse than buy-and-hold, five percentage points deeper, at the worst possible moment. The rebalancing premium is compensation for that risk, not an anomaly. If you'd like the equity version of that argument, the evidence on rebalancing through the 2008 and 2020 crashes covers it.
The sample bears the warning out. Over the decade, the tight 5% relative band compounded at 16.67% a year. The 2.5-point absolute band, with 21 trades instead of 516, compounded at 17.26%. Annual calendar rebalancing managed 21.10%. Never rebalancing at all returned 24.46% — and finished with bitcoin at 62% of the portfolio, which is not the portfolio anyone signed up for.
One trending decade proves nothing about the next. But it does show that the extra 495 trades bought no measured return here. They were sold as risk control, and a corridor 1.25 points wide delivers essentially the same risk control.
What would change the conclusion
Four things would.
If bitcoin's volatility keeps falling, the arithmetic softens. It hasn't converged yet — 67.2% against 18.1% over the sample decade — but a sleeve running at 25% volatility would breach a half-point corridor far less often, and the case for a special rule weakens.
If the correlation to equities rises, the diversification return shrinks fast. Willenbrock's formula is driven by the gap between the asset's variance and its covariance with the portfolio. A correlation of 0.7 rather than 0.27 would cut the sleeve's contribution substantially, and with it the reason to tolerate the turnover at all.
If trading and tax costs approach zero, the frequency argument mostly dissolves. Inside a wrapper with no disposals to report and negligible spreads, 516 trades and five trades differ mainly in operational risk. Most private investors aren't in that position for crypto.
And if the sleeve is large rather than small, this whole piece stops applying. Relative and absolute bands converge as the target weight approaches 25%. The problem is specific to small, fast-moving sleeves, in the same way that the drift of a plain 60/40 over one year is a much slower phenomenon than anything described here.
A reasonable reading
Applying one tolerance band uniformly across sleeves of wildly different volatility isn't a policy. It's an accident of arithmetic, and it produces a hundredfold difference in trading between the calmest holding and the fastest.
The evidence points at treating the band width as a function of the sleeve's volatility, not a single number copied across the whole portfolio. Absolute bands do that automatically for small holdings. Relative bands do the opposite. Either way, the thing worth measuring afterwards is turnover and disposals, because those are the costs a band actually controls — and the number the volatility figure alone won't tell you, as the difference between volatility and drawdown makes clear elsewhere.