What happens after you buy moves the result

Buying monthly, rebalancing, selling in a panic, gearing up and the timing of flows each changed the outcome by more than people expect. Investors in Australian crypto funds earned about 24 points a year less than their funds did.

01 · All at once, or a bit each month?

Monthly buying traded some return for smaller dips

Spreading a purchase out, dollar-cost averaging, is the classic way to take the edge off the fear of buying at a peak. The test is deliberately simple. For every start month from 2016, $12,000 into Bitcoin at once against $1,000 a month for a year, each valued three years after the first purchase.

All at once beat monthly buying in 69% of the start months for Bitcoin and 83% for global shares, because both mostly rose, and money waiting in cash falls behind. What monthly buying did change was the ride: the typical worst paper loss was 14% of the $12,000 against 20% for a lump sum, for a typical final value of 3.25x against 3.50x.

No start month since 2016 lost money over three years for either method, which says as much about one asset's run, from a few hundred dollars to six figures, as about the method. In the worst windows the lump sum finished at 1.21x and monthly buying at 1.08x, so monthly buying did not rescue the worst starts here.

02 · After the purchase

Rebalancing, drift and panic selling

Buying is one decision. The others come later: what to do when the slice grows, and what to do when it falls. Four simple rules, applied to a 5% Bitcoin slice in the balanced portfolio. Prices are from EODHD, in Australian dollars. The balanced portfolio is 35% VAS (Australian shares), 30% VGS (global shares) and 35% VAF (bonds), rebalanced yearly.

Never rebalancing looks brilliant from 2016, 26% a year against 15%, but it let Bitcoin grow to 88% of the portfolio, and the worst fall became −62% instead of −20%. It worked because Bitcoin kept winning; from 2018 the same decision cost only a little because Bitcoin did not run away. The panic rule cost between 1.5 and 1.8 points a year in the two windows: it locked in losses and missed the rebounds. Rebalancing on a threshold gave up some of the gains, but never let the slice grow past 11% of the portfolio.

03 · The range of outcomes

One history is one path

To see a spread, the daily returns from 2018 were resampled in 21-day blocks to build 4,000 alternative five-year futures, and each portfolio was held through all of them. The catch is that the resampled futures inherit Bitcoin's past return, which was high. So the test was run three ways: as in history, with Bitcoin's average return halved, and with it set to zero.

The no-Bitcoin portfolio's median was $14,562 (middle 80%: $11,284 to $18,290). A 5% slice finished behind it in 18% of histories if Bitcoin repeats its past, 40% if its return is half as good, and 65% if it earns nothing. A 20% slice had a one-in-four chance (26%) of a fall bigger than 30% even on Bitcoin's real history, and 39% if it earns nothing. The answer depends on one assumption, what Bitcoin returns from here, and the data cannot say what that is.

04 · What a 2x fund really does

The cost of leverage was in the falls, and on Bitcoin, everything

A daily-reset leveraged fund gives a multiple of each day's return, less the cost of borrowing the extra and a 0.6% fee. These simulations run from January 2018 on assets we hold, with borrowing at a constant 4%. Each extra 2 points of borrowing cost took about 2.4 points a year off the 2x result on global shares, about as much as the daily resetting itself.

On global shares, 2x made 21.0% a year against 13.6% for 1x, but the worst fall went from −23% to −42%. On Bitcoin, 2x lost 4.4% a year even though Bitcoin itself rose 23.2%, and fell 98% at its worst. At 3x, one bad day wipes it out, because Bitcoin has had days of −33% or worse. The drag from resetting daily is about the asset's variance: small for shares, very large for Bitcoin.

As a check against a real fund, GEAR (Betashares Geared Australian Equity) returned 10.2% a year from April 2014 to September 2026 against 7.9% for VAS, with volatility of 33% against 14% and a worst fall of −66%. Its daily returns correlated 0.97 with a simulated 2x VAS.

05 · What investors actually earned

Money followed the price, and the average dollar did far worse than the fund

A fund's published return assumes you held it from the start. Real investors add and withdraw money, usually after seeing prices move. The difference between a fund's return and the return on the average dollar actually invested is the investor return gap. Using the ASX monthly reports from January 2020 to August 2026, each fund's return, net flow and size, we measured it two ways: rebuilding the balance from flows and returns, and using reported fund sizes.

For crypto the two methods agree within a point (the rebuilt balance matches the reported fund size at 0.99), and the gap is about 24 points a year on both. For ETFs overall the sign of the gap depends on the method (+1.3 points on one, −2.0 on the other), because reported flows and fund sizes do not reconcile, so the data cannot tell a timing penalty from none. What is robust for the broad market is the flow behaviour: flows did not follow the previous month's return for ETFs overall, and for crypto they clearly did.

What this does not show

• One asset’s run. Every monthly-buying window ended in profit after three years, which is a feature of an asset that went from a few hundred dollars to six figures. A fair test of the insurance idea needs a market that fell for years.
• The simulations reshuffle the past. Resampling cannot create a new regime. The 2018 to 2026 sample has one boom, one collapse and a recovery, and Bitcoin’s average return in it is high. The three return assumptions are a sensitivity range, not a forecast.
• Leverage is simulated. Borrowing is a constant 2%, 4% or 6%, and real funds differ in tracking, spreads and how they borrow. Some, such as GEAR and GHHF, borrow and rebalance to a loan-to-value range instead of resetting daily. Check each fund’s product disclosure statement.
• The investor gap is small-sample. The crypto result rests on five funds over 26 months and a few large flow months, and the money-weighted method is sensitive when assets grow from almost nothing. For ASX ETFs as a whole the data cannot settle the sign of the gap: funds close and drop out, and reported figures contain some errors at source.
• Behaviour is not modelled. The panic rule is one mechanical rule. How any one person behaves through a 40 to 60% fall is not in the data.