How do you size something that can go to zero?
Two questions get conflated constantly. "Will this go up?" is unanswerable. "What happens to me if this goes to zero?" is entirely answerable, today, with arithmetic — and it is the question sizing depends on.
The arithmetic of recovery
Any asset that can lose 100% forces you to size by survivability rather than conviction, because recovery arithmetic is asymmetric and turns brutal quickly.
- A 2% position that goes to zero needs +2.04% from the rest of the portfolio to be made whole.
- A 25% position that goes to zero needs +33.3%.
- A 50% position needs +100%.
And you do not need a total loss to feel this. A −84% drawdown requires a +525% recovery merely to return to the previous level.
That number is not hypothetical. Bitcoin's documented peak-to-trough declines include roughly −93% in 2011, about −86% across 2013–2015, about −84% from the late-2017 peak to December 2018, and about −77% from November 2021 to November 2022. Over the same overall period the asset also produced extraordinary returns. Both facts are true, both are historical, and a sizing decision that acknowledges only one of them is not a decision.
Total losses are ordinary, not exotic
At the individual-token level, going to zero is the base case rather than the tail. A large majority of tokens listed in past cycles no longer trade in any meaningful sense.
The most instructive documented case needed no hack at all. In May 2022 the UST stablecoin — which maintained its peg through a mint-and-burn mechanism against the LUNA token rather than through reserves — lost its peg. The mechanism responded exactly as designed, minting LUNA to absorb the selling; LUNA's supply expanded from roughly 350 million tokens to trillions within days, and tens of billions of dollars of combined market value were effectively erased in about a week.
Nothing was exploited. The design did precisely what it was built to do, under conditions its designers had assumed away.
The claim that has repeatedly failed
The usual argument for a small allocation is diversification: it is uncorrelated, so a little of it improves the portfolio. That argument rests entirely on a correlation, and correlation is regime-dependent — the same lesson as in portfolio risk, with a sharper edge here.
On 12–13 March 2020, as equities fell sharply, bitcoin fell roughly 50% in two days. The day the hedge was needed was the day it behaved like the highest-beta asset in the portfolio, for the mundane reason that holders raising cash sell what they are able to sell. Through 2021 and 2022 the rolling correlation between bitcoin and US technology equities rose to historically elevated levels, and both fell together through 2022. Across other windows the same correlation has sat near zero.
So the honest statement is not "crypto is uncorrelated" and not "crypto is correlated." It is: the correlation is unstable, and it has repeatedly risen toward one during liquidity stress — that is, during precisely the episodes the diversification was supposed to help with. A diversification benefit estimated in calm markets is a calm-market number.
What that leaves you with
Not an allocation. This course does not have one, no course should, and anyone offering you a percentage is selling something. What it leaves you is a way to frame the question:
- Size by what a total loss would do to the plan, not by expected return.
- Stress the position with the diversification benefit set to zero and see whether it still makes sense.
- Decide the rebalancing rule before the position grows. A holding that multiplies becomes a concentration you never chose, arrived at by doing nothing.
- Add this course's other risks to the sizing. Custody, counterparty, contract, governance and regulatory risk sit on top of price risk — and they are correlated with each other, because they all bite hardest in the same conditions.
In the data
The full history of bitcoin, alongside a broad equity index:
Bitcoin's daily record in this data runs back to 13 July 2010, with a high and a low beside every close. Size a drawdown from closes alone and you will understate it, because the extreme sits in the high and the low, a trap for any 24/7 series whose daily boundary is an arbitrary UTC cut rather than an auction. And a provider's list of crypto pairs is today's list: tokens that died were removed, so the universe you size against is a list of survivors.
Try it now
- Switch both charts above to Monthly. On the crypto chart, Measure the largest peak-to-trough fall in each of the last three cycles, then compute the gain each one required to get back — an 80% fall needs 400%, and writing that number down is the exercise.
- Now use the equity chart to date the stress: early 2020, through 2022, and one long calm stretch of your choosing. Measure both charts across each of those three windows and write the pairs down. What you are after is the range of the relationship, not its average — in one window they move together, in another they do not, and the range is the finding.
- Complete this sentence with a number: "if this position went to zero tomorrow, the consequence for my plan would be ___." If you cannot complete it, the position is not sized. It is merely held.