Crypto Portfolio Allocation: The 0.84 Correlation That Breaks Most Plans

Most crypto allocation advice is about the split between coins. We think that is the second question, and the data says the first question swamps it. Across the eight largest non-stablecoin assets over the year to 3 August 2026, the average pairwise correlation of daily returns was 0.84, measured over 365 daily observations and 28 pairs.¹ Bitcoin against Ethereum was 0.89. Bitcoin against Solana, 0.86. The loosest pair in the whole set was Bitcoin and BNB at 0.77.

At 0.84 you are not holding eight positions. You are holding one position, expressed eight ways, with extra fees. So the allocation question that actually determines your outcome is not how to divide the crypto sleeve. It is how large the sleeve is relative to everything else you own — because the whole sleeve is going to move together, and it is going to move a long way.

How far it moves

Annualised volatility and the deepest peak-to-trough fall over the same year, per asset:

• Bitcoin — 44% volatility — worst drawdown −53.2%
• BNB — 53% — −58.5%
• Ethereum — 68% — −67.8%
• XRP — 69% — −68.8%
• Solana — 74% — −75.2%
• Chainlink — 77% — −73.5%
• Dogecoin — 80% — −76.2%
• Cardano — 81% — −85.2%

Every asset on that list had a year in which it lost more than half its value from its own high. The mildest of them, Bitcoin, is more volatile than almost anything in a conventional portfolio. This is not a bad year cherry-picked to make a point; it is simply the most recent 365 days.

What follows

Size the position to the drawdown, not to the average. A 5% allocation that falls 75% costs you 3.75% of your total — survivable, and you will still be holding it at the bottom, which is the only way any of this works. A 40% allocation doing the same thing costs 30%, and the practical problem is not the arithmetic, it is that almost nobody holds through that. The allocation you can keep is worth more than the allocation you can justify.

Inside the sleeve, act as though correlation is 1.0 until proven otherwise. Adding a ninth token to a portfolio of eight that move at 0.84 does not reduce risk in any measurable way; it adds a research obligation and a custody surface. If you want genuine diversification, it has to come from outside crypto, and we would rather say that plainly than sell a rotation strategy the correlation matrix does not support.

What this does not show

Correlations are not stable. Ours is a single year, and 0.84 is an average over a period in which the whole market fell together; in a market driven by asset-specific news the figure would be lower, though on this evidence we would not plan around it being much lower. Eight assets is also a narrow view, and adding smaller tokens would probably raise the average rather than lower it, since they tend to track the majors with amplification.

Nothing here is advice about your circumstances, and none of it forecasts returns. Volatility and drawdown describe what already happened; they are not a bound on what can happen next, and every one of these assets can fall further than it has. We will recompute the correlation matrix and the drawdown table on 3 November 2026.