Exploratory · backtest
Crypto testbed
An exploratory, backtest-only testbed — what dual momentum shows on Bitcoin and Ethereum, why their momentum window is shorter, and why none of it is traded automatically.
Crypto — exploratory testbed
The crypto universes are an exploratory testbed, kept separate from the production strategy. They are never traded by the engine: crypto has no automatic execution path, and the deployed portfolio remains the 70/30 described on the How it works page. This section documents what dual momentum does when applied to an extremely volatile asset — Bitcoin, then a Bitcoin/Ethereum rotation — and what a small crypto sleeve adds, in backtest only, to a diversified portfolio.
A shorter momentum window
The setting that matters most for crypto is the momentum window. Where the 70/30 measures momentum over a blend of 1-, 3- and 6-month windows, crypto demands a markedly shorter one — 1, 2 and 3 months. The reason is mechanical, not curve-fitting: an asset that can lose half its value in a few weeks turns trend too fast for a long window to follow; by the time a 6-month signal reacts, most of the fall is already taken. On Bitcoin alone (2016–2026), the short window yields a return-to-drawdown ratio (the gain per unit of worst decline) of about 1.6; the 70/30's long window, applied to the same asset, falls below 0.9 and rides Bitcoin down through its worst crashes.
The Bitcoin/Ethereum rotation
Adding Ethereum as a second risk asset turns absolute momentum (one asset, hold or cash) into a genuine rotation: each month the engine holds the stronger of the two — provided Bitcoin, taken as the crypto regime's barometer, itself clears the cash threshold. Over the window where both exist, the rotation beats Bitcoin alone on return. But that gain is not free: Ethereum, younger and more volatile, deepens the worst drawdown in the 2018 bear market (about −59% against −45% for Bitcoin alone). The extra return comes with extra risk over the full window — a point a shorter window, starting after Ethereum's worst excesses, would have hidden.
The real cost of execution
Every figure on this bench — ratios, returns, drawdowns — is computed on spot price series with no transaction costs. That is a convenient but misleading simplification: in the real world every switch has a cost, and that cost appears nowhere in the curves above.
On the actual execution rail (Alpaca, June 2026), a market order in Bitcoin or Ethereum costs 0.25% per leg (entry tier, under $100k of 30-day volume), charged in kind on the asset received — slightly less crypto on a buy, slightly less cash on a sell. In practice: entering or exiting (cash ↔ crypto) is one leg, about 0.25%; a Bitcoin ↔ Ethereum rotation is two legs, about 0.5%.
And this is where the short window turns against itself: the 1-2-3 setting that makes the signal responsive also makes it jumpy, multiplying switches. The more the signal turns over, the more the friction compounds. The ratio of roughly 1.6 shown above is therefore an optimistic ceiling: the return actually captured is reduced by these fees, all the more so when rotations are frequent — a cost the 70/30, which moves far less often, absorbs much more easily.
Finally, crypto held at a broker is not covered by investor-protection funds (neither SIPC in the US nor a European equivalent): a point developed on the Beware of CFDs page.
Last checked: June 2026.
A crypto sleeve in the portfolio
To measure what a small crypto sleeve would add to a diversified portfolio, we replace 5 points of the 70/30's equity sleeve with the Bitcoin/Ethereum rotation — a 65/5/30 portfolio, built as a probe, never as a target. Over 2016–2026 the return-to-drawdown ratio rises from about 1.6 (the 70/30) to about 1.9, for a near-unchanged worst drawdown: the crypto sleeve decorrelates from the rest — its catastrophic 2018 collapse lands when the equities are near a high, so even at 5% it barely deepens the portfolio's worst. That is the mark of genuine diversification — over this window.
Sizing is the only protection
These results are doubly sample-dependent. First, 2016–2026 is crypto's only window of existence, and it is a historic bull regime (Bitcoin rose roughly two-hundred-fold over it): any dose of crypto mechanically flatters return here. Second, the ratio edge depends entirely on the short momentum window, itself calibrated on that same period — with the 70/30's window, the 65/5/30's edge almost disappears. On top of that comes a risk specific to the class that dual momentum does not cover: a fall to zero is not excluded, and a collapse faster than the monthly step would leave the position exposed. Crypto therefore remains a testbed, to be managed — if it ever truly is — as a separate, capped sleeve whose size is the only real protection. None of this is traded.
Buying crypto where the stocks already live. To manage a small crypto pocket alongside a stock strategy, there is a practical case for buying it in an account that trades both stocks and crypto — such as Alpaca or tastytrade — rather than in a brokerage account paired with a dedicated crypto platform (such as Coinbase). In a single account, you adjust the relative size of the two pockets with a simple trade, without moving money between institutions: no delay, no friction, no round-trips through ordinary cash. One honest caveat: crypto held at a broker is not covered by SIPC (it is entrusted to a third-party custodian — a firm that holds the assets on your behalf) — so the advantage is operational, not a guarantee on custody.