Why momentum failed our test.
Momentum — winners keep winning — is the most documented factor in all of finance. We tested it on crypto and it didn’t hold. Here is what the data said, and the caveat we’re honest about.
Why momentum looked promising
Across stocks, bonds, commodities and currencies, momentum is the closest thing finance has to a law: assets that recently outperformed their peers tend to keep outperforming for a while. It would be strange if crypto were the one exception — so a cross-sectional momentum factor (rank coins by recent return relative to the field) was an obvious candidate for the IOX score.
How we tested it
We measured cross-sectional momentum across the liquid universe on non-overlapping windows (overlapping windows inflate significance), and — the decisive step — we split the result by market regime, checking whether the factor kept its direction in both calm/bull and volatile/bear conditions. Sign-stability across regimes is exactly the test that separated funding crowding from the pretenders.
Insignificant — and unstable
Two problems, either of which is enough to drop it:
- Not significant. The cross-sectional momentum signal came in at |t| < 1.1 on our universe — indistinguishable from noise.
- Sign-unstable. Worse than weak: its direction flipped between bull and bear regimes. In one regime higher-momentum coins led; in the other they lagged.
A factor whose direction reverses with the market isn’t an edge — it’s a coin flip that needs you to call the regime first. Funding crowding kept the same sign in both regimes; momentum didn’t.
Dropped — but honestly, “unproven”
Momentum does not enter the IOX score. But we won’t overclaim that it’s dead: our test ran on the coins that survived to today, and survivorship bias can quietly bury a factor that would show up on a clean, point-in-time universe (low volatility is the cautionary tale — its verdict flipped when we changed the sample).
What would make us reconsider
- A retest on a survivorship-free, point-in-time universe where momentum is sign-stable across regimes.
- And, as always, marginal information over funding crowding — not just a standalone tilt.
Common questions
Does momentum work in crypto?
On our universe, cross-sectional momentum (ranking coins by recent relative return) was statistically insignificant (|t| < 1.1) — and critically its sign flipped between bull and bear regimes. A factor that points one way in rallies and the other way in selloffs is not a stable edge.
Why does a flipping sign disqualify it?
A real cross-sectional factor should keep the same direction across regimes, the way funding crowding did. If the direction reverses with the market, you can’t trust it out-of-sample — you’d have to know the regime in advance, which is its own hard problem.
Is momentum definitively dead for you?
No — we call it unproven, not dead. Our test ran on the coins that survived to today, and survivorship can bury a real factor. A proper point-in-time-universe retest stays on the list. We just won’t ship it on the evidence we have.
Ioxer is research, not investment advice. IOX is a crowding read — not a price prediction, not a buy/sell signal.