Why low volatility failed our test.
This is the false positive we’re proudest of catching. Low volatility looked like a clear winner — until a deeper, survivorship-corrected test flipped the verdict completely. It’s the clearest proof that our gauntlet works on its own findings.
A textbook anomaly, and a great-looking backtest
In equities, lower-volatility stocks have historically delivered better risk-adjusted returns than the theory says they should — the “low-volatility anomaly.” Tested on about one year of crypto data, it looked like the anomaly carried over: a top-minus-equal-weight Sharpe around 1.1. On its own, that’s a result most tools would ship and market hard.
A down market, made of survivors
That one-year window had two problems baked in. It was a down market (so low-beta names naturally held up — that’s beta timing, not a cross-sectional edge), and it was measured only on coins that survived to today (so every quiet-but-doomed coin that delisted is invisible). Either bias alone can manufacture a Sharpe of 1.
So we did what the one-year test couldn’t: we re-ran it over a multi-regime history back to 2019, and then we injected the graveyard — delisted and collapsed coins — to see what survivorship had hidden.
The edge didn’t weaken — it inverted
On the deep, multi-regime sample the result flipped sign. Low volatility underperformed equal-weight, and lagged hardest exactly where it should pay off if it were real — in bull markets, about −1.19% per week long-short in risk-on (the one number survivorship can’t inflate). Its earlier “positive IC” turned out to be non-monotonic noise in the middle of the cross-section; the actual portfolio long-short was deeply negative.
The twist: adding the dead coins back briefly made low-vol look amazing again (portfolio Sharpe from −0.33 up to +1.92). But that resurrection was an untradeable artifact — it came entirely from shorting illiquid, dying microcaps down to −100%, with no borrow and no way to exit. A number you cannot actually trade is not an edge.
Two honest-looking tests gave opposite answers. Survivorship and regime flipped the verdict — and the rigour layer caught it before anything shipped.
Dropped — and the lesson kept
Low volatility is not in the IOX score, and won’t be on this evidence. More important than the factor is the lesson it taught us, now baked into every test: a single good backtest proves nothing; survivorship and regime are silent killers; and the portfolio gate — can you actually trade this after costs? — is the final arbiter, not a pretty correlation.
This is why we publish the failures. The factor that survived the same gauntlet — funding crowding — is worth more because low volatility didn’t.
Common questions
Didn’t low volatility look like a strong factor at first?
Yes — on about a year of crypto data it looked great, with a top-minus-equal-weight Sharpe around 1.1. That window was a down market made of survivors, and the result was an artifact of both. It’s exactly why one good-looking backtest is not proof.
What changed the verdict?
Two things. A deeper multi-regime backfill (2019 onward) inverted the edge — low-vol underperformed equal-weight and lagged hard in bull markets (the survivorship-immune number: about −1.19% per week long-short in risk-on). And when we injected delisted/dead coins, its apparent "resurrection" turned out to come entirely from shorting illiquid dying microcaps to −100% — untradeable.
So is low volatility dead for crypto ranking?
As a tradeable cross-sectional factor on our universe, yes. Its positive IC was non-monotonic mid-cross-section noise, and its portfolio long-short was deeply negative. We keep the finding public because catching this false positive is the whole point of the discipline.
Ioxer is research, not investment advice. IOX is a crowding read — not a price prediction, not a buy/sell signal.