Statistical arbitrage — stat arb — trades temporary deviations between assets that historically move together, betting the relationship will reassert. Long the one that’s cheap relative to the relationship, short the one that’s rich, wait for convergence.
The “statistical” part is the point: the relationship is probabilistic, not guaranteed, and the strategy earns from being right on average across many positions.
How it works
Identify assets with a stable historical relationship. Model the spread between them. When the spread widens beyond its normal range, position for reversion. Close when it normalises.
Run across dozens or hundreds of pairs simultaneously, the individual bets are noisy but the portfolio is smooth, and exposure to the overall market direction nets out to roughly zero.
Why it’s market-neutral
Because every long is paired with a short, the portfolio earns from relative movement, not from the market going up or down. A crash that hits both legs equally leaves the spread unchanged.
That’s the theory. In practice, correlations that held for years break in crises, and “market-neutral” strategies have lost heavily in exactly the moments they were meant to protect.
In crypto
Correlations between tokens are high and unstable, which makes stat arb both tempting and dangerous. Spreads between related tokens, between a perp and its spot, or between the same asset on different venues are all candidates. Execution cost is the main constraint: a spread of 0.3% is worthless if round-trip costs are 0.4%.