Statistical Arbitrage Bets on Relationships Rather Than Direction
Rather than forecasting whether a stock rises, the strategy identifies securities that historically move together and trades the deviations between them.
The Core Idea
Two securities with similar economic exposures should move together. When they diverge without an obvious reason, the divergence may revert.
Statistical arbitrage systematically identifies such relationships and trades the deviations: buying the one that fell relatively and shorting the one that rose, expecting the gap to close.
The position has no view on whether either security goes up. It has a view that the distance between them is temporarily too wide, which is a different and narrower claim.
Pairs Trading, the Simplest Version
The original form takes two similar companies, establishes the historical relationship between their prices, and trades when the current relationship deviates by a defined amount.
The position is long one and short the other in a ratio designed to neutralise market exposure. If both fall together, the position is roughly unaffected. It profits only if the gap narrows.
Modern implementations extend this to hundreds or thousands of securities simultaneously, using factor models to define what should move together rather than relying on identifying obvious pairs.
Why Breadth Is Essential
Each individual position has a small expected edge and substantial uncertainty. The strategy works through the number of independent positions rather than through the accuracy of any one.
This is the same relationship that governs any systematic strategy: the information ratio depends on skill multiplied by the square root of the number of genuinely independent bets.
A statistical arbitrage book therefore holds a very large number of small positions, and concentration in any one of them defeats the entire logic.
The Costs Determine Viability
| Cost | Effect |
|---|---|
| Commissions and fees | Directly reduce a thin edge |
| Bid ask spread | Paid on every entry and exit |
| Market impact | Binds as size grows |
| Borrow cost on shorts | Continuous drag |
Because the edge per trade is small and turnover is high, transaction costs can consume the entire return. Many relationships that appear profitable in a backtest are not tradeable once realistic costs are applied.
This is why the strategy is dominated by firms with excellent execution infrastructure. The edge is partly in the signal and substantially in the ability to trade cheaply.
When It Fails
The strategy assumes relationships revert. Sometimes they break permanently, because one company was genuinely deteriorating and the divergence was information rather than noise.
The strategy will keep adding to a position that is diverging, since a wider gap looks like a better opportunity, which means it accumulates exposure to exactly the relationships that have broken.
The August 2007 quant event demonstrated the systemic version. Multiple funds held similar positions, one began deleveraging, the forced selling moved prices against the same positions everyone else held, which triggered further deleveraging. Losses over a few days were severe, and the positions largely recovered afterwards, which was little comfort to anyone who had been forced to close.
The Decay Problem
Signals stop working. Once a relationship is identified and traded by enough capital, the opportunity is arbitraged away.
This makes the business a continuous research operation rather than a strategy that can be established and maintained. Firms must generate new signals at least as fast as existing ones decay, and the decay rate has increased as more capital and better technology entered the field.
The Bottom Line
Statistical arbitrage trades deviations in historically stable relationships rather than direction, relying on many small independent positions rather than accuracy on any one. Transaction costs determine whether a signal is tradeable at all, crowding makes deleveraging self reinforcing, and signals decay continuously, which makes it a research treadmill rather than a strategy you can hold.