The spread is the strategy: execution in thin markets

The Quantitative Strategies division runs a small book by the standards of the markets it trades in, and that is a deliberate constraint rather than a stage of growth. A small book can hold positions a large one cannot enter without moving the price against itself. The cost of that advantage is that the instruments available to us are thin: wide quoted spreads, shallow depth at the touch and long intervals between prints. A research process ignoring those three properties produces results that cannot be realised.

We separate a strategy's gross signal from its net outcome at the earliest possible stage. Every candidate is evaluated against a cost model built from our own executed trades rather than from quoted prices, because a quoted spread in a thin instrument is an invitation rather than a fact. The model charges half the effective spread, an impact term that scales with participation, and an opportunity cost for the part of the order that never fills. Most candidates fail at the second term.

Participation rate is the variable a research team controls and most often ignores. A strategy requiring ten per cent of a day's volume in an instrument trading four million dollars a day is a strategy with a four hundred thousand dollar limit, and the size of the book is set by that number rather than by the size of the signal. We size every strategy from the capacity side first, then ask whether the expected return on the capacity available justifies the operational cost of running it.

Thin markets punish urgency and reward patience unevenly. Passive execution earns the spread but accepts adverse selection, because the counterparty trading against a resting order in a quiet instrument usually knows something. Aggressive execution pays the spread but takes the price it wanted. The rule here is that a strategy must be viable under one execution style, not under a blend chosen after the fact. Blended assumptions are how a backtest borrows returns from a decision nobody made.

Data quality is the quiet constraint. In thin instruments a daily closing price can be a single small trade, and a series built from those prints will show volatility that does not exist and correlation that does not either. The Information and Data division maintains our reference series with a documented origin for every field, and strategies are tested against that series rather than against a convenient one. Where a series cannot be reconstructed from source records, the strategy is not approved for capital.

None of this is an argument for trading less. It is an argument for knowing the size of the room before buying the furniture. The division reports capacity, realised execution cost and slippage against model to the Risk and Valuation Committee each quarter, and a strategy whose realised cost exceeds model by more than a stated margin is reduced automatically. The arithmetic is not complicated. It is simply performed before the position rather than afterwards.

Published 2021-04-14 by the Quantitative Strategies division. Research is prepared for eligible counterparties and does not constitute advice.