Spread Widening

Spread widening is, in itself, legitimate microstructure. A market maker quoting a two-sided price is short an option to every better-informed counterparty; when a scheduled release is seconds away, the probability of being run over by informed or simply faster flow spikes, and the rational response is to quote wider or withdraw. Liquidity thins, the touch widens, and fills degrade — at every venue, for every participant. The institutional market is not exempt.

The measurement question that matters is not whether spreads widen but how a specific venue’s widening compares to the market’s. Three properties characterise a widening profile: magnitude (how many multiples of the baseline spread), duration (how long until the spread renormalises), and timing (whether widening anticipates the event, tracks it, or persists suspiciously after underlying liquidity has returned). Comparing one broker’s profile against a multi-venue reference at the same timestamps separates market-wide widening, which is physics, from venue-specific widening, which is policy.

The retail relevance is blunt. Advertised “typical” spreads are averages sampled predominantly in calm regimes, but retail order flow does not arrive uniformly across regimes: stops and pending orders cluster precisely around the events where spreads are widest, so the spread a trader’s orders actually experience can differ materially from the spread the marketing page reports. A backtest fed the advertised average inherits the same optimism — one of the quieter routes into backtest overfitting via understated execution cost.

Spread widening also interacts with feed integrity: a venue can disguise widening by throttling or suppressing ticks during the event window, which is the territory of censoring and quote staleness. The Broker Feed Auditor we are building is intended to measure widening profiles from recorded feeds rather than assert them.