How High-Frequency Market Makers Adjust Liquidity Based on Trader Sophistication

A working paper posted on SSRN, built on trader-level data from India's National Stock Exchange, finds that proprietary algorithmic liquidity providers systematically recalibrate their quotes according to the perceived sophistication of the opposing flow. According to the study, these market makers withdraw liquidity when technologically advanced participants generate stronger order imbalance — and expand provision when less sophisticated order flow dominates.

Conditional depth and the behavioral read

The finding reframes how displayed book size should be interpreted. Liquidity is not a static field but a state-dependent one, with the same instrument clearing at materially different effective spreads depending on who else is pressing the tape. For active traders operating on tight spreads and shallow pullbacks, this asymmetry translates directly into slippage variance that no chart pattern alone can isolate.

The mechanism appears behavioral. Proprietary desks seem to classify counterparties by execution signature rather than by identity, treating order imbalance as an information proxy. A surge from a sophisticated participant reads as adverse selection risk; a surge from a less informed participant reads as harvestable flow. The institutional footprint shifts accordingly — quietly, with no change in posted size.

The hardware layer is repricing in parallel

The study lands as the infrastructure underpinning fast execution is itself being re-rated. At the World Federation of Exchanges' late-August webinar, Beeks Group CEO Gordon MacArthur stated that AI-driven demand has pushed server pricing to 300–400% above levels seen 18 months ago, with lead times stretching back toward pandemic-era norms. That repricing sharpens the asymmetry above: firms running thinner colocation budgets fall further back in queue position, and that queue position itself becomes a signal read by their counterparties — a second-order liquidity void.

Operating in conditional markets

For practitioners, the operational read is to treat depth as state-dependent. Mean reversion setups in Indian equities may reflect not just price action but the recurring rhythm of informed-versus-uninformed flow cycles; sizing off relative depth, and tracking order imbalance asymmetry as a leading indicator of withdrawal, becomes more useful than relying on absolute spread. The same conditional logic extends across regions — India's structural ascent, including the creator-driven expansion of its media market, is one of several parallel shifts quietly rewriting how execution assumptions should be calibrated.