High-frequency trading behavior

Definition

High-frequency trading firms trade at microsecond-to- millisecond horizons, holding positions for seconds to minutes and ending most days flat. They are not one strategy but a technology tier running several documented behaviors: electronic market making (most HFT volume — inst-equity-market-makers at maximum speed), cross-venue and index arbitrage (enforcing ms-consolidated-tape price consistency), and short-horizon directional trading on order-flow signals. Brogaard-Hendershott-Riordan (2014, RFS) documents the net effect: HFT trades in the direction of permanent price changes and against transitory pricing errors — on average AIDING price discovery — while the stress-behavior record (2010) documents the caveat that matters for risk.

How it works / structure

• The behavioral repertoire (documented classes): MAKING — posting two-sided quotes, earning spread plus rebates, canceling fast when signals shift (the high order-to-trade ratios traders observe); ARBITRAGE — ETF/NAV, index/futures basis, cross-venue NBBO enforcement (why identical instruments rarely diverge for long); ANTICIPATION — inferring large institutional parent orders from child-order footprints (ms-execution-algos slicing exists BECAUSE of this) and trading ahead at short horizons — the documented adversarial layer, economically a tax on visible size.
• What retail actually experiences: near-instant fills and historically narrow spreads in liquid names (the documented liquidity benefit); quotes that fade when you try to hit stale prices (cancellation speed); and negligible direct interaction otherwise — retail-scale orders are below HFT anticipation thresholds and mostly internalized upstream (ms-payment-for-order-flow).
• What institutions experience: execution is a cat-and-mouse against anticipation — hence randomized slicing, dark venues, and TCA obsession (ms-implementation-shortfall); footprint hygiene is a real cost driver at size.
• Stress behavior: speed cuts both ways — HFT liquidity provision thins or inverts in dislocations (documented in the flash-crash record: some firms withdrew, some became aggressive sellers) — amplifying the fair-weather-liquidity property of the modern tape.

When it applies

Execution planning at any size (footprint awareness scales with order size); interpreting quote flicker and depth as ESTIMATES rather than commitments; intraday tactics (strategy-day-trading-styles — competing at HFT horizons directly is a losing proposition for humans; day-trading edges must live at horizons where speed doesn’t decide); market-quality regime reads (spread/depth behavior in stress).

Risk profile & failure modes

• Competing on speed: any strategy whose edge decays in milliseconds belongs to the co-located — retail latency arbitrage attempts are structurally dominated; the honest boundary is minutes-plus horizons.
• Depth illusion: resting size is continuously re-evaluated inventory (inst-equity-market-makers) — sizing market orders against displayed depth overestimates what will actually be there.
• Stress liquidity assumptions: the documented 2010 pattern — normal-times spread narrowness says nothing about crisis depth; stop-loss and liquidation plans priced at calm-market spreads carry hidden slippage risk (ms-slippage-friction).
• Narrative extremes: both “HFT is theft” and “HFT is pure liquidity” over-claim — the documented record is heterogeneous by strategy class; the KB carries the class-level distinctions, not the slogans.

Evidence & limits

Brogaard-Hendershott-Riordan (2014) anchors the price-discovery evidence; the SEC/CFTC flash-crash report documents stress behavior; anticipation economics are documented in the execution literature. Firm-level strategies are proprietary — class-level behavior is the honest resolution.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Visible resting depth at the NBBO fills at <60% of displayed size when hit during volatility spikes (depth-decay check)” — falsified by fill-rate data.
• “Randomized child-order execution reduces implementation shortfall vs deterministic slicing for orders >2% ADV (anticipation-cost thesis)” — falsified by paired TCA results.

Cross-references

• The parent business: inst-equity-market-makers
• The defense toolkit: ms-execution-algos, ms-implementation-shortfall, ms-order-types
• The plumbing: ms-consolidated-tape; the stress record: episode-flash-crash-2010