Quote stuffing
†deadFlood a venue with orders and instant cancellations to congest the data feed and slow competitors: a denial-of-service on the book. Manipulative under MAR and US rules; covered for detection.
The idea
Reference figure. This concept is explained in prose and diagram; the interactive widgets live on the flagship pages it links to under Where this fits.
What is quote stuffing, and why is it illegal?
Quote stuffing is rapidly entering and withdrawing a very large number of orders to deliberately overload the feed and the matching engine, so that competitors processing that feed fall behind. The orders are not meant to trade; their purpose is to congest the infrastructure, manufacturing a brief latency advantage for the stuffer over everyone forced to process the noise. It is, in effect, a denial-of-service attack against the market's plumbing.
The mechanism is simple. Every order, modification and cancellation is a message the venue must process and broadcast, and that every other participant must parse. If one participant emits tens of thousands of messages a second with no economic purpose, the feed and rivals' parsers congest, and their view of the market lags. The stuffer, who knows the burst is noise, acts on the real state while competitors are still digesting the flood. The defining feature (what separates it from a legitimately busy market maker) is that the message volume serves no economic function: a near-zero fill rate, no inventory purpose, and timing that coincides with the stuffer's own trades.
In the EU and UK it is explicitly named: MAR Article 12(2)(c) lists "quote stuffing" as an example of manipulation that disrupts the venue's functioning or makes it harder for others to identify genuine orders. It is in the regulation itself, not merely inferred. In the US, the CEA's disruptive-practices provisions (§4c(a)(5), as amended by Dodd-Frank) reach conduct showing intentional or reckless disregard for orderly execution, and SEC/FINRA pursue analogous securities conduct under the general anti-manipulation rules. Exchange rulebooks independently prohibit conduct that degrades system performance, so stuffing can be a rule violation before it is a statutory one. See the market-abuse regimes.
How is it detected?
Quote stuffing is invisible in the trade record and unmistakable in the message record, so detection runs on message-rate telemetry the venue already captures to the microsecond. Surveillance monitors each participant's messages-per-second, its order-to-trade ratio, the lifetime of its orders, and the correlation between its bursts and feed latency for everyone else.
Four telemetry tells combine. A message-rate anomaly: a spike in adds plus cancels per unit time far above the participant's and its peers' baselines. An exploding order-to-trade ratio: thousands of messages per execution, beyond any genuine quoting or hedging need. A tiny cancel lifetime: orders cancelled within microseconds, never resting long enough to plausibly trade. And a latency correlation: the burst coincides with a measurable rise in dissemination latency for peers (the very effect the conduct exists to produce) and is timed around the stuffer's own genuine trading. The full timestamped log makes the burst reconstructable, and machine-learning models trained on normal message-rate distributions flag the anomaly automatically.
How do venues and regulators price it out?
The structural defence does not rely on proving intent case by case; it makes message volume cost something, so stuffing becomes uneconomic and self-limiting. Four controls do the work. Message throttles cap each participant's messages-per-second; excess is rejected or queued, removing the congestion mechanism. Order-to-trade-ratio limits (mandated by MiFID II RTS 9, which requires venues to set a maximum OTR per instrument) throttle or penalise a participant whose messages vastly exceed its trades. Message-based fees charge per message (or per cancellation above a ratio), internalising the cost the burst imposes on the system. And minimum resting times or speed bumps require an order to rest for a minimum interval, removing the fire-and-cancel pattern stuffing depends on.
Where the conduct is nonetheless proven, regulators pursue it under MAR Article 12 or the CEA, and venues discipline it under their rulebooks. The combination (telemetry detection plus structural pricing) is why quote stuffing is now substantially mitigated relative to the early-2010s era when it was first identified. See the market-abuse regimes for the prohibition and how the OTR controls fit a firm's own monitoring obligations.
Is a busy market maker doing quote stuffing?
No, and the distinction is the whole point of the controls. A genuine market maker also sends huge message volumes, constantly re-quoting as the price moves. The difference is economic purpose and effect: the maker's messages maintain real, executable two-sided quotes that produce fills; the stuffer's serve only to congest, with near-zero fills and a deliberate latency effect on others. OTR limits are calibrated precisely to permit the former and catch the latter.
This is why a high-message strategy must be designed with the OTR limits and message-fee schedule in mind, not to evade them, but because legitimate quoting must stay within the envelope the venue defines for genuine activity, and because a poorly tuned re-quoting loop can look like stuffing on the telemetry. The maker's volume comes with fills and inventory purpose; the stuffer's comes with neither. The metrics above are built to read exactly that difference.
Worked example
A synthetic telemetry reconstruction, as of 2026: how it is caught, not how it is done. A participant's normal baseline on an instrument is ~50 messages/second with an order-to-trade ratio around 20:1, busy but economic. In a 200 ms window it emits ~12,000 add/cancel messages, an equivalent rate near 60,000/second, with zero executions in that window. The venue's dissemination latency to other participants rises measurably across the burst, and the participant's own genuine order trades immediately afterwards, into the lag.
Read the tells. The message rate runs ~1,200× the participant's own baseline for the window, a glaring anomaly against its history. The in-window order-to-trade ratio is undefined: 12,000 messages, zero trades. The cancel lifetime has a median order life of ~30 µs, far below any plausible execution intent. And the latency correlation shows feed latency for peers spiking exactly across the burst, then recovering: the effect the conduct exists to produce. Under a MiFID II OTR limit (RTS 9) and a per-message fee, this burst would be throttled or made directly expensive, the structural defence that bites before a case is even built.
The lesson: quote stuffing is invisible in the trade record and unmistakable in the message record, which is why message-rate telemetry, not the tape, is where it is caught and where the controls bite. The figures are synthetic and illustrative; real baselines, thresholds and OTR caps are calibrated per instrument and per venue. See market manipulation for the surrounding conduct.