What high-frequency actually means in retail crypto

“High-frequency trading” is the most abused phrase in consumer-facing crypto marketing. It is plastered on landing pages of products that route orders through a serverless function in the wrong continent, and the buyer has no easy way to know the difference. This post is the framing we wish someone had handed us before we started building.

What HFT means in finance, in one paragraph

In equities and futures, high-frequency trading refers to a small number of firms operating on infrastructure where the budget is measured in microseconds and the head-to-head competition is for queue position on the matching engine. The hardware is specialised (FPGAs, custom NICs, dedicated cross-connects). The teams are roughly half engineers, half quants. The strategies decay in value the moment they are publicly described. None of this is commercially available as a subscription product.

What “low-latency API execution” means in retail crypto

What a retail crypto product can actually deliver is much more modest, and still useful: a server in the same region as the exchange match engine, a well-tuned client with reconnect/replay discipline, fee-tier awareness, and an order-routing path that avoids cron-based scheduling. That gets you to typical round-trip times on the order of tens of milliseconds, not microseconds.

That is enough to compete on fill rate and slippage against a consumer bot running from a wrong-region cloud function. It is not enough to be in the same conversation as a real HFT firm, and pretending otherwise is dishonest.

A practical taxonomy

• Microsecond tier: co-located on the venue LAN, hardware acceleration, microsecond budgets. Institutional only.
• Millisecond tier: in-region cloud or datacenter, tuned client, ~10-50 ms round trips. Achievable for retail-grade platforms that take execution seriously.
• Hundreds-of-milliseconds tier: any-region cloud, REST-only path, sometimes routed through a webhook layer. Most consumer bots live here. Calling it HFT is marketing fiction.

How to tell which tier a platform is actually in

You do not need to take anyone's word for it. Three checks:

1. Ask for a continuously published median and 99th-percentile round-trip time per region per exchange. A serious operator will already have it on their proof page.
2. Ask which match-engine regions they deploy in, and whether the region is selected automatically by an order router or pinned by configuration.
3. Ask whether the order path is REST polling on a schedule, or an event-driven WebSocket session with reconnect/replay guarantees. The second is harder to build and cheaper to run once it works.

Why we say “low-latency execution” and not HFT

We are millisecond-tier. Our advantage over a wrong-region consumer bot is real and measurable in fill rate and slippage. Our advantage over an actual HFT firm is zero. Saying so is the only frame in which the things we do well — in-region deployment, fee-aware routing, transparent measurement — register as edges instead of as marketing fluff.

If the word “HFT” appears anywhere on a product page without a co-location address attached, treat it as evidence of either confusion or marketing pressure. We would rather be specific.