BotIndex
Compare
GUIDE · 6 MIN READ

High frequency trading: how HFT works, and why your bot isn't doing it

BotIndex editorial team · Updated 6 Nov 2026 · Editorial policy

SHORT ANSWER

High frequency trading (HFT) is proprietary trading by firms that use very fast computers, servers placed next to exchange matching engines, and very short holding periods to profit from tiny price differences, mostly by market making and arbitrage. Retail crypto bots work through public APIs with rate limits, so they cannot compete on speed and are not HFT.

What is high frequency trading?

High frequency trading (HFT) is a style of automated trading in which firms use extremely fast computers and network connections to place, change and cancel very large numbers of orders, holding positions for seconds or less. The profit per trade is tiny; the business works because it repeats across millions of trades.

There is no single legal definition. The US Securities and Exchange Commission said in its 2010 concept release on equity market structure that the term “is relatively new and is not yet clearly defined.” It typically refers to professional firms trading their own money, with five characteristics often attributed to them (SEC Release 34-61358):

  1. Extraordinarily high-speed, sophisticated programs to generate, route and execute orders.
  2. Colocation services and individual exchange data feeds to cut latency.
  3. Very short time frames for opening and closing positions.
  4. Many orders cancelled shortly after they are submitted.
  5. Ending the day as close to flat as possible, without large unhedged positions overnight.

The same release noted that estimates of HFT volume in US equities, at that time, were typically 50% of total volume or higher. In other words, high speed trading is not a fringe activity in modern markets; it is a large part of how prices are quoted.

How does high frequency trading work?

HFT is a contest of speed and infrastructure first, and of clever models second. Three ideas explain most of it.

Latency and colocation

Latency is the delay between something happening on an exchange and a firm’s order reaching the exchange in response. The SEC explains that speed matters both in absolute terms and “in the relative sense of being faster than competitors, even if only by a microsecond.”

Colocation is the main tool. Exchanges, or third parties that host their matching engines, rent rack space so firms can put their own servers physically next to the matching engine, shortening the cable and the number of network hops (SEC concept release). Firms also buy direct data feeds from each venue rather than waiting for consolidated data.

How fast is fast? The SEC’s 2020 staff report cites research in which response times for one type of HFT strategy fell from about 5 microseconds in 2011 to 300 nanoseconds in 2018 (SEC staff report, 2020).

Crypto has a version of this too. Large crypto exchanges run their matching engines in specific data centers or cloud regions, and professional firms place servers nearby. Bybit’s API documentation, for instance, carries a notice that it will move its trading and matching engine to a specific AWS Tokyo availability zone on 12 November 2026 (Bybit API docs). That kind of detail matters to an HFT firm and not at all to someone running a grid bot.

HFT trading algorithms: the main strategy types

The SEC grouped the strategies of these firms into four broad types:

  • Passive market making. Posting resting buy and sell orders and earning the spread between them, plus any rebates venues pay to liquidity providers. When firms layer many orders at different prices, cancellation rates can reach 90% or more, according to the SEC.
  • Arbitrage. Capturing price gaps between related products, such as an ETF and the basket of stocks it holds. In crypto, the equivalent is the same coin priced differently across exchanges or between spot and futures. See arbitrage in our glossary.
  • Structural. Exploiting features of market structure, such as being faster than other participants to react to a quote change.
  • Directional. Taking a short-term view on price direction. The SEC flagged two directional strategies, order anticipation and momentum ignition, as potentially problematic for long-term investors.

The release also noted that many of these strategies are not new. What changed is the technology that lets firms spot and execute them faster.

Who are high frequency traders?

The SEC describes them as proprietary firms: independent trading companies (which may or may not be registered broker-dealers), proprietary desks at multi-service broker-dealers, and hedge funds. They trade their own capital, not client money, and spend heavily on technology, data and exchange connectivity.

Is HFT good or bad for markets? Regulation and context

Regulators have looked at HFT closely without reaching a simple verdict.

  • Fair access. The SEC said colocation services offered by registered exchanges are subject to the Exchange Act, must be filed as rule changes, and must not be unfairly discriminatory. It asked the public whether colocation fees are high enough to act as a barrier for smaller firms (SEC concept release).
  • Stress events. A study of the May 6, 2010 Flash Crash by Kirilenko, Kyle, Samadi and Tuzun, authorized for public distribution by the CFTC, found that HFTs “did not cause the Flash Crash, but contributed to it by demanding immediacy ahead of other market participants.” The authors suggested short trading pauses when large order imbalances appear (Kirilenko et al.).
  • Speed bumps. Some venues deliberately slow everyone down. The SEC’s 2020 staff report describes the Investors Exchange (IEX) adding a 350-microsecond delay by routing external communications through a coil of fiber-optic cable (SEC staff report, 2020).

Crypto markets are less uniformly regulated than US equities, and the rules differ by country. Our country guides cover which bot platforms and exchanges are available where.

Why retail trading bots are not HFT

Search results for “HFT trading software” or “high speed trading software” often lead to retail bots marketed as high-frequency. Treat that label skeptically. Retail bots connect to exchanges through public APIs over the internet, and those APIs are built with hard limits.

Professional HFT Retail crypto bot
Connection Colocated servers, direct feeds Public REST/WebSocket API over the internet
Speed that matters Microseconds and below Seconds to minutes
Order limits Negotiated, high-capacity access Public rate limits per IP and account
Holding period Seconds or less Minutes to months
Edge Being first Strategy fit to market conditions

The rate limits alone rule out HFT-style order flow:

  • Bybit allows 600 HTTP requests per 5-second window per IP by default. Exceeding it returns “403, access too frequent”, and you must stop and wait at least 10 minutes. Order creation is further limited per account by default to between 10 and 20 requests per second, depending on the product (Bybit rate limits).
  • Binance returns HTTP 429 when you break a request limit and auto-bans IPs that keep sending requests after a 429 (HTTP 418), with bans scaling from 2 minutes to 3 days for repeat offenders. Its documentation also caps unfilled order counts per account (Binance API docs).

Add the round trip from your home or a cloud server to the exchange, and a retail bot is many orders of magnitude slower than the firms quoting the order book. That is fine, as long as the strategy does not depend on being first. Paying for a bot that claims to win speed races against those firms is paying for something it cannot deliver.

What a retail crypto trader can realistically automate instead

Retail automation works best where speed is not the edge. Practical options:

  • Grid bots place a ladder of buy and sell orders in a price range and profit from swings inside it. They are a slow, simple cousin of market making. Compare options in our best grid trading bots ranking and model a range with the grid profit calculator.
  • DCA bots buy on a schedule or on dips to build a position over time. See best DCA trading bots.
  • Arbitrage and market-making bots exist for retail, for example on Cryptohopper and HaasOnline, but expect thin, contested margins: you are competing with faster firms for the same gaps. Our guides to market making bots and crypto arbitrage bots explain the risks.
  • Rule-based and signal bots that act on indicators or TradingView alerts over minutes or hours.

If you want to build and test your own strategies, start with the research workflow in our guide to quantitative trading. For ready-made tools, our crypto trading bot rankings compare platforms on fees, strategies and availability. None of these will make you a high frequency trader, and none of them guarantee profit; they automate a plan you are responsible for.

FAQ

Can I buy HFT trading software as a retail trader?

You can buy or download software that trades often, but that is not high frequency trading in the professional sense. The SEC describes HFT as relying on colocation, individual exchange data feeds and extraordinarily high-speed programs; the speed comes from infrastructure, not from a program you install at home.

Is high frequency trading legal?

Yes, HFT is a legal activity, but it operates inside market rules. The SEC's 2010 concept release says colocation services offered by registered exchanges are subject to the Exchange Act and their terms must not be unfairly discriminatory. Manipulative practices are illegal whatever the speed.

Did high frequency traders cause the 2010 Flash Crash?

A study by Kirilenko, Kyle, Samadi and Tuzun, authorized for release by the CFTC, found that HFTs did not cause the May 6, 2010 Flash Crash but contributed to it by demanding immediacy ahead of other traders.

How fast is high speed trading?

Very fast. The SEC's 2020 staff report cites research where response times for one type of HFT strategy fell from about 5 microseconds in 2011 to 300 nanoseconds in 2018. A microsecond is a millionth of a second.

Sources

Not financial advice. Trading bots do not remove market risk.