High Frequency Trading

Not all algorithmic trading looks the same. A pension fund's algorithm might slice one large order into pieces over an hour. A high frequency trading (HFT) firm is doing something different in kind, not just in speed.

The workbook's definition: "High frequency trading (HFT) is a type of algorithmic trading which is latency sensitive and is characterized by a high daily portfolio turnover and high order-to-trade ratio (OTR)." Every word in that sentence is doing work. It is a subset of algorithmic trading, not a separate category. "Latency sensitive" means microseconds of delay matter to the strategy. "High order-to-trade ratio" means the firm places far more orders than it ever lets execute — most are cancelled or modified before anyone trades against them.

HFT strategies depend on speed, and the workbook's Chapter 3 describes the infrastructure that makes that speed possible even though it does not name HFT alongside it directly:

• Co-location (or proximity hosting) is a facility stock exchanges offer to brokers and data vendors, letting their trading or data-vending systems sit within or at close proximity to the exchange's own premises, connected through a direct and private network rather than the public internet. Shaving the physical distance to the exchange's matching engine shaves the latency an HFT strategy is built around.
• Algorithmic (ALGO) trading itself is any order generated using automated execution logic — software that, once its parameters are met, generates and pushes buy/sell orders into the exchange's trading system without manual entry. HFT sits inside this category as the fastest, highest-turnover end of it.

Regulatory guardrails that apply because HFT is algorithmic trading:

• Algorithmic trading systems must undergo a system audit every six months, by a certified system auditor, to check that SEBI/exchange requirements are effectively implemented.
• Under SEBI's 4 February 2025 guidelines on safer retail participation in algorithmic trading, brokers act as the principal and any algo provider or fintech/vendor acts as the broker's agent when its API is used; retail-built algorithms must be registered with the exchange, through the broker, once they exceed a specified order-per-second (OPS) threshold.
• SEBI has directed exchanges to ensure all algorithmic orders are routed through broker servers located in India, with risk controls including price checks, quantity limit checks and order value checks.

A proprietary trading firm runs an HFT strategy in a liquid large-cap stock. Over one trading session:

A very high fraction of its orders never trade at all — they are placed to test the market's reaction and cancelled within milliseconds if the price moves against the strategy. That 40:1 ratio, and the fact that the firm's portfolio turns over many times in a single day, is exactly what the workbook's definition is pointing at: latency sensitivity and a high order-to-trade ratio.

To run this, the firm has taken co-location at the exchange — its servers sit in the exchange's own data centre, connected by a direct private network, rather than routing orders over the public internet from an office across the city. Its algorithmic trading system, like every other broker-side algo system, is due for its six-monthly system audit by a certified system auditor. A retail investor trying to run a similar strategy from a home-built script would, once past the exchange's specified OPS threshold, have to register that algorithm with the exchange through their broker — the broker standing as principal, the investor's script-provider as agent.

Chapter 3 (Introduction to Securities Broking Operations), section 3.1.1, introduces Algorithmic Trading and High Frequency Trading together, immediately after Direct Market Access, and ties both to the co-location facility described a page earlier. Expect a definitional question — HFT is a type of algorithmic trading characterised by latency sensitivity, high turnover and a high order-to-trade ratio — and a question on what co-location is and who it is offered to.

• HFT is a type of algorithmic trading, not a separate facility. Every HFT order is an algo order; not every algo order is HFT.
• "High order-to-trade ratio" means many more orders than trades — most HFT orders are cancelled, not executed. A high OTR is a feature of the strategy, not a sign something has gone wrong.
• Co-location is offered by the exchange to brokers and data vendors, not directly to a retail client, and connects through a direct and private network — not simply "a fast internet connection".
• The six-monthly system audit applies to algorithmic trading systems generally, because HFT is algorithmic trading — do not treat it as an HFT-specific rule with a different cycle.
• DMA and algorithmic trading are different facilities that can be combined, not synonyms — DMA is about who enters the order (the client, directly), while algorithmic/HFT is about how the order is generated (automated logic).