What is High-Frequency Trading (HFT)?
Reviewed by Karthikeyan Anbarasan (Founder, TradeLyser) · Updated 19 Oct 2026 · Editorial policy
How high-frequency trading works
HFT is a subset of algorithmic trading defined by speed and turnover. Firms run servers close to the exchange’s matching engine, react to changes in the order book within microseconds and keep very little inventory overnight. Profit comes from many tiny gains, so the edge only exists if costs per trade are lower than the gain per trade.
Market making
Arbitrage
Short-term signals
Why the cost arithmetic rules out retail HFT
Take one Nifty futures lot of 65 bought and sold at 25,000, a turnover of ₹16,25,000 each way (illustrative). STT at 0.05% on the sell side is ₹812.50. Exchange charges at 0.00183% are about ₹29.74 a side, stamp duty on the buy side is ₹32.50, the SEBI fee is ₹3.25, and an assumed ₹20 brokerage per order plus 18% GST on brokerage and fees adds about ₹58.49. The total is about ₹966, close to 15 Nifty points.
A retail account trading through a broker API also has a longer path to the exchange than a server sitting next to it, so by the time its order arrives the price it reacted to may be gone. Retail strategies need edges measured in many points per trade, not fractions of one.
HFT vs scalping
| High-frequency trading | Retail scalping | |
|---|---|---|
| Holding time | Milliseconds to minutes | Seconds to tens of minutes |
| Decision maker | Code, no human per trade | Usually a person, sometimes a simple algo |
| Trades per day | Thousands or more | A handful to a few dozen |
| Edge per trade | Fractions of a point | Needs several points to cover costs |
| Infrastructure | Co-located servers, direct market data | Broker terminal or broker API |
What it means for your own trading
- Tag fast intraday trades as scalping, not HFT, so their statistics are judged on the costs you actually pay.
- Log the expected price and the fill for every market order; the gap is slippage, as in Figure 1.
- Compare average gross profit per trade with average charges per trade. If charges take most of it, the setup needs bigger targets or fewer trades.
- Prefer limit orders in thin option strikes, where walking the book costs the most.
Reference guide
| Context | Value | Reading |
|---|---|---|
| Scalping by hand or with a simple broker API | Not HFT | Tag it as scalping; its costs and speed are nothing like an HFT firm’s |
| Edge per trade smaller than round-trip charges | Negative after costs | A 1-point Nifty futures edge (₹65 a lot) cannot cover about ₹966 of charges |
| Reading HFT activity as a reason for every loss | Unhelpful excuse | Check your own entries, size and slippage in the journal first |
Questions
What is high-frequency trading in simple words?
Computer programs that trade very often, very quickly, for very small gains per trade, usually from servers placed next to the exchange. Holding times run from milliseconds to minutes.
Can retail traders do HFT in India?
Not in any real sense. HFT depends on co-located servers and very low costs per trade. A retail account pays STT and other charges of about ₹966 per Nifty futures lot round trip at 25,000, so sub-point edges cannot survive.
Is HFT legal in India?
Yes. Algorithmic and high-frequency trading are permitted for exchange members under SEBI and exchange rules covering approvals, risk controls and order limits. Check the current SEBI and NSE circulars for the details.
What is the difference between HFT and algo trading?
Algo trading is any trading where software places orders by rules. HFT is the fastest, highest-turnover part of it. A rule-based strategy holding positions for hours is algo trading but not HFT.
Does HFT make retail traders lose money?
HFT firms quote much of the spread you trade against, so their presence shows up in the price you get. Your own results still depend mostly on your entries, size, costs and slippage, which a journal can measure.