“High-frequency trading” sounds like one strategy. It isn’t. HFT is a speed regime: holding periods measured in milliseconds to minutes, thousands of small trades, and almost no overnight position. Inside that regime sit several very different ways of making money, and each one needs a different kind of speed, a different kind of capital and a different kind of venue.
This guide explains the main high-frequency trading strategies in plain terms: where the profit actually comes from, what goes wrong, and which of them a trader outside a large firm can realistically run. We use simple, clearly labelled illustrative numbers rather than marketing claims.
What makes a strategy “high-frequency”?
There is no legal or official definition, but practitioners and regulators tend to agree on a few traits:
- Very short holding periods – from microseconds to a few minutes.
- High message and order counts – many quotes and cancels for each trade that actually executes.
- Small edge per trade, large volume – a fraction of a pip or a tick, repeated thousands of times.
- Flat at the end of the session – little or no inventory carried overnight.
- Fully automated – no human is fast enough to click the trades.
Every HFT strategy gets paid from one of three sources. Keeping these three apart is the single most useful way to understand the field.
The six core high-frequency trading strategies
1. Market making
A market maker posts a buy price and a sell price at the same time and earns the difference when both sides fill. On a liquid FX pair that might mean quoting 1.08500 / 1.08510 and earning one pip on every clean round trip. Market making is widely reported to be the largest single category of HFT activity on exchanges, and it is the reason modern spreads are so tight.
The catch is adverse selection. When the price is about to move, the traders who know it first will hit the maker’s old quote. The maker ends up holding exactly the position it did not want. Speed matters here not to grab opportunities but to cancel stale quotes before someone else exploits them.
The arithmetic in Figure 2 is the heart of the business. If a clean round trip earns +1.0 pip and a picked-off fill loses 2.5 pips on average, a maker who is picked off 30% of the time loses 0.05 pip per trade. Get picked off only 20% of the time and the same book makes +0.30 pip per trade. That is why market makers spend so much on infrastructure: the money is in the cancel, not the quote.
2. Rebate capture (liquidity provision for rebates)
Many stock and futures exchanges run a maker-taker model: they pay a small rebate to whoever adds a resting order and charge a fee to whoever removes it. Some strategies are built to collect these rebates, often trading close to break-even on price and profiting from the rebate. This only exists on venues with that fee model. Most retail FX and CFD accounts have no maker rebate at all, so this strategy simply does not translate to an MT4 or MT5 account.
3. Latency arbitrage
Latency arbitrage trades the short window when one price source has already moved and another has not. The trader sees the new price first, buys or sells against the stale quote, and closes when the slower venue catches up. Between exchanges this is a microsecond race between colocated firms. Against a retail broker whose feed lags the wider market by tens or hundreds of milliseconds, the window is far wider – which is why it is the strategy most often discussed outside institutions. We cover the mechanics in depth in our guide to how latency arbitrage works.
Notice that latency arbitrage and market making are two sides of the same trade. One trader’s latency profit in Figure 2 is the market maker’s loss. That is also why brokers that act as the market maker watch for this flow closely.
4. Statistical arbitrage
Statistical arbitrage bets that a historically stable relationship between instruments will hold over the next few seconds, minutes or hours: two correlated currency pairs, a futures contract and its index, or a basket of related assets. When the spread between them stretches, the strategy sells the rich leg and buys the cheap one. At high frequency this means short horizons and many small trades; the edge comes from the model, and speed mostly protects it from slippage. If you want the full worked example with z-scores and cost stacks, read our statistical arbitrage and pairs trading explainer.
5. Event and news trading
Economic releases, central-bank decisions and company news hit the market at known times. Event-driven HFT reads machine-readable feeds and trades in the first milliseconds after the number is published. Institutional players pay for low-latency news feeds and colocated servers next to the release point. Retail traders face the opposite conditions at those moments: spreads widen, liquidity disappears and slippage is at its worst, so a “news bot” on an ordinary account usually trades at the worst prices of the day.
6. Order-flow and microstructure prediction
The order book contains short-term information: when buy orders heavily outweigh sell orders at the top of the book, the next tick is slightly more likely to be up. Microstructure strategies model imbalance, queue position and trade flow to predict the next few price changes. They need full depth-of-book data and very fast reaction, which is why they mostly live on exchanges with public order books (futures, equities, crypto), not on OTC FX accounts that show only a top-of-book quote.
HFT strategies compared
The table summarises the six strategies by where the edge comes from and what it takes to run them. Holding times and requirements are typical ranges, not rules.
| Strategy | Edge source | Typical holding time | Speed needed | Realistic outside a large firm? |
|---|---|---|---|---|
| Market making | Bid-ask spread | Seconds or less | Microseconds (to cancel stale quotes) | Rarely; some crypto venues only, with real inventory risk |
| Rebate capture | Exchange maker rebates | Seconds | Microseconds | No – needs exchange membership-level access and a maker-taker venue |
| Latency arbitrage | Stale quotes on a slower venue | Milliseconds to seconds | Faster than the slow venue’s feed | Yes, against slower broker feeds – subject to broker rules |
| Statistical arbitrage | Mean reversion of related prices | Seconds to days | Moderate; mostly to limit slippage | Yes, at the slower end of the horizon |
| Event / news | First reaction to new information | Milliseconds to minutes | Microseconds for the first move | Hard – retail fills are worst exactly at release time |
| Order-flow prediction | Order-book imbalance | Sub-second | Microseconds plus full depth data | Partly, on crypto exchanges with public books |
Same strategy, very different speed requirements
The question “can I do HFT?” really means “can my setup react fast enough for the strategy I picked?”. Figure 3 places each strategy on a log scale against the reaction time different setups can deliver.
Two things stand out. First, the strategies that dominate institutional HFT – exchange market making, rebate capture, cross-venue arbitrage – need reaction times only colocated hardware can deliver. Competing there from a VPS is like entering a Formula 1 race in a family car. Second, the strategies that remain open to individual traders work because the other side is slow (a retail broker’s lagging feed) or because the edge comes from a model rather than raw speed (stat arb). The layers that make up these setups – feeds, servers, colocation, networks – are explained in our overview of high-frequency trading infrastructure.
Which HFT strategies can a retail or prop trader actually run?
An honest shortlist, from most to least accessible:
- Latency arbitrage against a slower broker. The edge comes from the gap between a fast reference feed and the broker’s quote, so you do not need institutional speed – you need to be faster than the broker. The real constraints are broker policy, slippage and how long an account stays profitable before the broker reacts.
- Short-horizon statistical arbitrage. Minutes-to-hours holding periods keep speed requirements modest. The hard part is research: finding relationships that survive costs and do not break after you start trading them.
- Two-leg and hedge-style arbitrage. Opening offsetting positions at two venues reduces exposure to a single broker’s reaction, at the price of double costs.
- Crypto order-flow and cross-exchange strategies. Public order books and open APIs make microstructure data available to anyone, but competition from professional firms is intense and exchange risk is real.
- Market making and rebate capture. Almost always out of reach without exchange-level access and serious capital.
If you are still deciding between approaches, our guide on choosing the arbitrage strategy that fits your capital and broker walks through the trade-offs, and the complete HFT arbitrage guide covers the full pipeline from feed to execution.
Risks every HFT strategy shares
- Costs bigger than the edge. With profits of a fraction of a pip per trade, a small rise in spread, commission or slippage can turn a winning system into a losing one.
- Adverse selection. Whoever is slower in a given trade pays whoever is faster. Every strategy is someone else’s counterparty.
- Technology failure. A frozen feed, a disconnected server or a bug that sends orders in a loop can lose more in a minute than the strategy earns in a month. Kill switches and position limits are not optional.
- Regime change. Relationships that held for years can break after a central-bank shift or a change in market structure.
- Venue and broker rules. Some brokers and prop firms restrict certain high-frequency or arbitrage styles. Read the terms before you trade, not after a payout is refused.
“Strategies” that are not strategies: manipulation to avoid
Some practices that appear in HFT discussions are illegal in most major markets: spoofing (placing orders you intend to cancel to push the price), layering (a stack of such orders at several levels) and quote stuffing (flooding a venue with messages to slow others down). Regulators have fined and prosecuted firms for all three. They are market abuse, not edge, and no legitimate trading setup should include them.
Key takeaways
- HFT is a speed regime; the strategies inside it earn from liquidity, from being first, or from being right.
- Market making is the largest HFT activity, and its profit depends on cancelling stale quotes fast enough to avoid being picked off.
- Latency arbitrage and market making are opposite sides of the same trade.
- For individual traders, the realistic HFT strategies are the ones where the counterparty is slow or the edge comes from a model: latency arbitrage against slower brokers and short-horizon stat arb.
- Always model full costs, plan for technology failure and check venue rules before going live.
Frequently asked questions
What are the main high-frequency trading strategies?
The six core HFT strategies are market making, rebate capture, latency arbitrage, statistical arbitrage, event or news trading, and order-flow (microstructure) prediction. They earn money from one of three sources: providing liquidity, being first to new prices, or correctly forecasting short-term moves.
What is the most common HFT strategy?
Market making is widely reported to be the largest category of high-frequency trading on exchanges. Market makers continuously quote buy and sell prices and earn the spread, using speed mainly to cancel stale quotes before faster traders can pick them off.
Can retail traders do high-frequency trading?
Not in the institutional sense – exchange market making and cross-venue arbitrage need colocated hardware. But some HFT-style strategies are realistic for individuals: latency arbitrage against a broker whose feed lags the market, and short-horizon statistical arbitrage where the edge comes from a model rather than raw speed.
Is market making the same as latency arbitrage?
No – they are opposite sides of the same trade. A market maker earns the spread by quoting both sides; a latency arbitrageur profits when the maker's quote is stale. When a latency trader wins, the market maker on the other side usually loses.
Are HFT strategies legal?
Market making, arbitrage and statistical trading are legal in major markets. Spoofing, layering and quote stuffing are forms of market abuse and are illegal. Separately, individual brokers and prop firms may restrict certain high-frequency or arbitrage styles in their own terms, so check the rules of any venue you trade with.
Can you use high-frequency trading strategies in forex?
Yes, but the forex market is mostly over-the-counter, so exchange-specific strategies such as rebate capture do not apply to most FX accounts. In FX the realistic HFT-style approaches are latency arbitrage against slower broker feeds and short-horizon statistical arbitrage between related currency pairs.