Frequently asked questions

HFT is automated trading that competes on speed and order-book microstructure: a machine quotes, cancels and executes in microseconds-to-milliseconds, typically holding positions for seconds or less and ending most days near flat. It is a subset of algorithmic trading defined by latency sensitivity and very high message rates, not by any single strategy. Read more →

Quant trading uses mathematical models to decide what to trade (any horizon). Algorithmic trading automates how an order is executed. HFT is the latency-sensitive corner of both: quant signals plus algorithmic execution run fast enough that microseconds and queue position decide the P&L. All HFT is algo and quant; most quant and algo trading is neither fast nor HFT. Read more →

No. You need strong probability, statistics and programming, and real microstructure understanding; a PhD signals those but is not the gate, especially for independent crypto and prediction-market work. Sell-side HFT desks often prefer them; an open venue does not check credentials. This site front-loads the maths so the prerequisite is competence, not a certificate. Read more →

It depends on the strategy. Pure latency races need colocation and often FPGAs; a laptop cannot compete. But slower microstructure strategies (inventory-aware market making on a single venue, statistical arbitrage, mid-frequency order-flow signals) are runnable from a laptop, especially in 24/7 crypto and prediction markets where the latency arms race is younger. Read more →

Yes. HFT is legal and a large share of regulated-venue volume. What is illegal is manipulation (spoofing, layering, momentum ignition, quote stuffing) prohibited under the US Dodd-Frank Act §747, the EU Market Abuse Regulation (MAR, 2014) and MiFID II (2018). The technology is lawful; specific manipulative intent and conduct are not. Read more →

As an industry, yes, but margins have compressed for over a decade and aggregate HFT revenue is far below its 2009 peak. Pure speed in mature lit equities is a commoditised, capital-intensive utility dominated by a few firms. The live edge in 2026 sits in newer or fragmented venues (crypto, prediction markets) and in smarter microstructure signals, not raw latency. Honest status: structural, but not the open frontier it was. Read more →

At microsecond latency, large models are too slow to sit in the hot path, so the inner loop stays classical signal processing. AI’s real impact in 2026 is off the critical path: faster research and feature discovery, machine-readable news and sentiment for event trading, better simulation, and, crucially, collapsing the engineering cost of building a stack, which makes a small independent shop more viable than before. Read more →

For latency-race strategies in mature markets, no: that needs institutional capital and infrastructure. For microstructure-driven trading on open venues, increasingly yes: crypto and prediction markets give direct API access without a prime broker, cloud and AI tooling cut the build cost, and a one-person market-making or stat-arb shop is a real (if hard, capital-constrained) proposition. Educational only, not a promise of profit. Read more →

A limit order book (LOB) is the venue’s live, sorted record of all resting buy (bid) and sell (ask) limit orders at each price. The highest bid and lowest ask form the best bid and offer; the gap between them is the spread. Incoming market orders consume resting liquidity from the top down. It is the central data structure of every order-driven market. Read more →

Price-time priority is how most order-driven venues rank resting orders: better-priced orders fill first, and among orders at the same price, the one that arrived earliest fills first (FIFO). It is why queue position has economic value: being early in the queue at a price means you trade before later orders. Some venues use pro-rata or size-time hybrids instead. Read more →

The microprice (Stoikov, 2017) is a fair-value estimate that weights the bid and ask by the opposite side’s size, so it sits closer to the side with more pressure. Intuitively, when bids vastly outnumber asks, fair value is near the ask, not the mid. It is a better short-horizon predictor of the next mid-move than the simple midpoint, and a core input to order-flow market making. Read more →

Order-flow imbalance (OFI) measures the net pressure from changes in bid versus ask depth and trades over a short window. A strongly positive imbalance (bids growing, asks consumed) predicts a near-term upward price move; negative predicts down. It is one of the most robust short-horizon microstructure signals and underlies both microprice and many market-making quote adjustments. Read more →

Crypto exchanges publish full L2/L3 order-book data over free public WebSocket APIs, the most accessible real high-frequency data in 2026. Prediction-market venues (e.g. Polymarket) expose their books too. Equities tick/depth data is gettable but costly. For learning, capture a crypto feed yourself; for serious backtesting you need a clean, point-in-time L2/L3 dataset, the kind of resource on our datasets waitlist. Read more →

Because it breaks the assumptions slower data lets you make. It is enormous (terabytes per day per market), irregularly timed (events, not clock ticks), heavy-tailed and non-normal, and contaminated by microstructure noise such as bid-ask bounce. Timestamps need care, and naive resampling destroys exactly the signal you want. It demands event-driven, point-process thinking rather than fixed-interval statistics. Read more →

Bid-ask bounce is the artificial up-down sawtooth in transaction prices caused by trades alternating between hitting the bid and lifting the ask, with no change in the underlying fair value. It inflates measured volatility and creates spurious negative autocorrelation in returns. Roll’s 1984 model uses exactly this serial covariance to back out the effective spread from trade prices alone. Read more →

You infer the aggressor, because most feeds do not label it. The tick rule signs a trade by whether its price rose or fell from the prior trade. The Lee–Ready algorithm (1991) compares the trade price to the prevailing midpoint (above mid is buyer-initiated, below is seller-initiated) falling back to the tick rule at the mid. Both are imperfect; accuracy drops in fast or fragmented markets. Read more →

A market maker quotes a bid and an ask simultaneously and earns the spread when both sides fill against uninformed flow: buy low, sell high, repeatedly, holding little net position. The structural cost is adverse selection: informed traders fill you on the wrong side just before the price moves. Net P&L is spread captured plus rebates, minus adverse selection, minus inventory and infrastructure cost. Read more →

Adverse selection is the market maker’s core risk: a disproportionate share of your fills come from traders who know something you don’t, so you systematically buy just before prices fall and sell just before they rise. Formalised by Glosten–Milgrom (1985) and Kyle (1985), it is the reason spreads exist at all: they must compensate the maker for trading against the informed. Read more →

The Avellaneda–Stoikov model (2008) is the canonical inventory-aware market-making framework. It derives optimal bid and ask quotes around a reservation price that skews away from your current inventory, trading off the spread you earn against inventory risk over a finite horizon, governed by a risk-aversion parameter γ, volatility σ and order-arrival intensity. As inventory grows long, it quotes lower to offload; as it grows short, higher. Read more →

Inventory risk is the danger that filled quotes leave you holding a directional position that then moves against you, turning a spread-capture business into a bet on price. Managing it means skewing quotes against your inventory (quoting more aggressively on the side that flattens you) so the book mean-reverts toward zero. Avellaneda–Stoikov (2008) formalises exactly this skew. Read more →

Not necessarily, but they change the maths. On maker-taker venues, rebates can flip a thin or negative gross spread into a profitable one, and some strategies exist only to capture them. On crypto and prediction-market venues fee schedules differ; many have no rebate, so the edge must come from spread and signal alone. Whether rebates are essential depends entirely on the venue’s fee model. Read more →

Yes. Crypto is the most accessible market-making arena in 2026. Major exchanges offer direct order entry over public APIs with no prime broker, 24/7 order-driven books, and free L2/L3 data. The microstructure maths transfers directly; what changes is the infrastructure burden (you run your own), fee models, and counterparty and exchange risk. Educational only, not advice. Read more →

Yes, and the same order-book maths applies, but the instrument differs: prediction-market shares have bounded payoffs (0 or 1 at resolution), so fair value is a probability and volatility behaves differently near the bounds and around the resolving event. Books are thinner and event-driven, adverse selection spikes around news, and inventory risk is capped but skewed. A clean, distinct microstructure worth treating on its own terms. Read more →

There is no fixed minimum on open venues (crypto and prediction markets let you quote with modest size) but capital sets your inventory limits, your buffer against adverse-selection runs, and your survivable drawdown, so undercapitalisation is a common failure mode. You need enough to hold inventory through volatility and absorb fee and infrastructure costs. Educational, not a recommendation of any capital level. Read more →

Statistical arbitrage is a market-neutral approach that trades a portfolio of correlated instruments on the expectation that temporary statistical relationships (a spread, a residual, a factor exposure) revert to their historical norm. It bets on the average behaviour of many small, diversified positions rather than any single forecast. Returns come from frequent, low-edge trades whose statistics hold over a large sample. Read more →

Pairs trading is the simplest statistical arbitrage: find two instruments whose prices move together (ideally cointegrated, not merely correlated), and when their spread diverges, short the rich one and buy the cheap one, expecting the spread to revert. Entry and exit are usually triggered by the spread’s z-score. The classic edge is heavily competed in liquid equities; the method remains a clean teaching case and lives on in newer venues. Read more →

As a structural feature, yes; as an open opportunity, mostly no. Picking off stale quotes across venues still happens every day, but in mature lit markets it is a winner-takes-all utility owned by a handful of firms with microwave links and FPGAs; the marginal newcomer cannot win the race. Where it remains contestable is younger, fragmented venues (crypto across exchanges) with looser latency floors. Read more →

Event trading positions around discrete information events: scheduled (earnings, central-bank decisions, economic releases) or unscheduled (breaking news, outages). The edge is reacting to, and pricing in, new information faster or more accurately than the market re-prices. In HFT it overlaps with machine-readable news and the brief, violent microstructure dislocations around an event, where spreads widen and liquidity vanishes. Read more →

News trading reacts to information releases; in HFT it relies on machine-readable feeds: structured economic data, low-latency news wires, and increasingly NLP/LLM parsing of unstructured text. The edge is speed and accuracy of interpretation, not having the news first. In 2026, language models genuinely help off the hot path (classifying, summarising), but the microsecond reaction itself still runs on pre-computed, classical logic. Read more →

In mature lit equities and futures, realistically no: the speed frontier (colo, microwave, FPGA) costs more than an independent can fund, and physics has capped further gains. You compete instead on signal and on venue selection: smarter microstructure models, and trading where the latency arms race hasn’t matured. Speed is table stakes on the old battlefields and not yet decisive on the new ones. Read more →

VWAP (volume-weighted average price) and TWAP (time-weighted average price) are execution schedules that slice a large order into pieces to reduce market impact. TWAP spreads child orders evenly over time; VWAP front- or back-loads them to match the day’s expected volume curve, aiming to fill at the session’s volume-weighted average. Both are benchmarks and algorithms; you are measured against the price they target. Read more →

Almgren–Chriss (2000) is the canonical optimal-execution framework. It frames liquidating a large position as a trade-off between market impact (trade fast, move the price against yourself) and timing risk (trade slow, expose yourself to volatility). Minimising a mean-variance cost yields an optimal trading trajectory; a risk-aversion parameter sets where on the impact-versus-risk frontier you sit. Read more →

Implementation shortfall is the total cost of executing an order, measured as the difference between the price when you decided to trade and the price you actually achieved, including fees, spread, market impact and the opportunity cost of any unfilled portion. Introduced by Perold (1988), it is the honest, all-in execution-quality benchmark, harder to game than VWAP because it counts delay and missed fills. Read more →

A kill switch is an automated, pre-set control that halts a strategy or cancels all its orders when a safety threshold is breached: a loss limit, a position limit, an abnormal order rate, or a data-feed anomaly. After the 2012 Knight Capital loss (~$440m in 45 minutes from a runaway algo), kill switches and pre-trade risk checks became standard and, under regimes like MiFID II, effectively mandatory. Read more →

Spoofing is placing orders you intend to cancel before execution, to create a false impression of supply or demand and move the price; layering is spoofing with multiple orders stacked across price levels. Both are illegal market manipulation, explicitly banned by the US Dodd-Frank Act §747 (2010) and the EU Market Abuse Regulation. This site covers them for recognition and detection only. Read more →

Quote stuffing is flooding a venue with a huge volume of orders and immediate cancellations, with no intent to trade, to congest the market data feed and slow competitors: a denial-of-service attack on the order book. It is treated as manipulative or disruptive conduct under MAR and US rules. We describe it so it can be recognised in the data, not reproduced. Read more →

Yes. Spoofing is explicitly criminalised in the US under the Dodd-Frank Act §747 (2010) and prosecuted by the CFTC, SEC and DOJ, with custodial sentences imposed. In the EU and UK it is prohibited market abuse under MAR (2014). The defining element is the intent to cancel before execution. Read more →

Surveillance looks for the statistical fingerprint of intent: large orders placed away from the touch and cancelled at very high rates just before opposite-side executions, an order-to-trade ratio far above peers, and a recurring pattern of fills following one-sided pressure. Venues, regulators and firms run automated pattern detection over full order-book history. We teach the tell, not the technique. Read more →

Momentum ignition is an illegal strategy that deliberately triggers a rapid price move (often by aggressive orders intended to set off other participants’ stops or trend-following algos) so the manipulator can profit from the cascade it started. It is prohibited disruptive conduct under MAR and US rules. The tell is engineered order flow that initiates, rather than follows, a sharp move. Read more →

Regulation National Market System (Reg NMS, SEC 2005, effective 2007) is the rulebook governing US equity markets. Its core Order Protection Rule (Rule 611) requires orders to execute at the best displayed price across all venues (the NBBO), preventing trade-throughs. By linking fragmented venues into one price grid, it directly created the conditions for cross-venue latency arbitrage and the modern HFT landscape. Read more →

MiFID II with MiFIR is the EU framework in force since January 2018 governing trading venues and conduct. For HFT it adds direct obligations: algorithm registration and testing, mandatory pre-trade risk controls and kill switches, order-to-trade ratio limits, market-maker agreements, and synchronised clocks (microsecond-accurate timestamps). It is the EU’s deliberate regulatory response to the rise of high-frequency trading. Read more →

Circuit breakers are pre-set rules that pause trading when prices move too far too fast, to interrupt disorderly markets and give participants time to react. US equities use single-stock Limit Up-Limit Down (LULD) bands plus market-wide breakers tied to S&P 500 drops (7%, 13%, 20%). They were tightened after the 6 May 2010 Flash Crash. Most major venues, including crypto, now run some equivalent. Read more →

There is no single "HFT regulator"; it is governed by each market’s securities and derivatives authorities. In the US that is the SEC (equities, Reg NMS) and CFTC (futures), with the DOJ prosecuting manipulation. In the EU, ESMA and national regulators enforce MiFID II and MAR; in the UK, the FCA. Crypto and prediction-market oversight in 2026 is still fragmented and venue-dependent. Read more →

Yes. Every concept page, interactive illustration and the concept map are free to read and use, with no paywall and no login. Everything here stays free. Paid infrastructure (datasets, backtest harnesses, reference implementations) may come later, sold separately, never by gating the educational material. As of 2026 those products are pre-launch; only the waitlist exists. Read more →

Infrastructure, not signals. The planned paid products are tools that operationalise what the pages teach: clean, point-in-time L2/L3 limit-order-book datasets, a market-making and stat-arb backtest harness, and reference implementations of named models (Avellaneda–Stoikov, microprice). We sell data and code you’d otherwise have to build, never trade ideas, alerts, or anything with a performance promise. Read more →

It is how to register interest in the paid products before they launch. As of 2026 the datasets, harnesses and reference implementations are designed but not yet shipped, so the only action is joining the waitlist: you give an email against the specific resource a page implies and we notify you when it’s available. No charge, no commitment, no spam. Read more →

HFT Book is published by DT Global Ventures Ltd (CRN 16042336). Content is written for practising and aspiring quants, dated explicitly to 2026, and tied to primary sources: exchange specifications, regulator texts, and named papers with years. Each technique carries an honest status (still alpha / commoditised / dead / structural) and a review date, because microstructure regimes change and stale guidance is worse than none. Read more →

No. HFT Book is educational only. Nothing here is investment, financial, legal or tax advice, a recommendation to trade any instrument, or a promise of any result. Trading high-frequency and quantitative strategies carries substantial risk of loss. Examples and figures illustrate concepts, not opportunities. Do your own research and seek qualified professional advice before risking capital. Read more →