What do I need to start an HFT?

Participants in HFTs must have the following:

• Fast computers: HFT focuses on single-core throughput in most cases, and parallelism is not used by the strategies necessarily.
• Exchange proximity: While some countries restrict the use of shared places to have trading systems and exchange, in the US, we use co-location. This is a place where all the HFTs participants have their production servers. They will pay to have their computers co-located with an exchange's computer servers in the same data centers in order to decrease latency and shorten the time it takes to complete a deal – even by microseconds. The cables linking trading systems from all market participants with the server are the same length to guarantee that nobody has an advantage over another market participant. The SEC has issued a wide request for feedback on co-location fees, as well as other concerns impacting the equity market structure. To ensure fairness among market participants, it is important that co-location fees are reasonably priced. The SEC invites the co-locations to report their fees.
• Low latency: In HFT, latency is the time it takes for data to reach a trader's computer, for the trader to make an order in response to the data, and for the order to be accepted by an exchange. The order may enter the market alongside many other orders issued by other traders at the most profitable time. There is a danger of competing against a large number of other people in this circumstance. The order may not be as profitable as it may have been in this scenario. High-frequency traders are able to make orders at unfathomably quick speeds because of technology advertised as low-latency or ultra-low-latency. It is important to use gear designed to reduce the latency of shuffling data from one place to another.
• Computer algorithms, which are at the heart of AT and HFT, and real-time data feeds, which could damage earnings.

In the previous sections, we learned where high-frequency traders make business. We also talked about the technological prerequisites to trade faster. Let's now focus on what HFT is in depth.

What are HFT strategies?

HFT strategies are a subset of algorithmic trading strategies. They are executed in the order of the microseconds (and sometimes nanoseconds). The strategies must be aware of this time limitation to be efficient HFT strategies. They deploy cutting-edge technology advancements to obtain information faster than the competition. The main game of this type of strategy is the tick-to-trade, which is the response time to send an order responding to incoming market data. As we will explain in the next chapter, it is important to host trading strategies on cutting-edge machines, and they must also run in a co-location.

We will be defining the domain of applications and some vocab to have when talking about HFT strategies.

Asset classes

HFT strategies can be applied to any asset classes, such as stocks, futures, bonds, options, and FX. We also have cryptocurrencies being traded using HFT strategies, even if the definition of speed is different (because of the settlement time).

Liquidity

The desire of players to interact with regard to a certain asset is known as liquidity.

We define depth as the number of price levels for a given asset. We will say that a book is deep when there are many levels (layers) for a given asset. We will define a book as big or broad if the volume per layer is high. If a book is deep or large, we will define a liquidity of a given asset liquid. The consequence of this statement is that it will be easier for a trader to buy or sell this asset whenever they want to. As a result, trading exchanges with a lot of liquidities are wanted by traders. Crypto trading exchanges have difficulty finding liquidities at the moment.

Tick-by-tick data and data distribution

HFT generates orders every microsecond. Since there are a lot of participants, it is likely to have huge amounts of data. Storage of this data will be key when we study HFT data to create models for trading strategies.

Thousands of ticks (security price changes from one order to another) are generated per trading day on liquid marketplaces, which make up high-frequency data. This material is randomly spaced in time by its very nature. HFT data exhibits fat tail distributions. That means that the trading strategies need to take into account that we can have big losses.

They distribution of the market data can be grouped into two categories:

• Volatility clustering: Large changes follow large changes whether in terms of signs or numbers, while minor changes follow smaller changes.
• Long-range dependency (long memory) refers to the pace at which statistical dependence between two sites decays as the time interval or spatial distance between them increases.

Liquidity rebates

To support the provision of stock liquidity, most exchanges have used a maker-taker model. In this arrangement, investors and traders who place limit orders often earn a modest rebate from the exchange when their orders are executed since they are considered to have contributed to stock liquidity, or makers.

Those who place market orders, on the other hand, are considered takers of liquidity and are charged a small fee by the exchange. While the rebates are normally fractions of a penny per share, over the millions of shares exchanged daily by high-frequency traders, they may add up to large amounts. Many HFT businesses use trading techniques that are geared to take advantage of as many liquidity rebates as feasible.

Matching engine

The software program that forms the heart of an exchange's trading system and matches buy and sell orders on a continuous basis, a service traditionally done by trading floor professionals, called the matching engine is critical for guaranteeing the efficient operation of an exchange since it matches buyers and sellers for all stocks. The matching engine is housed on the exchange's computers, and it is the main reason why HFT businesses strive to be as near to the exchange servers as possible. We will learn about it in Chapter 3, Understanding the Trading Exchange Dynamics.

Market making

Before going into details on what market making is, we need to explain the difference between market takers and market makers.

Market taker/maker

Figure 1.1 represents the limit order book on an exchange. When a trading strategy places an order close to the top of the book (the layer representing the best price for bid and for ask), we say that this order is an aggressive order. It means that this order is likely to be matched with another order. If the order is executed, it means that liquidity has been removed from the market; it is a market taker. We will say that a trader crosses the spread when they place a buy order at the price of the ask on the top of the book. If the order is less aggressive (or passive), this order will not remove liquidities from the market; it is a market maker.

Figure 1.1 – Order book – passive/aggressive order

Let's look at the market-making strategies.

Market-making strategies

A trading corporation can provide market-making as a service on an exchange. Over time, a market maker assists in the matching of buyers and sellers. Rather than purchasing or selling securities based on their underlying assets, market makers maintain a continual offer to buy and sell securities and profit from the spread, which is the difference between the two offers.

To reduce the risk of keeping stocks for extended periods of time, every purchase should be matched with a sale and every sell should be matched with a buy. If a stock is trading at $100, a market maker can keep a buy offer at $99.50 and a sell offer at $100.50. If they are successful in finding both a buyer and a seller, it allows those who want to sell right now to do so even if no one else wants to purchase, and vice versa.

Market makers, in other words, supply liquidity—they make trading simpler. For the most traded stocks, this technique is not important; however, for smaller firms (less traded than the big ones), it can be critical to increase the trading volume to facilitate trading. Market making is one approach that many HFT businesses use. They out-compete everyone else by changing their quotations quickly and reducing the spread even further: they're willing to make less money each time since their market-making business can readily grow to massive quantities. However, an HFT firm's technology can be used for other purposes, such as arbitrage (making money on minor discrepancies between linked securities) or execution (breaking up huge institutions' trades to minimize market effect). I won't go into much more detail because the point is that HFT is capable of more than simply market-making. The only thing that matters is speed.

Market making can be done by the analysis of the order flow:

• A large volume of buy and sell can drive the market price of buying and selling on the basis of momentum.
• The flow of liquids (how big are the buy and sell orders: small, medium, or big).
• Exhaustion of momentum (when the order flow is drying off it may signal a price reversal).

Market-making is the most widely used trading strategy for high-frequency traders. We will talk about the other HFT strategies in the next sections.

Scalping

Scalping is a trading method that focuses on benefitting from tiny price movements and reselling for a quick profit. Scalping is a phrase used in day trading to describe a technique that focuses on generating large volumes from tiny profits. Scalping necessitates a tight exit plan since a single major loss might wipe out all of the modest wins the trader has worked so hard to achieve. For this technique to work, you'll need the necessary tools, such as a live feed, a direct-access broker, and the endurance to conduct a lot of trades.

The concept behind scalping is that most stocks will finish the first stage of a trend. But it's unclear where things will go from there. Some stocks stop rising after that early stage, while others continue to rise. The goal is to benefit from as many minor transactions as possible. The let your gains run mentality, on the other hand, aims to maximize good trading results by expanding the size of winning deals. By increasing the number of winners while compromising on the magnitude of the gains, this technique accomplishes outcomes. It's very uncommon for a trader with a long time period to produce good profits while winning just 50% of their transactions, or even less – the difference is that the wins are far larger than the losses.

Statistical arbitrage

The Efficient Market Hypothesis (EMH) claims financial markets are informationally efficient, which means that the prices of traded assets are accurate, and at any one moment represent all known information. Based on this hypothesis, the market should not fluctuate if there is not any fundamental news. However, this is not the case, and we can explain that with liquidity.

Throughout the day, many huge institutional trades have little to do with information and everything to do with liquidity. Investors who believe they are overexposed will aggressively hedge or sell their positions, impacting the price. Liquidity seekers are frequently ready to pay a premium to exit their positions, resulting in a profit for liquidity providers. Although this capacity to benefit from knowledge appears to violate efficient market theory, statistical arbitrage is based on it.

Statistical arbitrage seeks to profit from the correlation of price and liquidity by gaining from the perceived mispricing of assets based on the assets' anticipated value given by a statistical model.

Short-term price discrepancies in the same security sold on separate venues, or short-term price differences in related securities, are used in statistical arbitrage, often known as stat arb. Statistical arbitrage is based on the assumption that price differences in securities markets exist but go away quickly. Because the time period during which a price difference occurs might be as short as a fraction of a second, algorithmic trading is well suited to statistical arbitrage.

When trading the same security in several venues, for example, an algorithm tracks all of the locations where the security is exchanged. When a price difference arises, the algorithm buys in the lower market and sells in the higher market, resulting in a profit. Because the window of opportunity for such differences is small (less than 1 millisecond), algorithmic trading is ideally suited to this form of trade.

Statistical arbitrage becomes more challenging when investing in linked securities. An index and a single stock within that index, or a single stock and other stocks in the same sector, are examples of related securities. In linked securities, a statistical arbitrage approach entails gathering a large amount of historical data and estimating the usual connection between the two markets. The algorithm makes a buy or a sell whenever there is a variation from the norm.

Latency arbitrage

Modern equities markets are complicated, requiring highly technical systems to manage vast volumes of data. Because of its intricacy, data is invariably processed at varying speeds. Latency arbitrage takes advantage of market players' differing speeds. Latency arbitrage aims to take advantage of high-frequency traders' greater speed by leveraging high-speed fiber optics, superior bandwidth, co-located servers, and direct-price feeds from exchanges, among other things, to place trades ahead of other market players.

The hypothesis behind latency arbitrage is that in the US, the aggregated feed that determines the National Best Bid and Offer (NBBO) of all US stock exchanges is slower than the direct data feeds from stock exchanges available to high-frequency traders. An HFT program's algorithm can read transaction data more quickly than many other market players, seeing prices a fraction of a second ahead of the Securities Information Processor (SIP) feed, which is the consolidated US stock exchange price feed, thanks to its superior speed. This essentially provides information to the HFT software before it reaches the official market (the SIP feed), allowing high-frequency traders to observe where prices are heading ahead of other market players.

Impact of news

Information is at the heart of all trading, and it is used to make financial decisions. The utilization of news data by algorithmic trading systems to generate trading choices is referred to as information-driven strategies.

Algorithms have been developed to read and analyze news reports from major news organizations, as well as social media. Any news that has the potential to alter market prices causes the algorithm to purchase or sell.

High-frequency traders have gotten so accustomed to using information-driven methods that certain news agencies now package their press releases in a way that makes it simple for computers to analyze them. They employ predetermined keywords to characterize a favorable or bad occurrence, for example, so that an algorithm can act on keywords in a news release. Prior to their planned publication, news providers also place news reports on servers in crucial geographical regions (such as major financial centers). This reduces the amount of time it takes for data to move from one location to another. For this sort of service, news service providers charge an additional fee.

As seen by the hacking of the Associated Press Twitter feed, the use of social media for information-driven initiatives is growing. In 2013, a hacker tweeted that a bomb had gone off in the White House, injuring the president, causing an instantaneous plunge in equities markets throughout the world as algorithms analyzed the bad news from a trustworthy source and began selling in the market.

Next, let's learn about the momentum ignition trading technique.

Momentum ignition

You have the chance to trade financially if an order you send into the market may cause a price change and you know it can. The goal of momentum-ignition trading techniques is to achieve this. The objective is to get other algorithms and traders to start trading in a stock, causing a price change. In essence, a momentum ignition approach attempts to deceive other market players into believing that a large price movement is going to occur, causing them to trade. As a result, the price movement becomes a self-fulfilling prophecy: traders believe a price movement will occur, and their activities cause one to occur.

Sending enormous volumes of orders into the order book and then canceling them is a momentum-ignition approach. This creates the illusion of a huge shift in volume in the stock, which might prompt other traders to place orders, resulting in the start of a short-term price trend. Before attempting to ignite the market movement, the momentum ignition approach includes executing the real targeted trading position. This means that a deal is completed initially that does not significantly influence the market. This permits a trader using the momentum-ignition approach to enter the market before the price movement is initiated. The momentum ignition is set after the deal is completed by submitting a flurry of orders and canceling them in the hopes that other traders will follow suit and move the price.

The trader using the momentum-ignition technique then quits their initial position at a profit as the price begins to move.

Momentum-ignition methods require the use of specific order types, and traders may only utilize algorithms that can send and cancel huge numbers of orders in a short amount of time to execute them.

Rebate strategies

Market order traders must pay a fee to the exchange, whereas the limit order is reimbursed with rebates when they add liquidities. As a result, traders, particularly those engaged in HFT, submit limit orders to build markets, which in turn generates liquidity on the exchange. It is undoubtedly appealing to traders who place a large number of limit orders due to the pricing scheme's lower risk for the limit order.

There is also a charge structure called trader-maker pricing that is the polar opposite of market-taker pricing. In certain markets, it entails giving rebates to market order traders and collecting fees from limit order traders.

Pinging

Pinging is a strategy for learning about huge orders in trading exchanges and dark pools by placing tiny marketable orders (typically for 100 shares).

To lessen the market effect of large orders, buy-side businesses utilize this trading technique to split large orders into many small orders. This algorithm feeds these orders slowly into the exchange. In order to detect the presence of such large orders, HFT companies arrange bids and offers in 100-share lots for each listed stock.

These ping trades will alert HFT participants to the existence of a large order placed by the buy-side. HFTs will use this information to ensure risk-free profit from the buy-side.

Some significant market participants have compared pinging to baiting because its main objective is to entice institutions with huge orders to expose their hand.

Illegal activities

The SEC, the Federal Bureau of Investigation (FBI), and other regulators have launched crackdowns on alleged HFT violations in recent years. The following sections are examples of possible offenses.

Front-running

Placing an order based on information that has not been publicly released is called front-running. This technique has been outlawed by SEC and the Financial Industry Regulatory Authority (FINRA). Some have used the term front-running to describe a technique in which HFT firms utilize algorithmic trading technology to identify a large number of new orders for a given instrument. Before the large number of orders comes to the market, we place orders to benefit from this incoming large quantity. HFT corporations can earn almost instantly after purchasing assets by selling them to the original investors. Even if this way of trading is legal, regulators are concerned and may need to control this behavior moving forward.

Spoofing

Spoofing is not a legal trading strategy. It consists of a spoofing strategy sending orders that are not intended to be executed, just to have the other market participants react to these orders. They will probably send orders to get to this price level. Meanwhile, the initial orders are canceled and the spoofer takes advantage of the other orders remaining in the market.

The Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010 specifically targeted the practice, and even before that, FINRA regulations barred orders whose goal is to mislead the market. The first criminal spoofing case disclosed by legislators in 2014 related to a Chicago trader accused of faking futures markets.

Layering

Layering is the same as spoofing except that the orders are placed at different price levels to give the appearance that there is a lot of interest in a certain security. The outcome of this strategy is the same as with regular spoofing. Because of the rapid advancement of technology, massive market manipulation may take place in fractions of a second. Layering, like generic spoofing, is typically illegal and forbidden under FINRA rules.

Even if these strategies are now outlawed, we need to keep in mind that some exchanges are less or not regulated. We will see in Chapter 11, High Frequency FPGA and Crypto, about cryptocurrencies that these strategies can still work.